identification, manufacture and use of neoantigens

专利摘要:
A system and methods for determining alleles, neoantigens and the composition of the vaccine are disclosed here, as determined based on an individual's tumor mutations. Systems and methods for obtaining high-quality sequencing data for a tumor are also disclosed. In addition, systems and methods for identifying somatic changes in polymorphic genome data are described here. Finally, here described are exclusive vaccines against cancer.
公开号:BR112019021782A2
申请号:R112019021782-7
申请日:2018-04-19
公开日:2020-08-18
发明作者:Thomas Boucher；Brendan Bulik-Sullivan；Jennifer Busby；Roman Yelensky
申请人:Gritstone Oncology, Inc.；
IPC主号:

专利说明:

[001] [001] Therapeutic vaccines based on tumor-specific neoantigens hold great promise as a next generation of personalized cancer immunotherapy. ' Cancers with a high mutational burden, such as non-small cell lung cancer (NSCLC) and melanoma, are particularly attractive targets for this therapy, given the relatively greater likelihood of generating neoantigens. ** Initial evidence shows that vaccination based on neoantigens can trigger T cell responses and that cell therapy targeting neoantigens can cause tumor regression under certain circumstances in selected patients. * Both MHC class I and MHC class II impact the responses of T7º cells ”!,
[002] [002] A question for the design of the neoantigen vaccine is which of the many coding mutations present in the subject tumors can generate the “best” therapeutic neoantigens, for example, antigens that can cause antitumor immunity and cause tumor regression.
[003] [003] Initial methods have been proposed incorporating mutation-based analysis using next generation sequencing, expression of RNA genes and prediction of MHC binding affinity of candidate neoantigen peptides. * However, these proposed methods may fail to model the entire epitope generation process, which contains many steps (for example, TAP transport, proteasome cleavage, MHC binding, transport of the MHC-peptide complex to the cell surface and / or TCR recognition for MHC-I; endocytosis or autophagy, cleavage by extracellular or lysosomal proteases (eg cathepsins), competition with the CLIP peptide for HLA-DM-catalyzed HLA binding, transport of the MHC-peptide complex to the surface and / or recognition and TCR for MHC-II), in addition to gene expression and MHC binding ”. Consequently, existing methods are likely to suffer from a low low positive predictive value (PPV). (Figure 1A)
[004] [004] Indeed, analyzes of peptides presented by tumor cells carried out by various groups have shown that <5% of the peptides that are expected to be presented using gene expression and MHC binding affinity can be found on the surface of the MHC tumor '*!' (Figure 1B). This low correlation between binding prediction and MHC presentation was further reinforced by recent observations of the lack of improvement in predictive accuracy of neoantigens with binding restriction for the checkpoint inhibitor response on the number of isolated mutations. !
[005] [005] This low positive predictive value (PPV) of existing methods for predicting presentation presents a problem for the design of vaccines based on neoantigens. If vaccines are designed using low PPV predictions, it is unlikely that most patients will receive a therapeutic neoantigen and even less will receive more than one (even assuming that all of the peptides presented are immunogenic). Thus, it is unlikely that vaccination of neoantigens with current methods will be successful in a substantial number of subjects with tumors. (Figure 1C)
[006] [006] In addition, previous approaches have generated candidate neoantigens using only cis-action mutations and have been largely neglected to consider additional sources of neo-ORFs, including mutations in junction factors, which occur in various types of tumors and lead to aberrant junction of many genes " and mutations that create or remove protease cleavage sites.
[007] [007] Finally, standard approaches for tumor genome and transcriptome analysis may lose somatic mutations that give rise to candidate neoantigens due to sub-ideal conditions in the construction of libraries, exome and transcriptome capture, sequencing or data analysis. Likewise, standard approaches to tumor analysis may inadvertently promote sequence artifacts or germline polymorphisms as neoantigens, leading to inefficient use of vaccine capacity or the risk of autoimmunity, respectively. SUMMARY
[008] [008] An optimized approach to identify and select neoantigens for personalized cancer vaccines is disclosed here. First, optimized approaches to tumor exome and transcriptome analysis for identifying neoantigen candidates using next generation sequencing (NGS). These methods are based on standard approaches for the analysis of NGS tumors to ensure that candidates for neoantigens of greater sensitivity and specificity are advanced in all classes of genomic alteration. Second, new approaches for the selection of neoantigens with high PPV are presented to overcome the specificity problem and ensure that advanced neoantigens for inclusion in vaccines are more likely to elicit antitumor immunity. These approaches include, depending on the modality, trained statistical regression or non-linear deep learning models that jointly model mappings of peptide alleles, as well as motifs by allele for peptides of multiple lengths, sharing the statistical strength between peptides of different lengths. Non-linear models of deep learning, in particular, can be designed and trained to treat different MHC alleles in the same cell as independent, addressing problems with linear models that would have interfered with each other. Finally, additional considerations for the personalized design and manufacture of vaccines based on neoantigens are addressed. BRIEF DESCRIPTION OF THE VARIOUS VIEWS OF THE FIGURES
[009] [009] These and other characteristics, aspects and advantages of the present invention will be better understood in relation to the following description and attached figures, where:
[010] [010] Figure (FIG.) 1 A shows the current clinical approaches for identifying neoantigens.
[011] [011] FIG. 1B shows that &lt; 5% of the predicted bound peptides are presented in the tumor cells.
[012] [012] FIG. 1C shows the impact of the neoantigens prediction specificity problem.
[013] [013] FIG. 1D shows that the prediction of binding is not sufficient for the identification of neoantigens.
[014] [014] FIG. 1E shows probability of presenting MHC-I as a function of peptide length.
[015] [015] FIG. 1F shows an example of a peptide spectrum generated from Promega's dynamic range pattern. The figure discloses SEQ ID NO: 1.
[016] [016] FIG. 1G shows how adding features increases the positive predictive value of the model.
[017] [017] FIG. 2A is an overview of an environment to identify probabilities of presentation of peptides in patients, according to one modality.
[018] [018] FIG. 2B and 2C illustrate a method for obtaining presentation information, according to a modality. Figure 2B discloses SEQ ID NO: 3. Figure 2C discloses SEQ ID NOS 3-8, respectively, in order of appearance.
[019] [019] FIG. 3 is a high-level block diagram that illustrates the computer's logical components of the presentation identification system, according to a modality.
[020] [020] FIG. 4 illustrates an example of a training data set, according to a modality. The Figure discloses the "Peptide Sequences" as SEQ ID NOS 10-13 and the "Flanking C Sequences" as SEQ ID NOS 14, 19-20 and 20, respectively, in order of appearance.
[021] [021] FIG. 5 illustrates an example of a network model in association with an MHC allele. FIG. 6A illustrates an example of an NNH (-) network model shared by MHC alleles, according to one modality.
[022] [022] FIG. 6B illustrates an example of a NNuH (-) network model shared by MHC alleles, according to another modality.
[023] [023] FIG. 7 illustrates the generation of a presentation probability for a peptide in association with an MHC allele using an example network model.
[024] [024] FIG. 8 illustrates the generation of a presentation probability for a peptide in association with an MHC allele using example network models.
[025] [025] FIG. 9 illustrates the generation of a presentation probability for a peptide in association with MHC alleles using example network models.
[026] [026] FIG. 10 illustrates the generation of a presentation probability for a peptide in association with MHC alleles using example network models.
[027] [027] FIG. 11 illustrates the generation of a presentation probability for a peptide in association with MHC alleles using example network models.
[028] [028] FIG. 12 illustrates the generation of a presentation probability for a peptide in association with MHC alleles using example network models.
[029] [029] FIG. 13A is a histogram of peptide lengths eluted from MHC class II alleles in human tumor cells and tumor infiltrating lymphocytes (TIL) using mass spectrometry.
[030] [030] FIG. 13B illustrates the dependency between mRNA quantification and the peptides presented per residue for two sample data sets.
[031] [031] FIG. 13C compares performance results, for example, trained and tested presentation models using two sets of sample data.
[032] [032] FIG. 13D is a histogram that describes the amount of peptides sequenced using mass spectrometry for each sample out of a total of 39 samples comprising HLA class II molecules.
[033] [033] FIG. 13E is a histogram that describes the number of samples in which a specific MHC class II molecule allele has been identified.
[034] [034] FIG. 13F is a histogram that describes the proportion of peptides presented by MHC class II molecules in the 39 total samples, for each peptide length in a range of peptide lengths.
[035] [035] FIG. 13G is a line graph that represents the relationship between gene expression and the prevalence of presentation of the gene expression product by a class II MHC molecule, for genes present in the 39 samples.
[036] [036] FIG. 13H is a line graph that compares the performance of identical models with variable inputs, to predict the likelihood that peptides in a peptide test data set will be presented by a class II MHC molecule.
[037] [037] FIG. 131 is a line graph that compares the performance of four different models in predicting the likelihood that peptides in a peptide test data set will be presented by a class II MHC molecule.
[038] [038] FIG. 13J is a line graph that compares the performance of a prior art model of the best class, using two different criteria and the presentation model disclosed here with two different inputs, to predict the likelihood that peptides in a test data set of peptides are presented by a class II MHC molecule.
[039] [039] FIG. 14 illustrates an example computer for implementing the entities shown in FIGS. 1e3.
[040] [040] In general, the terms used in the claims and in the specification should be understood as having the clear meaning understood by someone skilled in the art.
[041] [041] As used in this document, the term "antigen" is a substance that induces an immune response.
[042] [042] As used in this document, the term “neoantigen” is an antigen that has at least one alteration that distinguishes it from the corresponding parental antigen of the wild type, for example, via mutation in a tumor cell or specific post-translational modification tumor cell. A neoantigen may include a polypeptide sequence or a nucleotide sequence. A mutation can include a frame shift or no frame shift, missense or nonsense, change at the junction site, genomic rearrangement or gene fusion, or any genomic or expression change that gives rise to a neoORF. A mutation can also include a splice variant. Post-translational modifications specific to a tumor cell can include aberrant phosphorylation. Post-translational modifications specific to a tumor cell may also include a spliced antigen generated by a proteasome. See Liepe et al., A large fraction of HLA class I ligands are proteasome-generated spliced peptides; Science. October 21, 2016; 354 (6310): 354-358.
[043] [043] As used in this document, the term "tumor neoantigen" is a neoantigen present in a subject's tumor cell or tissue, but not in the subject's corresponding normal cell or tissue.
[044] [044] As used in this document, the term "neoantigen-based vaccine" is a vaccine construct based on one or more neoantigens, for example, a plurality of neoantigens.
[045] [045] As used in this document, the term “candidate for neoantigen” is a mutation or other aberration that gives rise to a new sequence that may represent a neoantigen.
[046] [046] As used in this document, the term "coding region" is the portion (s) of a gene that encodes the protein.
[047] [047] As used in this document, the term "coding mutation" is a mutation that occurs in a coding region.
[048] [048] As used in this document, the term “ORF” means open reading frame.
[049] [049] As used in this document, the term "NEO-ORF" is a tumor-specific ORF resulting from a mutation or other aberration, such as junctions.
[050] [050] As used in this document, the term “missense mutation” is a mutation that causes a substitution from one amino acid to another.
[051] [051] As used in this document, the term "meaningless mutation" is a mutation that causes a substitution of an amino acid for a stop codon.
[052] [052] As used in this document, the term "frame shift mutation" is a mutation that causes a change in the protein's frame.
[053] [053] As used in this document, the term "indel" is an insertion or exclusion of one or more nucleic acids.
[054] [054] As used in this document, the term "identity" percentage, in the context of two or more nucleic acid or polypeptide sequences, refers to two or more sequences or subsequences that have a specified percentage of nucleotides or amino acid residues equal when compared and aligned for maximum matching, as measured using one of the sequence comparison algorithms described below (for example, BLASTP and BLASTN or other algorithms available to qualified persons) or by visual inspection. Depending on the application, the “identity” percentage may exist in a region of the sequence being compared, for example, in a functional domain or, alternatively, in the entire length of the two sequences to be compared.
[055] [055] For sequence comparison, typically a sequence acts as a reference sequence, with which the test sequences are compared. When using a sequence comparison algorithm, the test and reference sequences are entered into a computer, subsequence coordinates are designated, if necessary, and sequence algorithm program parameters are designated. The sequence comparison algorithm then calculates the percentage of sequence identity for the test sequence (s) relative to the reference sequence, based on the designated program parameters. Alternatively, sequence similarity or dissimilarity can be established by the combined presence or absence of specific nucleotides or, for translated sequences, amino acids at selected sequence positions (for example, sequence motifs).
[056] [056] The ideal sequence alignment for comparison can be accomplished, for example, by the local homology algorithm of Smith & Waterman, Adv. Appl. Math. 2: 482 (1981), by Needleman & Wunsch's homology alignment algorithm, J. Mol. Biol. 48: 443 (1970), by the similarity search method by Pearson & Lipman, Proc. Nat'l. Acad. Sci. USA 85: 2444 (1988), by computerized implementations of these algorithms (GAP, BESTFIT, FASTA, and TFASTA in Wisconsin Genetics Software Package, Genetics Computer Group, 575 Science Dr., Madison, WD), or by visual inspection (see , in general Ausubel et al., infra).
[057] [057] An example of an algorithm that is suitable for determining the percentage of identity and sequence similarity is the BLAST algorithm, described in Altschul et al., J. Mol. Biol. 215: 403-410 (1990). The BLAST analysis program is publicly available through the National Biotechnology Information Center.
[058] [058] As used in this document, the term “without interruption or reading” is a mutation that causes the removal of the natural stop codon.
[059] [059] As used in this document, the term "epitope" is the specific portion of an antigen typically bound to an antibody or T cell receptor.
[060] [060] As used in this document, the term "immunogenic" is the ability to obtain an immune response, for example, via T cells, B cells or both.
[061] [061] As used in this document, the term "HLA binding affinity" "MHC binding affinity" means binding affinity between a specific antigen and a specific MHC allele.
[062] [062] As used in this document, the term "bait" is a nucleic acid probe used to enrich a specific DNA or RNA sequence from a sample.
[063] [063] As used in this document, the term "variant" is a difference between a subject's nucleic acids and the reference human genome used as a control.
[064] [064] As used in this document, the term "variant call" is an algorithmic determination of the presence of a variant, typically from sequencing.
[065] [065] As used in this document, the term "polymorphism" is a variant of the germline, that is, a variant found in all of an individual's DNA-carrying cells.
[066] [066] As used in this document, the term “somatic variant” is a variant that appears in cells of an individual's non-germ line.
[067] [067] As used in this document, the term "allele" is a version of a gene or a version of a genetic sequence or a version of a protein.
[068] [068] As used in this document, the term "HLA type" is the complement of the HLA gene alleles.
[069] [069] As used in this document, the term "meaningless mediated decay" or "NMD" is a degradation of an mRNA by a cell due to a premature stop codon.
[070] [070] As used in this document, the term "truncal mutation" is a mutation that originates in the early development of a tumor and is present in a substantial portion of the tumor cells.
[071] [071] As used in this document, the term "subclonal mutation" is a mutation that originated later in the development of a tumor and is present in only a subset of the tumor cells.
[072] [072] As used in this document, the term "exome" is a subset of the genome that encodes proteins. An exome can be the collective exon of a genome.
[073] [073] As used in this document, the term “logistic regression” is a regression model for binary data from statistics, where the logit of the probability that the dependent variable is equal to one is modeled as a linear function of the dependent variables .
[074] [074] As used in this document, the term “neural network” is a machine learning model for classification or regression that consists of several layers of linear transformations followed by hand linearities by elements normally trained by stochastic gradient descent and rear propagation.
[075] [075] As used in this document, the term "proteome" is the set of all proteins expressed and / or translated by a cell, group of cells or individual.
[076] [076] As used in this document, the term "peptidoma" is the set of all peptides presented by MHC-I or MHC-II on the cell surface. The peptidoma can refer to a property of a cell or to a collection of cells (for example, the tumor peptidoma, meaning the union of the peptidomas of all the cells that make up the tumor).
[077] [077] As used in this document, the term "ELISPOT" means enzyme-linked immunosorbent assay - which is a common method for monitoring immune responses in humans and animals.
[078] [078] As used in this document, the term "dextramers" is a dextran-based MHC peptide multimer used for staining antigen-specific T cells in flow cytometry.
[079] [079] As used in this document, the term "immune tolerance or tolerance" is a state of immune non-response to one or more antigens, for example, autoantigens.
[080] [080] As used in this document, the term “central tolerance” is an affected tolerance in the thymus, either by excluding self-reactive T cell clones or by promoting self-reactive T cell clones to differentiate immunosuppressive regulatory T cells (Tregs).
[081] [081] As used in this document, the term "peripheral tolerance" is an affected peripheral tolerance by overregulating or anergizing autoreactive T cells that survive central tolerance or promoting these T cells to differentiate into Tregs.
[082] [082] The term “sample” can include a single cell or multiple cells or cell fragments or an aliquot of body fluid, taken from a subject, by means including venipuncture, excretion, ejaculation, massage, biopsy, needle aspiration , washing sample, scraping, surgical incision or intervention or other means known in the art.
[083] [083] The term "subject" encompasses a cell, tissue or organism, human or non-human, whether in vivo, ex vivo or in vitro, male or female. The term subject includes mammals, including humans.
[084] [084] The term “mammal” encompasses humans and non-humans and includes, but is not limited to, humans, non-human primates, canines, felines, murines, cattle, horses and pigs.
[085] [085] The term "clinical factor" refers to a measure of a subject's condition, for example, disease activity or severity. The “clinical factor” covers all markers of a subject's health status, including non-sample markers and / or other characteristics of a subject, such as, without limitation, age and sex. A clinical factor can be a score, a value, or a set of values that can be obtained by evaluating a sample (or sample population) of a subject or subject under a given condition. A clinical factor can also be predicted by markers and / or other parameters, as substitutes for gene expression. Clinical factors may include tumor type, tumor subtype and smoking history.
[086] [086] Abbreviations: MHC: major histocompatibility complex; HLA: human leukocyte antigen, or the human MHC gene locus; NGS: next generation sequencing; PPV: positive predictive value; TSNA: tumor-specific neoantigen; FFPE: fixed in formalin, embedded in paraffin; NMD: decay mediated by nonsense; NSCLC: non-small cell lung cancer; DC: dendritic cell.
[087] [087] It should be noted that, as used in this specification and the appended claims, the singular forms "one" "one" and "o / a" include plural referents, unless the context clearly indicates otherwise.
[088] [088] Any terms not defined directly in this document should be understood as having the meanings commonly associated with them, as understood in the technique of the invention. Certain terms are discussed here to provide additional guidance to the practitioner in describing the compositions, devices, methods and the like of aspects of the invention, and how to make or use them. It will be appreciated that the same thing can be said in more than one way. Consequently, the alternative language and synonyms can be used for any one or more of the terms discussed in this document. It should not be stressed whether a term is elaborated or discussed here. Some synonyms or replaceable methods, materials and the like are provided. The recital of one or a few synonyms or equivalents does not exclude the use of other synonyms or equivalents, unless explicitly stated. The use of examples, including example terms, is for illustrative purposes only and does not limit the scope and meaning of the aspects of the invention described herein.
[089] [089] All references, issued patents and patent applications cited in the body of the specification are incorporated by reference in their entirety, for all purposes.
[090] [090] Methods for identifying neoantigens of a tumor of a subject that are likely to be presented on the tumor cell surface or immune cells, including professional antigen presenting cells, such as dendritic cells, and / or probably immunogenic cells, are disclosed here. As an example, one of these methods may comprise the steps of: obtaining at least one of the tumor nucleotide sequencing data from the exome, transcriptome or entire genome from the subject's tumor cell, where the tumor nucleotide sequencing data is used to obtain data representing peptide sequences for each of a set of neoantigens, and wherein the peptide sequence for each neoantigen comprises at least one alteration that distinguishes it from the corresponding wild-type parental peptide sequence; introduce the peptide sequence of each neoantigen into one or more presentation models to generate a set of numerical probabilities that each neoantigen is presented by one or more MHC alleles on the surface of the tumor cell of the subject's tumor cell or cells present in the tumor, the set of numerical probabilities having been identified at least based on the data received from mass spectrometry; and select a subset of the set of neoantigens based on the set of numerical probabilities to generate a set of selected neoantigens.
[091] [091] The presentation model can comprise a statistical regression model or machine learning (for example, deep learning) trained on a reference data set (also known as training data set) that comprises a set of labels corresponding, where the reference data set is obtained from each of a plurality of distinct subjects, where optionally some subjects may have a tumor, and where the reference data set comprises at least one among: data representing nucleotide sequences tumor tissue exome, data representing nucleotide sequences of the normal tissue exome, data representing tumor tissue transcriptome nucleotide sequences, data representing tumor tissue proteomic sequences and data representing MHC peptide sequences of tumor tissue and data representing MHC peptide sequences of normal tissue. The reference data may further comprise mass spectrometry data, sequencing data, RNA sequencing data and proteomic data for single allele cell lines designed to express a predetermined MHC allele that is subsequently exposed to synthetic proteins, human cell lines normal and tumor and fresh and frozen primary samples and T cell assays (eg ELISPOT). In some respects, the reference data set includes each form of reference data.
[092] [092] The presentation model may comprise a set of characteristics derived, at least in part, from the reference data set and where the set of characteristics comprises at least one of the allele-dependent and allele-independent characteristics. In certain respects, each feature is included.
[093] [093] Also disclosed here are methods for generating an outlet for the construction of a personalized cancer vaccine, identifying one or more neoantigens from one or more tumor cells of a subject that are likely to be presented on a tumor cell surface. As an example, one of these methods may comprise the steps of: obtaining at least one of the exome, transcriptome, or entire genome nucleotide sequencing data from the subject's tumor cells and normal cells, where the nucleotide sequencing data is used to obtain data representing peptide sequences for each of a set of neoantigens identified by comparing the nucleotide sequencing data for tumor cells and the nucleotide sequencing data for normal cells and where the peptide sequence for each neoantigen comprises at least one change that distinguishes it from the corresponding wild-type peptide sequence identified from the subject's normal cells; encode the peptide sequences of each of the neoantigens in a corresponding numeric vector, each numeric vector including information about a plurality of amino acids that make up the peptide sequence and a set of positions of the amino acids in the peptide sequence; introduce numerical vectors, using a computer processor, into a profound learning presentation model to generate a set of presentation probabilities for the set of neoantigens, each presentation probability in the set representing the probability that a corresponding neoantigen will be presented by a or more class II MHC alleles on the surface of the subject's tumor cells, the deep learning presentation model; select a subset of the set of neoantigens based on the set of presentation probabilities to generate a set of selected neoantigens; and generate the way for the construction of a personalized cancer vaccine based on the set of selected neoantigens.
[094] [094] In some modalities, the presentation model comprises a plurality of parameters identified at least based on a set of training data and a function that represents a relationship between the numeric vector received as input and the probability of presentation generated as output based on the numeric vector and the parameters. In certain embodiments, the training data set comprises markers obtained by mass spectrometry that measures the presence of peptides linked to at least one class II MHC allele identified as present in at least one of a plurality of samples, peptide sequences training codes encoded as numeric vectors, including information about a plurality of amino acids that make up the peptide sequence and a set of amino acid positions in the peptide sequence and at least one HLA allele associated with the training peptide sequences.
[095] [095] The presentation of dendritic cells to characteristics of naive T cells may comprise at least one of the following: A characteristic described above. The dose and type of antigen in the vaccine. (for example, peptide, MRNA, virus, etc.): (1) The pathway by which dendritic cells (DCs) assume the type of antigen (for example, endocytosis, micropinocytosis); and / or (2) the effectiveness with which the antigen is absorbed by DCs. The dose and type of adjuvant in the vaccine. The length of the vaccine antigen sequence. The number and places of administration of the vaccine. Immune functioning of the patient's baseline (for example, as measured by history of recent infections, blood count, etc.). For RNA vaccines: (1) the rate of turnover of the mRNA protein product in the dendritic cell; (2) the rate of translation of MRNA after uptake by dendritic cells, measured in in vitro or in vivo experiments; and / or (3) the number or cycles of mRNA translation after uptake by dendritic cells, as measured by in vivo or in vitro experiments. The presence of protease cleavage motifs in the peptide, optionally giving additional weight to proteases typically expressed in dendritic cells (as measured by RNA-seq or mass spectrometry). The level of expression of the proteasome and immunoproteasome in typical activated dendritic cells (which can be measured by RNA-sec, mass spectrometry, immunohistochemistry or other standard techniques). The levels of expression of the specific MHC allele in the subject in question (for example, measured by RNA-seq or mass spectrometry), optionally measured specifically in activated dendritic cells or other immune cells. The likelihood of presentation of the peptide by the specific MHC allele in other individuals expressing the specific MHC allele, optionally measured specifically in activated dendritic cells or other immune cells. The probability of presenting peptides by the MHC alleles in the same family of molecules (for example, HLA-A, HLA-B, HLA-C, HLA-DQ, HLA-DR, HLA-DP) in other individuals, optionally measured specifically in dendritic cells or other immune cells.
[096] [096] Escape characteristics of immune tolerance may comprise at least one of the following: Direct measurement of autopeptidoma by protein mass spectrometry, performed on one or more types of cells. Estimation of the autopeptide, taking the union of all substrates k-mer (for example, 5-25) of autoproteins. The estimation of the auto-peptide, using a presentation model similar to the presentation model described above, was applied to all non-mutant autoproteins, optionally accounting for germline variants.
[097] [097] Classification can be performed using the plurality of neoantigens provided by at least one model based at least in part on numerical probabilities. After classification, a selection can be made to select a subset of neoantigens classified according to a selection criterion. After selection, a subset of the classified peptides can be provided as an output.
[098] [098] A number from the set of selected neoantigens can be 20.
[099] [099] The presentation model may represent dependence between the presence of a pair of one of the particular MHC alleles and a specific amino acid at a specific position in a peptide sequence; and probability of presentation on the surface of the tumor cell, by the particular of the MHC alleles of the pair, of such a peptide sequence comprising the particular amino acid in the particular position.
[0100] [0100] A method disclosed in this document may also include applying one or more presentation models to the peptide sequence of the corresponding neoantigen to generate a dependency score for each of the one or more MHC alleles, indicating whether the MHC allele will present the neoantigen corresponding based on at least amino acid positions of the peptide sequence of the corresponding neoantigen.
[0101] [0101] A method disclosed in this document may also include transforming the dependency scores to generate a probability by corresponding allele for each MHC allele indicating a probability that the corresponding MHC allele has the corresponding neoantigen; and combine the probabilities by allele to generate the numerical probability.
[0102] [0102] The step of transforming the dependency scores can model the presentation of the peptide sequence of the corresponding neoantigen as mutually exclusive.
[0103] [0103] A method disclosed in this document may also include transforming a combination of the dependency scores to generate the numerical probability.
[0104] [0104] The step of transforming the combination of the dependency scores can model the presentation of the peptide sequence of the corresponding neoantigen as interfering between the MHC alleles.
[0105] [0105] The set of numerical probabilities can be further identified by at least one non-interactive allele characteristics, and a method disclosed here may also include applying an allele that does not interact with one or more presentation models to the non-interactive characteristics of allele to generate a dependency score for non-interactive allele characteristics indicating whether the peptide sequence of the corresponding neoantigen will be presented based on the non-interactive allele characteristics.
[0106] [0106] A method disclosed in this document may also include combining the dependency score for each MHC allele into one or more MHC alleles with the dependency score for the characteristic that does not interact with alleles; transform the combined dependency scores for each MHC allele to generate a probability per corresponding allele for the MHC allele, indicating a probability that the corresponding MHC allele will have the corresponding neoantigen; and combine the probabilities by allele to generate the numerical probability.
[0107] [0107] A method disclosed in this document may also include transforming a combination of the dependency scores for each of the
[0108] [0108] A set of numerical parameters for the presentation model can be trained based on a set of training data including at least one set of training peptide sequences identified as present in a plurality of samples and one or more MHC alleles associated with each training peptide sequence, in which the training peptide sequences are identified by mass spectrometry on isolated peptides eluted from MHC alleles derived from the plurality of samples.
[0109] [0109] Samples can also include cell lines engineered to express a single MHC class I or class II allele.
[0110] [0110] Samples can also include cell lines engineered to express a plurality of MHC class I or class II alleles.
[0111] [0111] Samples can also include human cell lines obtained from or derived from a plurality of patients.
[0112] [0112] Samples can also include fresh or frozen tumor samples obtained from a plurality of patients.
[0113] [0113] Samples can also include fresh or frozen tissue samples obtained from a plurality of patients.
[0114] [0114] Samples can also include peptides identified using T cell assays.
[0115] [0115] The training data set may also include data associated with: abundance of peptides from the set of training peptides present in the samples; length of the peptide from the set of training peptides in the samples.
[0116] [0116] The training data set can be generated by comparing the set of training peptide sequences via alignment to a database comprising a set of known protein sequences, where the set of training protein sequences are greater than and include the training peptide sequences.
[0117] [0117] The training data set can be generated based on performing or executing nucleotide sequencing on a cell line to obtain at least one of the exome, transcriptome or complete genome sequencing data from the cell line, including sequencing data at least one nucleotide sequence including an alteration.
[0118] [0118] The training data set can be generated based on obtaining at least one of the normal nucleotide sequencing data for the exome, transcriptome and entire genome from normal tissue samples.
[0119] [0119] The training data set may also include data associated with proteomic sequences associated with the samples.
[0120] [0120] The training data set may also include data associated with the MHC peptide sequences associated with the samples.
[0121] [0121] The training dataset may further include data associated with measurements of MHC peptide-binding affinity for at least one of the isolated peptides.
[0122] [0122] The training dataset may further include data associated with measurements of peptide-MHC binding stability for at least one of the isolated peptides.
[0123] [0123] The training data set may also include data associated with the transcriptomes associated with the samples.
[0124] [0124] The training data set may also include data associated with genomes associated with the samples.
[0125] [0125] The peptide training sequences can have lengths within a range of k-mers where k is between 8-15, including for MHC class I or 6- including for MHC class II.
[0126] [0126] A method disclosed in this document may also include encoding the peptide sequence using a one-hot coding scheme.
[0127] [0127] A method disclosed in this document may also include encoding the training peptide sequences using a one-hot coding scheme completed on the left.
[0128] [0128] A method of treating a subject with a tumor, comprising performing the steps of claim 1, and further comprising obtaining a tumor vaccine comprising the set of selected neoantigens and administering the tumor vaccine to the subject.
[0129] [0129] A method disclosed in this document may also include identifying one or more T cells that are specific to the antigen for at least one of the neoantigens in the subset. In some embodiments, identification comprises co-culturing one or more T cells with one or more neoantigens in the subset under conditions that expand one or more antigen-specific T cells. In other embodiments, identification comprises contacting one or more T cells with a tetramer that comprises one or more of the neoantigens in the subset under conditions that permit the connection between the T cell and the tetramer. In yet other embodiments, the method disclosed herein may also include identifying one or more T cell receptors (TCR) from one or more identified T cells. In certain embodiments, the identification of one or more T cell receptors comprises sequencing the T cell receptor sequences of one or more identified T cells. The method disclosed herein may further comprise the genetic engineering of a plurality of T cells to express at least one of the one or more T cell receptors identified; culturing the plurality of T cells under conditions that expand the plurality of T cells; and infuse the expanded T cells into the subject. In some embodiments, genetic engineering of the plurality of T cells to express at least one of the one or more identified T cell receptors comprises cloning the T cell receptor sequences of one or more identified T cells into an expression vector; and transfecting each of the plurality of T cells with the expression vector. In some embodiments, the method disclosed herein further comprises culturing one or more identified T cells under conditions that expand the one or more identified T cells; and infuse the expanded T cells into the subject.
[0130] [0130] Also disclosed here is an isolated T cell that is antigen specific for at least one neoantigen selected in the subset.
[0131] [0131] Also disclosed here is a method for making a tumor vaccine, comprising the steps of: obtaining at least one of the exome, transcriptome, or entire genome tumor nucleotide sequencing data from the subject's tumor cell, in which the tumor nucleotide sequencing are used to obtain data representing peptide sequences from each of a set of neoantigens and wherein the peptide sequence of each neoantigen comprises at least one mutation that distinguishes it from the corresponding parental peptide sequence of the wild type;
[0132] [0132] A tumor vaccine is also disclosed here, including a set of selected neoantigens, selected for executing the method comprising the steps of: obtaining at least one of the tumor nucleotide sequencing data from the exome, transcriptome or entire genome from subject's tumor cell, in which the nucleotide sequencing data of the tumor is used to obtain data representing peptide sequences from each of a set of neoantigens, and in which the peptide sequence of each neoantigen comprises at least one mutation that distinguishes it from the sequence corresponding parental peptide of the wild type; introduce the peptide sequence of each neoantigen into one or more presentation models to generate a set of numerical probabilities that each of the neoantigens is presented by one or more MHC alleles on the surface of the tumor cell of the subject's tumor cell, the set of numerical probabilities having been identified at least based on data received from mass spectrometry; and select a subset of the set of neoantigens based on the set of numerical probabilities to generate a set of selected neoantigens; and producing or having produced a tumor vaccine comprising the set of selected neoantigens.
[0133] [0133] The tumor vaccine can include one or more of a nucleotide sequence, a polypeptide sequence, RNA, DNA, a cell, a plasmid or a vector.
[0134] [0134] The tumor vaccine may include one or more neoantigens presented on the cell surface of the tumor.
[0135] [0135] The tumor vaccine may include one or more neoantigens that are immunogenic in the subject.
[0136] [0136] The tumor vaccine may not include one or more neoantigens that induce an autoimmune response against normal tissue in the subject.
[0137] [0137] The tumor vaccine may include an adjuvant.
[0138] [0138] The tumor vaccine may include an excipient.
[0139] [0139] A method disclosed in this document may also include selecting neoantigens that have an increased probability of being presented on the surface of the tumor cell in relation to neoantigens not selected based on the presentation model.
[0140] [0140] A method disclosed in this document may also include selecting neoantigens that have an increased probability of being able to induce a tumor-specific immune response in the subject in relation to unselected neoantigens based on the presentation model.
[0141] [0141] A method disclosed in this document may also include selecting neoantigens that have an increased probability of being able to be presented to naive T cells by professional antigen presenting cells (APCs) in relation to unselected neoantigens based on the presentation model, optionally where the APC is a dendritic cell (DC).
[0142] [0142] A method disclosed in this document may also include selecting neoantigens that are less likely to be subject to inhibition via central or peripheral tolerance compared to neoantigens not selected based on the presentation model.
[0143] [0143] A method disclosed in this document may also include selecting neoantigens that have a reduced probability of being able to induce an autoimmune response to normal tissue in the subject in relation to unselected neoantigens based on the presentation model.
[0144] [0144] Exome or transcriptome nucleotide sequencing data can be obtained by performing sequencing on tumor tissue.
[0145] [0145] Sequencing can be next generation sequencing (NGS) or any massively parallel sequencing approach.
[0146] [0146] The set of numerical probabilities can be further identified by at least interactive characteristics of the MHC allele comprising at least one of: the predicted affinity with which the MHC allele and the peptide encoded by neoantigen bind; the predicted stability of the neoantigen-encoded peptide-MHC complex; the sequence and length of the peptide encoded by neoantigen; the probability of presenting peptides encoded with neoantigens with a similar sequence in cells of other individuals that express the specific MHC allele, assessed by proteomics by mass spectrometry or by other means; the expression levels of the specific MHC allele in the subject in question (for example, as measured by RNA-seq or mass spectrometry); the general probability of independent presentation of the peptide sequence encoded by neoantigen by the specific MHC allele in other distinct subjects that express the specific MHC allele; the probability of independent presentation of a peptide sequence encoded by general neoantigen by MHC alleles in the same family of molecules (for example, HLA-A, HLA-B, HLA-C, HLA-DQ, HLA-DR, HLA-DP) in other distinct subjects.
[0147] [0147] The set of numerical probabilities is further identified by at least non-interactive characteristics of the MHC allele comprising at least one of: C and N-terminal sequences that flank the peptide encoded by neoantigen within its original protein sequence; the presence of protease cleavage motifs in the neoantigen encoded peptide, optionally weighted according to the expression of the corresponding proteases in the tumor cells (as measured by RNA-seq or mass spectrometry); the turnover rate of the source protein, measured in the appropriate cell type; the length of the origin protein, optionally considering the specific junction variants (“isoforms”) most highly expressed in tumor cells, measured by RNA-seq or proteome mass spectrometry, or as predicted from germline annotation or somatic mutations junction detected in DNA or RNA sequence data; the level of expression of the proteasome, immunoproteasome, thymoproteasome or other proteases in tumor cells (which can be measured by RNA-sec, proteome mass spectrometry or immunohistochemistry); the expression of the gene of origin of the peptide encoded by neoantigen (for example, as measured by RNA-seq or mass spectrometry); the typical tissue-specific expression of the originating peptide gene encoded by neoantigen during various stages of the cell cycle; a comprehensive catalog of characteristics of the source protein and / or its domains, as can be found at, for example, uniProt or PDB http: // www. resb. org / pdb / home / home. of; characteristics that describe the properties of the source protein domain that contains the peptide, for example: secondary or tertiary structure (for example, alpha helix vs beta leaf); alternative junction; the probability of presenting peptides from the protein of origin of the peptide encoded by the neoantigen in question in other distinct subjects; the probability that the peptide will not be detected or over-represented by mass spectrometry due to technical biases; the expression of various gene modules / pathways, measured by RNASeq (which do not need to contain the protein of origin of the peptide) that are informative about the state of tumor cells, stroma or tumor infiltrating lymphocytes (TILs); the number of copies of the gene of origin of the peptide encoded by neoantigen in the tumor cells; the likelihood that the peptide will bind to TAP or the measured or predicted binding affinity of the peptide to TAP; the level of TAP expression in tumor cells (which can be measured by RNA-sec, proteome mass spectrometry, immunohistochemistry); presence or absence of tumor mutations, including, but not limited to: driver mutations in cancer driver genes known as EGFR, KRAS, ALK, RET, ROS1, TP53, CDKN2A, CDKN2B, NTRK1, NTRK2, NTRK3 and in genes that encode the proteins involved in the antigen presentation machines (for example, B2M, HLA-A, HLA-B, HLA-C, TAP-I, TAP-2, TAPBP, CALR, CNX, ERP57, HLA-DM, HLA -DMA, HLA-DMB, HLA-DO, HLA-DOA, HLA-DOB, HLA-DP, HLA-DPAI, HLA-DPB1, HLA-DQ, HLA-DQAI, HLA-DQA2, HLA-DQB1, HLA-DQB2 , HLA-DR, HLA-DRA, HLA-DRB1, HLA-DRB3, HLA-DRB4, HLA-DRB5 or any of the genes encoding components of the proteasome or immunoproteasome). Peptides whose presentation depends on a component of the antigen presentation mechanism that is subject to a loss of function mutation in the tumor have a reduced probability of presentation; presence or absence of functional germline polymorphisms, including, but not limited to: genes encoding proteins involved in the antigen presentation mechanism (eg, B2M, HLA-A, HLA-B, HLA-C, TAP-1, TAP -2, TAPBP, CALR, CNX, ERP57, HLA-DM, HLA-DMA, HLA-DMB, HLA-DO, HLA-DOA, HLA-DOB, HLA-DP, HLA-DPA1, HLA-DPB1, HLA-DQ , HLA-DQA1, HLA-DQA2, HLA-DQB1, HLA-DQB2, HLA-DR, HLA-DRA, HLA-DRB1, HLA-DRB3, HLA-DRB4, HLA-DRB5 or any of the genes encoding components of the proteasome or immunoproteasome); tumor type (for example, NSCLC, melanoma); clinical tumor subtype (eg, squamous vs. non-squamous lung cancer); smoking history; the typical expression of the gene of origin of the peptide in the relevant tumor type or clinical subtype, optionally stratified by driver mutation.
[0148] [0148] The at least one mutation can be a frame shift indel or no frame shift, missense or nonsense, change at the junction site, genomic rearrangement or gene fusion, or any genomic or expression change that gives rise to a neoORF.
[0149] [0149] The tumor cell can be selected from the group consisting of: lung cancer, melanoma, breast cancer, ovarian cancer, prostate cancer, kidney cancer, gastric cancer, colon cancer, testicular cancer, cancer of head and neck, pancreatic cancer, brain cancer, B-cell lymphoma, acute myeloid leukemia, chronic myeloid leukemia, chronic lymphocytic leukemia and T-cell lymphocytic leukemia, non-small cell lung cancer and small cell lung cancer.
[0150] [0150] A method disclosed in this document may also include obtaining a tumor vaccine comprising the set of selected neoantigens or a subset of them, optionally further comprising administering the tumor vaccine to the subject.
[0151] [0151] At least one of the neoantigens in the set of selected neoantigens, when in the form of polypeptide, may include at least one of: an MHC binding affinity with an IC50 value less than 1,000 nM, for class MHC polypeptides I with a length of 8 to 15, 8, 9, 10, 11, 12, 13, 14 or amino acids, for MHC class II polypeptides, with a length of 6 to 30, 6, 7, 8,9, 10, 11 , 12, 13, 14, 15, 16, 17, 18,19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29 or 30 amino acids, presence of sequence motifs in or near the polypeptide following the original protein promoting cleavage of the proteasome and the presence of sequence motifs promoting TAP transport. For MHC Class II, presence of sequence motifs within or near the peptide promoting cleavage by extracellular or lysosomal proteases (eg, cathepsins) or HLA-DM-catalyzed HLA binding.
[0152] [0152] Also disclosed here is a method for generating a model to identify one or more neoantigens that are likely to be presented on the cell surface of a tumor cell, comprising the steps of: receiving mass spectrometry data that comprises data associated with a plurality of isolated peptides eluted from the major histocompatibility complex (MHC) derived from a plurality of samples; obtain a set of training data identifying at least one set of training peptide sequences present in the samples and one or more MHCs associated with each training peptide sequence; training a set of numerical parameters from a presentation model using the training data set comprising the training peptide sequences, the presentation model providing a plurality of numerical probabilities that the tumor cell peptide sequences are presented by one or more MHC alleles on the surface of the tumor cell.
[0153] [0153] The presentation model can represent dependence between: presence of a specific amino acid in a specific position of a peptide sequence; and probability of presenting, by one of the MHC alleles in the tumor cell, the peptide sequence that contains the particular amino acid at the particular position.
[0154] [0154] Samples can also include cell lines engineered to express a single MHC class I or class II allele.
[0155] [0155] Samples can also include cell lines engineered to express a plurality of MHC class I or class II alleles.
[0156] [0156] Samples can also include human cell lines obtained from or derived from a plurality of patients.
[0157] [0157] Samples can also include fresh or frozen tumor samples obtained from a plurality of patients.
[0158] [0158] Samples can also include peptides identified using T cell assays.
[0159] [0159] The training data set may also include data associated with: abundance of peptides from the set of training peptides present in the samples; length of the peptide from the set of training peptides in the samples.
[0160] [0160] A method disclosed in this document may also include obtaining a set of training protein sequences based on the training peptide sequences by comparing the set of training peptide sequences via alignment to a database comprising a set of known protein sequences, wherein the set of training protein sequences is longer than and includes the training peptide sequences.
[0161] [0161] A method disclosed in this document may also include performing or making mass spectrometry happen on a cell line to obtain at least one of the exome, transcriptome or entire genome nucleotide sequencing data from the cell line, the sequencing data nucleotide including at least one protein sequence including a mutation.
[0162] [0162] A method disclosed in this document may also include encoding the training peptide sequences using a one-hot coding scheme.
[0163] [0163] A method disclosed herein may also include obtaining at least one of the normal exome, transcriptome and entire genome nucleotide sequencing data from normal tissue samples; and training the set of parameters of the presentation model using normal nucleotide sequencing data.
[0164] [0164] The training data set may also include data associated with proteomic sequences associated with the samples.
[0165] [0165] The training data set may also include data associated with the MHC peptide sequences associated with the samples.
[0166] [0166] The training dataset may further include data associated with measurements of MHC peptide-binding affinity for at least one of the isolated peptides.
[0167] [0167] The training dataset may further include data associated with measurements of peptide-MHC binding stability for at least one of the isolated peptides.
[0168] [0168] The training data set may also include data associated with the transcriptomes associated with the samples.
[0169] [0169] The training data set may also include data associated with genomes associated with the samples.
[0170] [0170] A method disclosed in this document may also include logistic regression of the set of parameters.
[0171] [0171] Training peptide sequences can have lengths within a range of k-mers where k is between 8-15, including for MHC class I or 6- including for MHC class II.
[0172] [0172] A method disclosed in this document may also include encoding the training peptide sequences using a one-hot coding scheme completed on the left.
[0173] [0173] A method disclosed in this document may also include determining values for the set of parameters using a deep learning algorithm.
[0174] [0174] Methods are disclosed here to identify one or more neoantigens that are likely to be presented on the surface of a tumor cell in a tumor cell, comprising performing the steps of: receiving mass spectrometry data comprising data associated with a plurality of isolated peptides eluted from the major histocompatibility complex (MHC) derived from a plurality of fresh or frozen tumor samples; obtain a set of training data identifying at least one set of training peptide sequences present in the tumor samples and presented in one or more MHC alleles associated with each training peptide sequence; obtain a set of training protein sequences based on the training peptide sequences; and training a set of numerical parameters from a presentation model using the training protein sequences and the training peptide sequences, the presentation model that provides a plurality of numerical probabilities that the tumor cell peptide sequences are presented by one or more MHC alleles on the surface of the tumor cell.
[0175] [0175] The presentation model can represent dependency between: the presence of a pair of one of the particular MHC alleles and a specific amino acid at a specific position in a peptide sequence; and probability of presentation on the surface of the tumor cell, by the particular of the MHC alleles of the pair, of such a peptide sequence comprising the particular amino acid in the particular position.
[0176] [0176] A method disclosed in this document may also include selecting a subset of neoantigens, where the subset of neoantigens is selected because each has an increased probability of being presented on the tumor cell surface in relation to one or more distinct tumor neoantigens.
[0177] [0177] A method disclosed in this document may also include selecting a subset of neoantigens, in which the subset of neoantigens is selected because each has an increased probability that it is capable of inducing a tumor-specific immune response in the subject in relation to a or more distinct tumor neoantigens.
[0178] [0178] A method disclosed in this document may also include selecting a subset of neoantigens, in which the subset of neoantigens is selected because each has an increased likelihood that it will be able to be presented to naive T cells by professional antigen presenting cells ( APCs) relative to one or more distinct tumor neoantigens, optionally in which the APC is a dendritic cell (DC).
[0179] [0179] A method disclosed in this document may also include selecting a subset of neoantigens, where the subset of neoantigens is selected because each is less likely to be subject to inhibition via central or peripheral tolerance in relation to one or more tumor neoantigens distinct.
[0180] [0180] A method disclosed in this document may also include selecting a subset of neoantigens, in which the subset of neoantigens is selected because each is less likely to be able to induce an autoimmune response to normal tissue in the subject relative to one or more more distinct tumor neoantigens.
[0181] [0181] A method disclosed in this document may also include selecting a subset of neoantigens, where the subset of neoantigens is selected because each has a reduced probability that it will be differently post-translationally modified in tumor cells versus APCs, optionally in which the APC is a dendritic cell (DC).
[0182] [0182] The practice of the methods described herein will employ, unless otherwise indicated, conventional methods of protein chemistry, biochemistry, recombinant DNA techniques and pharmacology, within the skill of the art. Such techniques are fully explained in the literature. See, for example, T. E. Creighton, Proteins: Structures and Molecular Properties (W. H. Freeman and Company, 1993);
[0183] [0183] Also disclosed here are methods for identifying certain mutations (for example, the variants or alleles that are present in cancer cells). In particular, these mutations may be present in the genome, transcriptome, proteome, or exome of cancer cells in a subject with cancer, but not in the subject's normal tissue.
[0184] [0184] Genetic mutations in tumors can be considered useful for the immunological targeting of tumors if they lead to changes in the amino acid sequence of a protein exclusively in the tumor. Useful mutations include: (1) non-synonymous mutations leading to different amino acids in the protein; (2) reading mutations in which a stop codon is modified or deleted, leading to the translation of a longer protein with a new specific tumor sequence at the C terminal; (3) mutations at the junction site that lead to the inclusion of an intron in the mature MRNA and, therefore, in a unique tumor-specific protein sequence; (4) chromosomal rearrangements that give rise to a chimeric protein with specific tumor sequences at the junction of 2 proteins (ie, gene fusion); (5) frame shift mutations or deletions that lead to a new open reading frame with a new tumor-specific protein sequence. Mutations can also include one or more indel substitutions without frame shift, missense or nonsense, alteration at the junction site, genomic rearrangement or gene fusion or any genomic or expression alteration that gives rise to a neoORF.
[0185] [0185] Peptides with mutations or mutated polypeptides resulting from, for example, mutations at the junction site, frame shift, gene reading or fusion in tumor cells can be identified by sequencing DNA, RNA or protein in the tumor versus normal cells.
[0186] [0186] Mutations may also include tumor-specific mutations previously identified. Known tumor mutations can be found in the database of the Catalog of Somatic Mutations in Cancer (COSMIC).
[0187] [0187] A variety of methods are available to detect the presence of a specific mutation or allele in an individual's DNA or RNA. Advances in this field have provided large-scale, accurate, easy and inexpensive SNP genotyping. For example, several techniques have been described, including dynamic hybridization of specific alleles (DASH), diagonal microplate gel electrophoresis (MADGE), pyrosquencing, specific oligonucleotide binding, the TaqMan system and various DNA chip technologies, such as Affymetrix Chips SNP. These methods use the amplification of a target genetic region, typically by PCR. Still other methods, based on the generation of small signal molecules by invasive cleavage followed by mass spectrometry or immobilized padlock probes and amplification of rolling circles. Several of the methods known in the art for detecting specific mutations are summarized below.
[0188] [0188] PCR-based detection means can include multiplex amplification of a plurality of markers simultaneously. For example, it is well known in the art to select PCR primers to generate PCR products that do not overlap in size and can be analyzed simultaneously. Alternatively, it is possible to amplify different markers with differentially labeled primers and, thus, each one can be differentially detected. Obviously, hybridization-based detection means allow differential detection of multiple PCR products in a sample. Other techniques are known in the art to allow multiplex analysis of a plurality of markers.
[0189] [0189] Several methods have been developed to facilitate the analysis of single nucleotide polymorphisms in genomic DNA or cellular RNA. For example, a single base polymorphism can be detected using a specialized exonuclease-resistant nucleotide, as disclosed, for example, in Mundy, C. R. (US Patent
[0190] [0190] A solution-based method can be used to determine the nucleotide identity of a polymorphic site. Cohen, D. etal. (French Patent
[0191] [0191] An alternative method, known as Genetic Bit Analysis or GBA, is described by Goelet, P. et al. (PCT Application 92/15712). The method by Goelet, P. et al. uses mixtures of labeled terminators and a primer complementary to the 3 'sequence of a polymorphic site. The tagged terminator that is incorporated is thus determined by, and complementary to, the nucleotide present in the polymorphic site of the target molecule being evaluated. In contrast to the method by Cohen et al. (French Patent
[0192] [0192] Various primer-guided nucleotide incorporation procedures for analyzing polymorphic sites in DNA have been described (Kombher, JSetal., Nucl. Acids. Res. 17: 7779-7784 (1989); Sokolov, BP, Nucl. Acids Res. 18: 3671 (1990); Syvanen, A.-C., et al., Genomics 8: 684-692 (1990); Kuppuswamy, MN et al., Proc. Natl. Acad. Sci. (USA) 88 : 1143-1147 (1991); Prezant, TR et al., Hum. Mutat. 1: 159-164 (1992); Ugozzoli, L. etal., GATA 9: 107-112 (1992); Nyren, P. etal ., Anal. Biochem. 208: 171-175 (1993)). These methods differ from GBA in that they use the incorporation of labeled deoxynucleotides to discriminate between bases at a polymorphic site. In this format, since the signal is proportional to the number of deoxynucleotides incorporated, polymorphisms that occur in executions of the same nucleotide can result in signals that are proportional to the duration of the execution (Syvanen, A.-C ,, et al., Amer. J. Hum. Genet. 52: 46-59 (1993)).
[0193] [0193] Several initiatives obtain sequence information directly from millions of individual DNA or RNA molecules in parallel. Sequencing technologies for synthesis of unique molecules in real time depend on the detection of fluorescent nucleotides as they are incorporated into a nascent strand of DNA that is complementary to the template being sequenced. In one method, oligonucleotides 30 to 50 bases in length are anchored covalently at the 5 'end of the glass coverslips. These anchored wires perform two functions. First, they act as capture sites for the target mold strips if the molds are configured with capture tails complementary to the surface-bound oligonucleotides. They also act as primers for the extension of the primer directed to the template that forms the basis for reading the sequence. The capture primers function as a fixed position site for sequence determination using various cycles of synthesis, detection and chemical cleavage of the dye ligand to remove the dye. Each cycle consists of adding the labeled polymerase / nucleotide mixture, rinsing, imaging and dye cleavage. In an alternative method, the polymerase is modified with a fluorescent donor molecule and immobilized on a coverslip, while each nucleotide is color-coded with an acceptor fluorescent fraction bound to a gamma phosphate. The system detects the interaction between a fluorescence-labeled polymerase and a fluorescence-modified nucleotide when the nucleotide becomes incorporated into the strand again. Other synthesis sequencing technologies also exist.
[0194] [0194] Any suitable synthesis sequencing platform can be used to identify mutations. As described above, four main synthesis sequencing platforms are currently available: Roche / 454 Life Sciences genome sequencers, Illumina / Solexa 1G Analyzer, Applied BioSystems SOLID system and Helicos Biosciences Heliscope system. Synthesis sequencing platforms have also been described by Pacific BioSciences and VisiGen Biotechnologies. In some embodiments, a plurality of nucleic acid molecules being sequenced is attached to a support (for example, solid support). To immobilize the nucleic acid on a support, a capture sequence / universal initiation site can be added at the 3 'and / or 5' ends of the template. Nucleic acids can be linked to the support by hybridizing the capture sequence to a complementary sequence covalently linked to the support. The capture sequence (also referred to as a universal capture sequence) is a nucleic acid sequence complementary to a sequence attached to a support that can serve double as a universal primer.
[0195] [0195] As an alternative to a capture sequence, a member of a coupling pair (such as, for example, antibody / antigen, receptor / ligand or the avidin-biotin pair, as described in, for example, US Patent Application 2006 / 0252077) can be attached to each fragment to be captured on a surface coated with a respective second member of that coupling pair.
[0196] [0196] After capture, the sequence can be analyzed, for example, by single molecule detection / sequencing, for example, as described in the Examples and in US Patent 7,283,337, including sequencing by template-dependent synthesis. In sequencing by synthesis, the molecule bound to the surface is exposed to a plurality of nucleotide triphosphates labeled in the presence of polymerase. The model sequence is determined by the order of the labeled nucleotides incorporated at the 3 'end of the growth chain. This can be done in real time or in repeat and step mode. For real-time analysis, different optical markers for each nucleotide can be incorporated and multiple lasers can be used for stimulation of incorporated nucleotides.
[0197] [0197] Sequencing can also include other techniques and platforms for massively parallel or next generation sequencing (NGS). Additional examples of massively parallel sequencing techniques and platforms are Illumina HiSeq or MiSeg, Thermo PGM or Proton, Pac Bio RS II or Sequel, Qiagen's Gene Reader and Oxford Nanopore MinION. Additional current current massively parallel sequencing technologies can be used, as well as future generations of these technologies.
[0198] [0198] Any type of cell or tissue can be used to obtain nucleic acid samples for use in the methods described here. For example, a sample of DNA or RNA can be obtained from a tumor or body fluid, for example, blood, obtained by known techniques (for example, venipuncture) or saliva. Alternatively, nucleic acid tests can be performed on dry samples (for example, hair or skin). In addition, a sample can be obtained for sequencing a tumor and another sample can be obtained from normal tissue for sequencing in which the normal tissue is the same type of tissue as the tumor. One sample can be obtained for sequencing a tumor and another sample can be obtained from normal tissue for sequencing in which the normal tissue is of a different tissue type in relation to the tumor.
[0199] [0199] Tumors can include one or more of lung cancer, melanoma, breast cancer, ovarian cancer, prostate cancer, kidney cancer, gastric cancer, colon cancer, testicular cancer, head and neck cancer, pancreatic cancer, brain cancer, B-cell lymphoma, acute myeloid leukemia, chronic myeloid leukemia, chronic lymphocytic leukemia and T-cell lymphocytic leukemia, non-small cell lung cancer and small cell lung cancer.
[0200] [0200] Alternatively, protein mass spectrometry can be used to identify or validate the presence of mutated peptides linked to MHC proteins in tumor cells. The peptides can be eluted with acid from tumor cells or HLA molecules that are immunoprecipitated from the tumor and then identified by mass spectrometry.
[0201] [0201] Neoantigens can include nucleotides or polypeptides. For example, a neoantigen may be an RNA sequence that codes for a polypeptide sequence. Neoantigens useful in vaccines can therefore include nucleotide sequences or polypeptide sequences.
[0202] [0202] Isolated peptides comprising tumor-specific mutations identified by the methods disclosed herein, peptides comprising tumor-specific mutations known and mutant polypeptides or fragments thereof identified by the methods disclosed herein are disclosed. Neoantigenic peptides can be described in the context of their coding sequence, where a neoantigen includes the nucleotide sequence (for example, DNA or RNA) that encodes the related polypeptide sequence.
[0203] [0203] One or more polypeptides encoded by a sequence of neoantigenic nucleotides may comprise at least one of: an MHC binding affinity with an IC50 value less than 1,000nM, for MHC Class peptides
[0204] [0204] One or more neoantigens can be presented on the surface of a tumor.
[0205] [0205] One or more neoantigens may be immunogenic in a subject with a tumor, for example, capable of eliciting a T cell response or a B cell response in the subject.
[0206] [0206] One or more neoantigens that induce an autoimmune response in a subject can be excluded from consideration in the context of generating the vaccine for a subject with a tumor.
[0207] [0207] The size of at least one neoantigenic peptide molecule can comprise, but is not limited to, about 5, about 6, about 7, about 8, about 9, about 10, about 11, about 12, about 13, about 14, about 15, about 16, about 17, about 18, about 19, about 20, about 21, about 22, about 23, about 23 24, about 25, about 26, about 27, about 28, about 29, about 30, about 31, about 32, about 33, about 34, about 35, about 35, about 36, about 37, about 38, about 39, about 40, about 41, about 42, about 43, about 44, about 45, about 46, about 47, about 48, about 48, about 49, about 50, about 60, about 70, about 80, about 90, about 100, about 110, about 120 or more residues of amino molecules and any derivable range therein. In specific embodiments, the neoantigenic peptide molecules are equal to or less than 50 amino acids.
[0208] [0208] Neo-antigenic peptides and polypeptides can be: for Class I MHC 15 residues or less in length and generally consist of between about 8 and about 11 residues, particularly 9 or 10 residues; for Class II MHC, 6- residues, inclusive.
[0209] [0209] If desired, a longer peptide can be designed in several ways. In one case, when the probabilities of presenting peptides in the HLA alleles are predicted or known, a longer peptide may consist of: (1) individual presented peptides with extensions of 2-5 amino acids towards the N and C terminal of each corresponding genetic product; (2) a concatenation of some or all of the peptides presented with sequences extended to each one. In another case, when the sequencing reveals a long sequence of neoepitopes (> 10 residues) present in the tumor (for example, due to a frame shift, reading or intron inclusion that leads to a new peptide sequence), a longer peptide would consist of: (3) the whole stretch of new tumor-specific amino acids - thus ignoring the need for computational or in vitro selection of the strongest peptide presented by the strongest HLA. In both cases, the use of a longer peptide allows endogenous processing by the patient's cells and can lead to a more effective presentation of the antigen and induction of T cell responses.
[0210] [0210] Neo-antigenic peptides and polypeptides can be presented in an HLA protein. In some respects, neoantigenic peptides and polypeptides are presented in an HLA protein with greater affinity than a wild type peptide. In some respects, a neoantigenic peptide or polypeptide may have an IC 50 of at least less than 5,000 nM, at least less than 1,000 nM, at least less than 500 nM, at least less than 250 nM, at least less than 200 nM, in at least less than 150 nM, at least less than 100 nM, at least less than 50 nM or less.
[0211] [0211] In some respects, neoantigenic peptides and polypeptides do not induce an autoimmune response and / or invoke immune tolerance when administered to a subject.
[0212] [0212] Compositions comprising at least two or more neoantigenic peptides are also provided. In some embodiments, the composition contains at least two distinct peptides. At least two distinct peptides can be derived from the same polypeptide. By distinct polypeptides, it is understood that the peptide varies in length, amino acid sequence, or both. The peptides are derived from any polypeptide known or found to contain a specific tumor mutation. Suitable polypeptides from which neoantigenic peptides can be derived can be found, for example, in the COSMIC database. COSMIC organizes comprehensive information on somatic mutations in human cancer. The peptide contains the tumor-specific mutation. In some ways, the specific tumor mutation is a driver mutation for a specific type of cancer.
[0213] [0213] Neo-antigenic peptides and polypeptides with a desired activity or property can be modified to provide certain desired attributes, for example, improved pharmacological characteristics, while increasing or at least substantially retaining all the biological activity of the unmodified peptide to bind the molecule of Desired MHC and activate the appropriate T cell. For example, neoantigenic peptides and polypeptides can be subject to several changes, such as substitutions, conservative or non-conservative, where these changes can provide certain advantages in their use, such as better binding, stability or presentation to the MHC. Conservative substitutions mean the replacement of an amino acid residue with another that is biologically and / or chemically similar, for example, one hydrophobic residue for another, or a polar residue for another. Substitutions include combinations such as Gly, Ala; Val, Ile, Leu, Met; Asp, Glu; Asn, Gln; Ser, Thr; Lys, Arg; and Phe, Tyr. The effect of single amino acid substitutions can also be investigated using D-amino acids. Such modifications can be made using well-known peptide synthesis procedures, as described in, for example, Merrifield, Science 232: 341-347 (1986), Barany & Merrifield, The Peptides, Gross & Meienhofer, eds. (N. Y., Academic Press), pp. 1-284 (1979); and Stewart & Young, Solid Phase Peptide Synthesis, (Rockford, II, Pierce), 2nd Ed. (1984).
[0214] [0214] Modifications of peptides and polypeptides with various amino acid mimetics or unnatural amino acids can be particularly useful in increasing the stability of the peptide and polypeptide in vivo. Stability can be tested in several ways. For example, peptidases and various biological media, such as plasma and human serum, have been used to test stability. See, for example, Verhoef et al., Eur. J. Drug Metab Pharmacokin. 11: 291-302 (1986). The peptide half-life can be determined conveniently using a 25% (v / v) human serum assay. The protocol is usually as follows. The combined human serum (Type AB, inactivated without heating) is removed by centrifugation before use. The serum is then diluted to 25% with RPM tissue culture medium! and used to test the stability of the peptide. At predetermined time intervals, a small amount of reaction solution is removed and added to 6% aqueous trichloracetic acid or ethanol. The cloudy reaction sample is cooled (4 degrees C) for 15 minutes and then centrifuged to pellet the precipitated serum proteins. The presence of the peptides is then determined by reverse phase HPLC using specific stability chromatography conditions.
[0215] [0215] Peptides and polypeptides can be modified to provide desired attributes other than improved serum half-life. For example, the ability of the peptides to induce CTL activity can be enhanced by binding to a sequence that contains at least one epitope that is capable of inducing a helper T cell response. The immunogenic peptide / T helper conjugates can be linked by a spacer molecule. The spacer is typically composed of relatively small neutral molecules, such as amino acids or amino acid mimetics, which are substantially unloaded under physiological conditions. Spacers are typically selected from, for example, Ala, Gly or other neutral spacers of non-polar amino acids or neutral polar amino acids. It will be understood that the optionally present spacer does not need to be composed of the same residues and, therefore, can be a hetero or homo-oligomer. When present, the spacer will generally be at least one or two residues, more generally three to six residues. Alternatively, the peptide can be attached to the helper T peptide without a spacer.
[0216] [0216] A neoantigenic peptide can be attached to the auxiliary T peptide, directly or through a spacer, at the amino or carboxy terminus of the peptide. The amino terminus of the neoantigenic peptide or T helper peptide can be acylated. Exemplary T helper peptides include tetanus toxoid 830-843, influenza 307-319, malaria circumsporozoite 382-398 and 3578-389.
[0217] [0217] Proteins or peptides can be produced by any technique known to those skilled in the art, including the expression of proteins, polypeptides or peptides using standard molecular biological techniques, the isolation of proteins or peptides from natural sources or the chemical synthesis of proteins or peptides. The nucleotide and protein sequences, polypeptides and peptides corresponding to several genes have been previously disclosed and can be found in computerized databases known to those skilled in the art. One such database is the National Center for Biotechnology Information Genbank and GenPept database,
[0218] [0218] In an additional aspect, a neoantigen includes a nucleic acid (eg, polynucleotide) that encodes a neoantigenic peptide or a portion thereof. The polynucleotide can be, for example, DNA, cDNA, PNA, CNA, RNA (for example, mRNA), single-stranded and / or double-stranded or native or stabilized polynucleotides, such as polynucleotides with a backbone phosphorothed, or combinations thereof and may or may not contain introns. A still further aspect provides an expression vector capable of expressing a polypeptide or a portion thereof. Expression vectors for different cell types are well known in the art and can be selected without undue experimentation. Generally, DNA is inserted into an expression vector, such as a plasmid, in the proper orientation and in the correct reading frame for expression. If necessary, the DNA can be linked to the transcriptional and translational regulatory control nucleotide sequences recognized by the desired host, although such controls are generally available in the expression vector. The vector is then introduced into the host using standard techniques. Guidelines can be found, for example, in Sambrook et al. (1989) Molecular Cloning, A Laboratory Manual, Cold Spring Harbor Laboratory, Cold Spring Harbor, N. Y.
[0219] [0219] Also disclosed herein is an immunogenic composition, for example, a vaccine composition, capable of enhancing a specific immune response, for example, a tumor specific immune response. Vaccine compositions typically comprise a plurality of neoantigens, for example, selected using a method described herein. Vaccine compositions can also be referred to as vaccines.
[0220] [0220] A vaccine can contain between 1 and 30 peptides, 2, 3, 4, 5, 6, 7.8, 9, 10, 11, 12,13, 14, 15, 16, 17, 18, 19, 20 , 21, 22, 23, 24, 25, 26, 27, 28, 29 or 30 different peptides, 6, 7, 8, 9, 10 11, 12, 13 or 14 different peptides, or 12, 13 or 14 different peptides . The peptides can include post-translational modifications. A vaccine can contain between 1 and 100 or more nucleotide sequences, 2, 3, 4, 5, 6.7, 8.9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 , 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45 , 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70 , 71, 72.73, 74.75, 76.77, 78.79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94.95 , 96, 97, 98, 99, 100 or more different nucleotide sequences, 6, 7, 8, 9, 10 11, 12, 13 or 14 different nucleotide sequences or 12, 13 or 14 different nucleotide sequences. A vaccine can contain between 1 and 30 neoantigenic sequences, 2, 3, 4, 5, 6,7, 8,9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22,23, 24, 25,26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46 , 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69.70, 71 , 72, 73, 74.75, 76.77, 78.79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94.95, 96 , 97, 98, 99, 100 or more different neoantigen sequences, 6, 7, 8, 9, 10 11, 12, 13 or 14 different neoantigen sequences or 12, 13 or 14 different neoantigen sequences.
[0221] [0221] In one embodiment, different peptides and / or polypeptides or nucleotide sequences that encode them are selected so that peptides and / or polypeptides capable of associating with different MHC molecules, such as different MHC class I molecules and / or different MHC class II molecules. In some respects, a vaccine composition comprises the coding sequence for peptides and / or polypeptides capable of associating with the most frequently occurring MHC class I molecules and / or MHC class II molecules. Therefore, vaccine compositions can comprise different fragments capable of associating with at least 2 preferred, at least 3 preferred molecules or at least at least 4 preferred MHC class I molecules and / or MHC class II molecules.
[0222] [0222] The vaccine composition may be able to increase a specific response of cytotoxic T cells and / or a specific response of helper T cells.
[0223] [0223] A vaccine composition may further comprise an adjuvant and / or a carrier. Examples of useful adjuvants and carriers are given below. A composition can be associated with a carrier, for example, a protein or an antigen presenting cell, such as, for example, a dendritic cell (DC) capable of presenting the peptide to a T cell.
[0224] [0224] Adjuvants are any substance whose mixture in a vaccine composition increases or modifies the immune response to a neoantigen. The carriers can be scaffold structures, for example, a polypeptide or a polysaccharide, to which a neoantigen is capable of being associated. Optionally, adjuvants are conjugated covalently or non-covalently.
[0225] [0225] The ability of an adjuvant to increase an immune response to an antigen is typically manifested by a significant or substantial increase in an immunity-mediated reaction or reduction in the symptoms of the disease. For example, an increase in humoral immunity is typically manifested by a significant increase in the titer of increased antibodies to the antigen, and an increase in T cell activity is typically manifested in increased cell proliferation, or cell cytotoxicity or cytokine secretion. An adjuvant can also alter an immune response, for example, by changing a mainly humoral or Th response to a mainly cellular or Th response.
[0226] [0226] Suitable adjuvants include, but are not limited to, 1018 ISS, alum, aluminum salts, Amplivax, AS15, BCG, CP-870.893, CpG7909, CyaA, dSLIM, GM-CSF, IC30, IC31, Imiquimod, ImuFact IMP321, IS Patch, ISS, ISCOMATRIX, JuvImmune, LipoVac, MF59, monophosphoryl lipid A, Montanide IMS 1312, Montanide ISA 206, Montanide ISA 50V, Montanide ISA-S1, OK-432, OM-174, OM-197-MP-EC, ONTAK , PepTel vector system, PLG microparticles, resiquimod, SRL 172, virosomes and other virus-like particles, YF-17D, VEGF trap, R848, beta-glucan, Pam3Cys, Aquila QS21 stimulus (Aquila Biotech, Worcester, Mass., USA), derived from saponin, mycobacterial extracts and imitations of synthetic bacterial cell wall and other proprietary adjuvants, such as Ribi's Detox. Quil or Superfos. Adjuvants such as incomplete Freund or GM-CSF are useful. Various immunological adjuvants (eg, MF59) specific for dendritic cells and their preparation have been described previously (Dupuis M, et al., Cell Immunol. 1998; 186 (1): 18-27; Allison AC; Dev Biol Stand. 1998; 92: 3-11). Cytokines can also be used. Several cytokines have been directly linked to influencing the migration of dendritic cells to lymphoid tissues (eg, TNF-alpha), accelerating the maturation of dendritic cells in cells presenting antigen efficient for T lymphocytes (eg, GM-CSF, IL-1 and IL -4) (US Patent 5,849,589, hereby specifically incorporated by reference in its entirety) and acting as immunoadjuvants (for example, IL-12) (Gabrilovich D 1, et al., J Immunother Emphasis Tumor Immunol. 1996 ( 6): 414-418).
[0227] [0227] CpG immunostimulatory oligonucleotides have also been reported to increase the effects of adjuvants in a vaccine setting. Other TLR binding molecules can also be used, such as RNA binding, TLR 7, TLR 8 and / or TLR 9.
[0228] [0228] Other examples of useful adjuvants include, but are not limited to, chemically modified CpGs (eg, CpR, Idera), Poly (I: O) (eg, polii: CILU), non-CpG bacterial RNA or DNA , as well as small immunoactive molecules and antibodies such as cyclophosphamide, sunitinib, bevacizumab, celebrex, NOX- 4016, sildenafil, tadalafil, vardenafil, sorafinib, XL-999, CP-547632, pazopanib, ZD2171, AZD2171, ipilimumabe e17 an adjuvant. The amounts and concentrations of adjuvants and additives can be readily determined by one skilled in the art without undue experimentation. Additional adjuvants include colony-stimulating factors, such as Granulocyte Macrophage Colony-Stimulating Factor (GM-CSF, sargramostim).
[0229] [0229] A vaccine composition can comprise more than one different adjuvant. In addition, a therapeutic composition can comprise any adjuvant substance including any of the above alternatives or combinations thereof. It is also contemplated that a vaccine and an adjuvant can be administered together or separately in any appropriate sequence.
[0230] [0230] A vehicle (or excipient) can be present independently of an adjuvant. The function of a carrier may for example be to increase the molecular weight of mutants in particular to increase activity or immunogenicity, to confer stability, to increase biological activity or to increase serum half-life. In addition, a carrier can assist in presenting peptides to T cells. A vehicle can be any suitable vehicle known to the person skilled in the art, for example, a protein or antigen presenting cell. A carrier protein can be, but is not limited to keyhole keyhole limpet hemocyanin, serum proteins such as transferrin, bovine serum albumin, human serum albumin, thyroglobulin or ovalbumin, immunoglobulins or hormones, such as insulin or palmitic acid. For human immunization, the vehicle is generally a physiologically acceptable, human-acceptable and safe vehicle. However, tetanus toxoid and / or diphtheria toxoid are suitable carriers. Alternatively, the carrier can be dextran, for example, sepharose.
[0231] [0231] Cytotoxic T cells (CTLs) recognize an antigen in the form of a peptide attached to an MHC molecule instead of the intact foreign antigen itself. The MHC molecule itself is located on the cell surface of an antigen presenting cell. Thus, an activation of CTLs is possible if a trimeric complex of peptide antigen, MHC molecule and APC is present. Correspondingly, it can improve the immune response if not only the peptide is used for the activation of CTLs, but if additional APCs are added with the respective MHC molecule. Therefore, in some embodiments, a vaccine composition additionally contains at least one antigen presenting cell.
[0232] [0232] Neoantigens can also be included in vector-based viral vaccine platforms, such as vaccinia, chickenpox, self-replicating alphavirus, marabavirus, adenovirus (see, for example, Tatsis et al., Adenoviruses, Molecular Therapy (2004) 10 , 616—629), or lentivirus, including, but not limited to, second, third, or second / third generation hybrid lentiviruses and recombinant lentiviruses of any generation designed to target specific cell types or receptors (see, for example, Hu et al ., Immunization Delivered by Lentiviral Vectors for Cancer and Infectious Diseases, Immunol Rev. (2011) 239 (1): 45-61, Sakuma et al., Lentiviral vectors: basic to translational, Biochem J. (2012) 443 (3) : 603-18, Cooper et al., Rescue of splicing-mediated intron loss maximizes expression in lentiviral vectors containing the human ubiquitin C promoter, Nucl. Acids Res. (2015) 43 (1): 682-690, Zufferey et al. , Self Inactivating Lentivirus Vector for Safe and Efficient In Vivo Gene Delivery, J. Virol. (1998) 72 (12): 9873-9880). Depending on the packaging capacity of the viral vector-based vaccine platforms mentioned above, this approach may provide one or more nucleotide sequences that encode one or more neoantigen peptides. Sequences can be flanked by non-mutated sequences, they can be separated by ligands, or they can be preceded by one or more sequences targeting a subcellular compartment (see, for example, Gros et al., Prospective identification of neoantigen-specific lymphocytes in the peripheral blood of melanoma patients, Nat Med. (2016) 22 (4): 433-8, Stronen et al., Targeting of cancer neoantigens with donor-derived T cell receptor repertoires, Science. (2016) 352 (6291): 1337- 41, Lu et al., Efficient identification of mutated cancer antigens recognized by T cells associated with durable tumor regressions, Clin Cancer Res. (2014) 20 (13): 3401-10). After introduction into a host, infected cells express the neoantigens and thus elicit a host immune response (eg, CTL) against the peptide (s). The vaccinia vectors and methods useful in immunization protocols are described in, for example, US Patent 4,722,848. Another vector is BCG (Bacille Calmette Guerin). BCG vectors are described in Stover et al. (Nature 351: 456-460 (1991)). A wide variety of other vaccine vectors useful for therapeutic administration or immunization of neoantigens, for example, Salmonella typhi vectors and the like will be apparent to those skilled in the art from the description here. IV. A. Additional Considerations for Vaccine Design and Manufacturing IV. A. 1. Determination of a set of peptides that cover all tumor subclones
[0233] [0233] Truncal peptides, meaning those presented by all or most tumor subclones, will be prioritized for inclusion in the vaccine. ** Optionally, if there are no truncal peptides predicted to be presented and immunogenic with high probability, or if the number of truncal peptides predicted to be presented and immunogenic with high probability is small enough that additional non-truncal peptides can be included in the vaccine , then other peptides can be prioritized by estimating the number and identity of tumor subclones and choosing peptides to maximize the number of tumor subclones covered by the vaccine. * IV. A. 2. Prioritization of neoantigens
[0234] [0234] After all the above neoantigen filters have been applied, more candidate neoantigens may still be available for vaccine inclusion than the vaccine technology can support. In addition, uncertainty about various aspects of the analysis of neoantigens may remain and there may be exchanges between different properties of the vaccine candidate antigen. Thus, instead of predetermined filters at each stage of the selection process, an integrated multidimensional model can be considered that places the candidate neoantigens in a space with at least the following axes and optimizes the selection using an integrative approach. LL Risk of autoimmunity or tolerance (risk of germline) (the lowest risk of autoimmunity is generally preferred)
[0235] [0235] In addition, optionally, neoantigens can be de-prioritized (for example, excluded) from vaccination, if they are predicted to be presented by lost or inactivated HLA alleles in all or part of the patient's tumor. The loss of the HLA allele can occur by somatic mutation, loss of heterozygosis or homozygous exclusion of the locus. Methods for detecting somatic mutation in the HLA allele are well known in the art, for example (Shukla et al., 2015). Methods for somatic LOH detection and homozygous deletion (including for HLA locus) are also well described. (Carter et al., 2012; McGranahan et al., 2017; Van Loo et al., 2010).
[0236] [0236] A method is also provided to induce a tumor specific immune response in a subject, vaccinate against a tumor, treat and alleviate a cancer symptom in a subject, by administering to the subject one or more neoantigens, such as a plurality of neoantigens identified using methods disclosed herein.
[0237] [0237] In some ways, a subject has been diagnosed with cancer or is at risk of developing cancer. A subject can be a human, dog, cat, horse or any animal in which a tumor specific immune response is desired. A tumor can be any solid tumor, such as breast, ovary, prostate, lung, kidney, gastric, colon, testicular, head and neck, pancreas, brain, melanoma and other tissue organ tumors and hematological tumors such as lymphomas and leukemias, including acute myeloid leukemia, chronic myeloid leukemia, chronic lymphocytic leukemia, T-cell lymphocytic leukemia and B-cell lymphomas.
[0238] [0238] A neoantigen can be administered in an amount sufficient to induce a CTL response.
[0239] [0239] A neoantigen can be administered alone or in combination with other therapeutic agents. The therapeutic agent is, for example, a chemotherapeutic agent, radiation or immunotherapy. Any therapeutic treatment suitable for a particular cancer can be administered.
[0240] [0240] In addition, a subject may further be administered an anti-immunosuppressive / immunostimulatory agent, such as a checkpoint inhibitor. For example, the subject can also be administered an anti-CTLA or anti-PD-1 or anti-PD-L1 antibody. Blocking CTLA-4 or PD-L1 by antibodies can improve the immune response to cancer cells in the patient. In particular, blocking CTLA-4 has been shown to be effective when following a vaccination protocol.
[0241] [0241] The ideal amount of each neoantigen to be included in a vaccine composition and the ideal dosage regimen can be determined. For example, a neoantigen or its variant can be prepared for intravenous injection (1.v.), subcutaneous injection (s.c.), intradermal injection (i.d.), intraperitoneal injection (i.p.), intramuscular injection (i.m.). Injection methods include s.c., 1.d., 1.p., .m., Ei.v. DNA or RNA injection methods include i.d., i.m., s.c., 1.p. eiv. Other methods of administering the vaccine composition are known to those skilled in the art.
[0242] [0242] A vaccine can be compiled so that the selection, number and / or quantity of neoantigens present in the composition are / are tissues, cancer and / or patient specific. For example, the exact selection of peptides can be guided by patterns of expression of parental proteins in a given tissue. Selection may depend on the specific type of cancer, the disease status, previous treatment regimens, the patient's immune status and, of course, the patient's HLA haplotype. In addition, a vaccine can contain individualized components, according to the personal needs of the particular patient. Examples include varying the selection of neoantigens according to the expression of the neoantigen in the particular patient or adjustments for secondary treatments after a first round or treatment regimen.
[0243] [0243] For a composition to be used as a cancer vaccine, neoantigens with similar normal autopeptides that are expressed in large amounts in normal tissues can be avoided or be present in small amounts in a composition described herein. On the other hand, if a patient's tumor is known to express large quantities of a certain neoantigen, the respective pharmaceutical composition for the treatment of that cancer may be present in large quantities and / or more than one specific neoantigen for that neoantigen or the pathway. of this neoantigen can be included.
[0244] [0244] Compositions comprising a neoantigen can be administered to an individual who is already suffering from cancer. In therapeutic applications, the compositions are administered to a patient in an amount sufficient to elicit an effective CTL response to the tumor antigen and to cure or at least partially stop symptoms and / or complications. An adequate amount to accomplish this is defined as a “therapeutically effective dose. “The effective amounts for this use will depend, for example, on the composition, the mode of administration, the state and severity of the disease to be treated, the weight and general health of the patient, and the judgment of the prescribing physician. It should be kept in mind that the compositions can generally be used in severe disease states, that is, life-threatening or potentially life-threatening situations, especially when cancer is metastasized. In such cases, in view of the minimization of foreign substances and the relative non-toxic nature of a neoantigen, it is possible and may be considered desirable by the attending physician to administer substantial excesses of these compositions.
[0245] [0245] For therapeutic use, administration can begin with the detection or surgical removal of tumors. This is followed by increasing the doses until at least the symptoms are substantially reduced and for a later period.
[0246] [0246] Pharmaceutical compositions (for example, vaccine compositions) for therapeutic treatment are intended for parenteral, topical, nasal, oral or local administration. The pharmaceutical compositions can be administered parenterally, for example, intravenously, subcutaneously, intradermally or intramuscularly. The compositions can be administered at the site of surgical excision to induce a local immune response to the tumor. Compositions for parenteral administration comprising a solution of the neoantigen are disclosed herein and the vaccine compositions are dissolved or suspended in an acceptable vehicle, for example, an aqueous vehicle. A variety of aqueous vehicles can be used, for example, water, buffered water, 0.9% saline, 0.3% glycine, hyaluronic acid and the like. These compositions can be sterilized by conventional well-known sterilization techniques or can be filtered until sterilized. The resulting aqueous solutions can be packaged for use as is, or lyophilized, the lyophilized preparation being combined with a sterile solution prior to administration. The compositions can contain pharmaceutically acceptable auxiliary substances, as needed to approach physiological conditions, such as pH adjustment and buffering agents, tonicity adjusting agents, wetting agents and the like, for example, sodium acetate, sodium lactate, chloride sodium, potassium chloride, calcium chloride, sorbitan monolaurate, triethanolamine oleate, etc.
[0247] [0247] Neoantigens can also be administered via liposomes, which direct them to a specific cell tissue, such as lymphoid tissue. Liposomes are also useful for increasing the half-life. Liposomes include emulsions, foams, micelles, insoluble monolayers, liquid crystals, phospholipid dispersions, lamellar layers and the like. In these preparations, the neoantigen to be distributed is incorporated as part of a liposome, alone or together with a molecule that binds, for example, to a receptor prevalent among lymphoid cells, such as monoclonal antibodies that bind to the CDA45 antigen or with other therapeutic or immunogenic compositions.
[0248] [0248] To target immune cells, a ligand to be incorporated into the liposome may include, for example, antibodies or their specific fragments for determinants of the cell surface of the desired cells of the immune system. A liposome suspension can be administered intravenously, locally, topically, etc. in a dose that varies according to, inter alia, the mode of administration, the peptide being distributed and the stage of the disease to be treated.
[0249] [0249] For therapeutic or immunization purposes, nucleic acids encoding a peptide and, optionally, one or more of the peptides described herein can also be administered to the patient. Various methods are conveniently used to deliver nucleic acids to the patient. For example, nucleic acid can be distributed directly, like "naked DNA". This approach is described, for example, in Wolff et al., Science 247: 1465-1468 (1990), as well as in US Patent 5,580,859 and
[0250] [0250] Nucleic acids can also be delivered complexed to cationic compounds, such as cationic lipids. Methods of delivering lipid-mediated genes are described, for example, in 9618372WO0AWO 96/18372; 9324640 WOAWO 93/24640; Mannino & Gould-Fogerite, BioTechniques 6 (7): 682-691
[0251] [0251] Neoantigens can also be included in vector-based viral vaccine platforms, such as vaccinia, chickenpox, self-replicating alphavirus, marabavirus, adenovirus (see, for example, Tatsis et al., Adenoviruses, Molecular Therapy (2004) 10 , 616—629), or lentivirus, including, but not limited to, second, third, or second / third generation hybrid lentiviruses and recombinant lentiviruses of any generation designed to target specific cell types or receptors (see, for example, Hu et al ., Immunization Delivered by Lentiviral Vectors for Cancer and Infectious Diseases, Immunol Rev. (2011) 239 (1): 45-61, Sakuma et al., Lentiviral vectors: basic to translational, Biochem J. (2012) 443 (3) : 603-18, Cooper et al., Rescue of splicing-mediated intron loss maximizes expression in lentiviral vectors containing the human ubiquitin C promoter, Nucl. Acids Res. (2015) 43 (1): 682-690, Zufferey et al. , Self Inactivating Lentivirus Vector for Safe and Efficient In Vivo Gene Delivery, J. Virol. (1998) 72 (12): 9873-9880). Depending on the packaging capacity of the viral vector-based vaccine platforms mentioned above, this approach may provide one or more nucleotide sequences that encode one or more neoantigen peptides. Sequences can be flanked by non-mutated sequences, they can be separated by ligands, or they can be preceded by one or more sequences targeting a subcellular compartment (see, for example, Gros et al., Prospective identification of neoantigen-specific lymphocytes in the peripheral blood of melanoma patients, Nat Med. (2016) 22 (4): 433-8, Stronen et al., Targeting of cancer neoantigens with donor-derived T cell receptor repertoires, Science. (2016) 352 (6291): 1337- 41, Lu et al., Efficient identification of mutated cancer antigens recognized by T cells associated with durable tumor regressions, Clin Cancer Res. (2014) 20 (13): 3401-10). After introduction into a host, infected cells express the neoantigens and thus elicit a host immune response (eg, CTL) against the peptide (s). Vectors and vaccinia methods useful in immunization protocols are described in, for example, US Patent 4,722,848. Another vector is BCG (Bacille Calmette Guerin). BCG vectors are described in Stover et al. (Nature 351: 456-460 (1991)). A wide variety of other vaccine vectors useful for therapeutic administration or immunization of neoantigens, for example, Salmonella typhi vectors and the like will be apparent to those skilled in the art from the description here.
[0252] [0252] A means of administering nucleic acids uses minigene constructs that encode one or more epitopes. To create a DNA sequence that encodes the selected CTL epitopes (minigene) for expression in human cells, the amino acid sequences of the epitopes are translated inversely. A human codon usage chart is used to guide the choice of codon for each amino acid. These epitope encoding DNA sequences are directly linked, creating a continuous polypeptide sequence. To optimize expression and / or immunogenicity, additional elements can be incorporated into the design of the minigene. Examples of amino acid sequences that can be translated inversely and included in the minigene sequence include: helper T lymphocyte, epitopes, a leader sequence (signal) and an endoplasmic reticulum retention signal. In addition, the MHC presentation of the CTL epitopes can be improved by including synthetic (e.g., polyalanine) or naturally occurring flanking sequences adjacent to the CTL epitopes. The minigene sequence is converted into DNA by assembling oligonucleotides that encode the minigene strands more and less. The overlapping oligonucleotides (30-100 bases in length) are synthesized, phosphorylated, purified and annealed under appropriate conditions using well-known techniques. The ends of the oligonucleotides are joined using T4 DNA ligase. This synthetic minigene, which encodes the CTL epitope polypeptide, can then be cloned into the desired expression vector.
[0253] [0253] Purified plasmid DNA can be prepared for injection using a variety of formulations. The simplest of these is the reconstitution of lyophilized DNA in sterile phosphate buffered saline (PBS). A variety of methods have been described and new techniques may become available. As noted above, nucleic acids are conveniently formulated with cationic lipids. In addition, glycolipids, fusogenic liposomes, peptides and compounds collectively referred to as protective, interactive and non-condensing (PINC) can also be complex for purified plasmid DNA to influence variables such as stability, intramuscular dispersion or trafficking for specific organs or cell types.
[0254] [0254] A method of making a tumor vaccine is also disclosed, comprising performing the steps of a method disclosed herein; and producing a tumor vaccine comprising a plurality of neoantigens or a subset of the plurality of neoantigens.
[0255] [0255] The neoantigens disclosed herein can be manufactured using methods known in the art. For example, a method of producing a neoantigen or vector (for example, a vector including at least one sequence that encodes one or more neoantigens) disclosed here may include growing a host cell in conditions suitable for expressing the neoantigen or vector in which the host cell comprises at least one polynucleotide that encodes the neoantigen or vector and purifies the neoantigen or vector. Standard purification methods include chromatographic techniques, electrophoretic, immunological, precipitation, dialysis, filtration, concentration and chromatofocus techniques.
[0256] [0256] Host cells may include a Chinese Hamster Ovary (CHO) cell, NS0 cell, yeast or HEK293 cell. Host cells can be transformed with one or more polynucleotides comprising at least one nucleic acid sequence that encodes a neoantigen or vector disclosed herein, optionally wherein the isolated polynucleotide further comprises a promoter sequence operably linked to at least one nucleic acid sequence that encodes the neoantigen or vector. In certain embodiments, the isolated polynucleotide may be cDNA.
[0257] [0257] T cells can be isolated from patients' blood, lymph nodes or tumors. T cells can be enriched for antigen-specific T cells, for example, by classifying cells that bind to the MHC antigen tetramer or by classifying activated cells stimulated in an in vitro c-culture of T cells and cells presenting antigen-pulsed antigen. Various reagents are known in the art for identifying antigen-specific T cells, including antigen-loaded tetramers and other MHC-based reagents.
[0258] [0258] TCR alpha-beta (or gamma-delta) dimers relevant to the antigen can be identified by single cell sequencing of antigen-specific T cell TCRs. Alternatively, mass TCR sequencing of antigen-specific T cells can be performed and alpha-beta pairs with a high probability of matching can be determined using a TCR matching method known in the art.
[0259] [0259] Alternatively or in addition, antigen-specific T cells can be obtained through the in vitro initiation of naive T cells from healthy donors. T cells obtained from PBMCSs, lymph nodes or umbilical cord blood can be repeatedly stimulated by antigen-pulsed antigen presenting cells to initiate differentiation of T cells with antigen experience. TCRs can then be identified in the same way as described above for patient antigen-specific T cells.
[0260] [0260] Research methods for NGS analysis of tumors and normal exomes and transcriptomes have been described and applied in the neoantigen identification space. ó! *! 5 The example below considers certain optimizations for greater sensitivity and specificity in the identification of neoantigens in the clinical setting. These optimizations can be grouped into two areas, those related to laboratory processes and those related to NGS data analysis.
[0261] [0261] The process improvements presented here address challenges in the discovery of high precision neoantigens from clinical samples with low tumor content and small volumes, extending concepts developed for reliable evaluation of cancer driver genes in targeted cancer panels! ” for the set of exomes and transcripts required for the identification of neoantigens. Specifically, these improvements include: LL Target a single medium deep (> 500x) coverage across the entire exome of the tumor to detect mutations present at low frequency of the mutant allele due to the low tumor content or subclonal state.
[0262] [0262] Improvements in analytical methods address sub-optimal sensitivity and specificity of common mutation-calling approaches to research and specifically consider customizations relevant to the identification of neoantigens in the clinical setting. These include:
[0263] [0263] In samples with polyadenylated RNA, the presence of viral and microbial RNA in the RNA-seq data will be assessed using CoMPASS ”** RNA or a similar method, to identify additional factors that may predict the response of the patient.
[0264] [0264] The isolation of HLA peptide molecules was performed using classic immunoprecipitation (IP) methods after lysis and solubilization of the tissue sample ****. A clarified lysate was used for specific HLA PI.
[0265] [0265] Immunoprecipitation was performed using antibodies coupled to beads where the antibody is specific for HLA molecules. For a pan-Class HLA immunoprecipitation, a pan-Class I CR antibody is used; for HLA-DR Class II,
[0266] [0266] The clarified tissue lysate is added to the antibody beads for immunoprecipitation. After immunoprecipitation, the beads are removed from the lysate and the lysate is stored for further experiments, including additional PIs. The IP beads are washed to remove the non-specific binding and the HLA / peptide complex is eluted from the beads using standard techniques. Protein components are removed from the peptides using a molecular weight spin column or C18 fractionation. The resulting peptides are brought to dryness by SpeedVac evaporation and, in some cases, are stored at -20ºC before analysis by MS.
[0267] [0267] The dried peptides are reconstituted in an HPLC buffer suitable for reverse phase chromatography and loaded onto a C-18 microcapillary HPLC column for gradient elution on a Fusion Lumos (Thermo) mass spectrometer. The MS1 peptide mass / charge spectra (m / z) were collected in the high resolution Orbitrap detector, followed by low resolution MS2 scans collected in the ion trap detector after HCD fragmentation of the selected ion. In addition, MS2 spectra can be obtained using CID or ETD fragmentation methods or any combination of the three techniques to obtain greater amino acid coverage of the peptide. The MS2 spectrum can also be measured with high resolution mass accuracy on the Orbitrap detector.
[0268] [0268] The MS2 spectra of each analysis are searched in a protein database using Comet "* and the peptide identification is scored using Percolator ***. Additional sequencing is performed using the PEAKS studio (Bioinformatics Solutions Inc. ) and other search engines or sequencing methods may be used, including spectral matching and de-sequencing ”.
[0269] [0269] Using the peptide YVYVADVAAK (SEQ ID NO: 1) it was determined what the detection limits are using different amounts of peptide loaded in the LC column. The amounts of peptide tested were 1 pmol, 100 fmol, 10 fmol, 1 fmol and 100amol. (Table 1) The results are shown in Figure 1F. These results indicate that the lower limit of detection (LoD) is in the atomol range (** 18), that the dynamic range covers five orders of magnitude and that the noise signal seems sufficient for sequencing in low femtomol ranges ( "* 15) Peptide m / z Loaded on Column | Copies / Cell in 1e9 cells 566,830 1 pmol 600 559,816 10 fmol 6 VII. VILA Presentation Model. System Overview
[0270] [0270] FIG. 2A is an overview of an environment 100 to identify probabilities of presentation of peptides in patients, according to one modality. Environment 100 provides context for introducing a presentation identification system 160, including a presentation information store 165 itself.
[0271] [0271] The presentation identification system 160 is one or more computer models, incorporated into a computing system as discussed below in relation to FIG. 14, which receives peptide sequences associated with a set of MHC alleles and determines the probabilities that the peptide sequences are presented by one or more of the set of associated MHC alleles. The presentation identification system 160 can be applied to MHC class I and II alleles. This is useful in several contexts. A specific use case for the presentation identification system 160 is that it is able to receive nucleotide sequences from candidate neoantigens associated with a set of MHC alleles of tumor cells from a patient 110 and determine the probabilities that the candidate neoantigens will be presented by one or more of the MHC alleles associated with the tumor and / or induce immunogenic responses in the patient's immune system 110. Candidate neoantigens with high probabilities, as determined by system 160, can be selected for inclusion in a vaccine 118, a response antitumor immune response can be elicited from the patient's 110 immune system that supplies the tumor cells.
[0272] [0272] The presentation identification system 160 determines the presentation probabilities through one or more presentation models. Specifically, presentation models generate probabilities of whether certain peptide sequences will be presented for a set of associated MHC alleles and are generated based on presentation information stored in reserve 165. For example, presentation models can generate probabilities for a sequence peptide “YVYVADVAAK (SEQ ID NO: 1)” be presented for the HLA-A * 02: 01, HLA-A * 03: 01, HLA-B * 07: 02, HLA-B * 08: 03 allele set, HLA-C * 01: 04 on the cell surface of the sample. Presentation information 165 contains information on whether the peptides bind to different types of MHC alleles, so that those peptides are presented by the MHC alleles, which in the models are determined depending on the positions of the amino acids in the peptide sequences. The presentation model can predict whether an unrecognized peptide sequence will be presented in association with an associated set of MHC alleles based on presentation information 165. As mentioned earlier, presentation models can be applied to MHC class I alleles. and II. VII.B. Presentation Information
[0273] [0273] FIG. 2 illustrates a method for obtaining presentation information, according to a modality. Presentation information 165 includes two general categories of information: interactive allele information and non-interactive allele information. Interactive allele information includes information that influences the presentation of peptide sequences that are dependent on the type of MHC allele. Non-interactive allele information includes information that influences the presentation of peptide sequences that are independent of the type of MHC allele.
[0274] [0274] Interactive allele information mainly includes identified peptide sequences that are known to have been presented by one or more identified MHC molecules from humans, mice, etc. Notably, this may or may not include data obtained from tumor samples. The peptide sequences shown can be identified from cells that express a single MHC allele. In this case, the peptide sequences presented are generally collected from single allele cell lines that are engineered to express a predetermined MHC allele and that are subsequently exposed to the synthetic protein. The peptides presented in the MHC allele are isolated by techniques such as acid elution and identified by mass spectrometry. FIG. 2B shows an example of this, in which the example peptide YEMFNDKSQORAPDDKMEF (SEQ ID NO: 2), shown in the predetermined MHC allele HLA-DRBI1 * 12: 01, is isolated and identified by mass spectrometry. Since in this situation the peptides are identified by cells modified to express a single predetermined MHC protein, the direct association between a displayed peptide and the MHC protein to which it was bound is definitely known.
[0275] [0275] The peptide sequences shown can also be collected from cells that express multiple MHC alleles. Typically in humans, 6 different types of MHC-I and up to 12 different types of MHC-II molecules are expressed for a cell. Such peptide sequences presented can be identified from cell lines of multiple alleles that are engineered to express multiple predetermined MHC alleles. Such peptide sequences shown can also be identified from tissue samples, from normal tissue samples or from tumor tissue samples. In this particular case, the MHC molecules can be immunoprecipitated from normal or tumor tissue. The peptides presented in the multiple MHC alleles can be isolated by techniques such as acid elution and identified by mass spectrometry. FIG. 2C shows an example of this, where the six example peptides, YEMFNDKSF (SEQ ID NO: 3), HROEIFSHDFJ (SEQ ID NO: 4), FIIEJFOESS (SEQ ID NO: 5), NEIOREIREI (SEQ ID NO: 6), JFKSIFEMMSJDSSUIFLKSJFIEIFJ (SEQ ID NO: 7) and KNFLENFIESOFI (SEQ ID NO: 8), are shown in the MHC class 1 alleles identified HLA-A * 01: 01, HLA-A * 02: 01, HLA-B * 07: 02, HLA-B * 08: 01 and MHC alleles of IT class HLA-DRB1 * 10: 01, HLA-DRB1: 11: 01 and are isolated and identified by mass spectrometry. In contrast to single allele cell lines, the direct association between a presented peptide and the MHC protein to which it has been linked may be unknown, since the bound peptides are isolated from the MHC molecules before being identified.
[0276] [0276] Interactive allele information can also include the ion stream of mass spectrometry, which depends on the concentration of the peptide-MHC molecule complexes and the efficiency of peptide ionization. Ionization efficiency varies from peptide to peptide in a sequence dependent manner. Generally, the efficiency of ionization varies from peptide to peptide by approximately two orders of magnitude, while the concentration of peptide-MHC complexes varies in a greater range than this.
[0277] [0277] Interactive allele information can also include measurements or predictions of binding affinity between a given MHC allele and a particular peptide. (72, 73, 74) One or more affinity models can generate such predictions. For example, going back to the example shown in FIG. 1D, presentation information 165 may include a 1,000nM binding affinity prediction between the YEMEFNDKSF peptide (SEQ ID NO: 3) and the HLA-A * 01: 01 class I allele. Few peptides with IC50> 1,000nm are presented by the MHC, and lower IC50 values increase the likelihood of presentation. Presentation information 165 may include a prediction of binding affinity between the KNFLENFIESOFI peptide and the HLA-DRBI1: 11: 01 class II allele.
[0278] [0278] Interactive allele information may also include measurements or predictions of MHC complex stability. One or more stability models that can generate such predictions. More stable peptide-MHC complexes (ie, complexes with a longer half-life) are more likely to be presented with a high number of copies in tumor cells and in antigen presenting cells that encounter the vaccine antigen. For example, going back to the example shown in FIG. 2C, presentation information 165 may include a 1h half-life stability prediction for the HLA-A * 01: Ol class I molecule. Presentation information 165 may also include a prediction of half-life stability for the class II molecule HLA-DRBI1: 11:01.
[0279] [0279] Interactive allele information can also include the measured or predicted rate of the MHC-peptide complex formation reaction. Complexes that form at a higher rate are more likely to be present on the cell surface in high concentration.
[0280] [0280] Interactive allele information can also include the sequence and length of the peptide. Class I MHC molecules typically prefer to present peptides with lengths between 8-15 peptides. 60-80% of the peptides shown are of length 9. Class II MHC molecules typically prefer to present peptides with lengths between 6-30 peptides.
[0281] [0281] Interactive allele information may also include the presence of kinase sequence motifs in the neoantigen-encoded peptide and the absence or presence of specific post-translational modifications in the neoantigen-encoded peptide. The presence of kinase motifs affects the likelihood of post-translational modification, which can improve or interfere with MHC binding.
[0282] [0282] Interactive allele information can also include the levels of expression or activity of proteins involved in the post-translation modification process, for example, kinases (as measured or predicted from RNA sec, mass spectrometry or other methods) .
[0283] [0283] Interactive allele information may also include the likelihood of presenting peptides with a similar sequence in cells of other individuals expressing the specific MHC allele, as assessed by proteomics by mass spectrometry or by other means.
[0284] [0284] Interactive allele information may also include the levels of expression of the specific MHC allele in the subject in question (for example, as measured by RNA-sec or mass spectrometry). Peptides that bind more strongly to an MHC allele that is expressed at high levels are more likely to be presented than peptides that bind more strongly to an MHC allele that is expressed at a low level.
[0285] [0285] Interactive allele information can also include the overall likelihood of independent presentation of the peptide sequence encoded by neoantigen by the specific MHC allele in other individuals expressing the specific MHC allele.
[0286] [0286] Interactive allele information may also include the general likelihood of independent presentation by peptide sequence by the MHC alleles in the same family of molecules (for example, HLA-A, HLA-B, HLA-C, HLA-DQ, HLA -DR, HLA-DP) in other individuals. For example, HLA-C molecules are typically expressed at lower levels than HLA-A or HLA-B molecules and, consequently, presentation of a peptide by HLA-C is a priori less likely than presentation by HLA -A or HLA-B. For another example, HLA-DP is typically expressed at lower levels than HLA-DR or HLA-DQ; consequently, the presentation of a peptide by HLA-DP is less likely than the presentation by HLA-DR or HLA-DQ.
[0287] [0287] Interactive allele information can also include the protein sequence of the specific MHC allele.
[0288] [0288] Any information that does not interact with the MHC allele listed in the section below can also be modeled as interactive MHC allele information.
[0289] [0289] Non-interactive allele information may include C-terminal sequences that flank the peptide encoded by neoantigen within its original protein sequence. For MHC-I, the C-terminal flanking sequences can affect the proteasome processing of peptides. However, the C-terminal flanking sequence is cleaved from the peptide by the proteasome before the peptide is transported to the endoplasmic reticulum and finds MHC alleles on cell surfaces. Consequently, the MHC molecules do not receive information about the C-terminal flanking sequence and, therefore, the effect of the C-terminal flanking sequence cannot vary depending on the type of MHC allele. For example, going back to the example shown in FIG. 2C, presentation information 165 may include the C-terminal flanking sequence FOEIFNDKSLDKFII (SEQ ID NO: 9) of the displayed FJIEJFOESS peptide (SEQ ID NO: 5) identified from the peptide source protein.
[0290] [0290] Non-interactive allele information may also include MRNA quantification measures. For example, MRNA quantification data can be obtained for the same samples that provide mass spectrometry training data. As later described with reference to FIG. The expression of 13G, RNA was identified as a strong predictor of the presentation of peptides. In one embodiment, mRNA quantification measurements are identified using the RSEM software tool. The detailed implementation of the RSEM software tool can be found at Bo Li and Colin N. Dewey. RSEM: accurate transcript quantification from —RNA-Segq data with or without a reference genome. BMC Bioinformatics, 12: 323, August 2011. In one embodiment, the quantification of MRNA is measured in units of fragments per kilobase of transcription per Million readings mapped (FPKM).
[0291] [0291] Non-interactive allele information can also include the N-terminal sequences that flank the peptide within its original protein sequence.
[0292] [0292] Non-interactive allele information can also include the gene of origin of the peptide sequence. The source gene can be defined as the Ensembl protein family of the peptide sequence. In other examples, the source gene can be defined as the source DNA or the RNA and source of the peptide sequence. The source gene can, for example, be represented as a chain of nucleotides that code for a protein or, alternatively, be represented more categorically based on a named set of known DNA or RNA sequences that code for specific proteins. In another example, non-interactive allele information may also include the original transcript or isoform or the set of possible transcripts or isoforms of origin of the peptide sequence extracted from a database such as Ensembl or RefSeq.
[0293] [0293] Non-interactive allele information may also include the presence of protease cleavage motifs in the peptide, optionally weighted according to the expression of corresponding proteases in tumor cells (as measured by RNA-sec or mass spectrometry). Peptides that contain protease cleavage motifs are less likely to be presented, because they will be more easily degraded by proteases and therefore will be less stable within the cell.
[0294] [0294] Non-interactive allele information may also include the turnover rate of the source protein, as measured in the appropriate cell type. A faster turnover rate (ie, lower half-life) increases the likelihood of presentation; however, the predictive power of this characteristic is low if measured in a different cell type.
[0295] [0295] Non-interactive allele information can also include the length of the orgiem protein, optionally considering the specific junction variants (“isoforms”) most highly expressed in tumor cells, measured by RNA-seq or proteome mass spectrometry, or as predicted in the annotation of mutations in the germline or in somatic processing detected in the DNA or RNA sequence data.
[0296] [0296] Non-interactive allele information may also include the level of expression of the proteasome, immunoproteasome, thymoproteasome or other proteases in tumor cells (which can be measured by RNA-seq, proteome mass spectrometry or immunohistochemistry). Different proteasomes have preferences for different cleavage sites. More weight will be given to the cleavage preferences of each type of proteasome in proportion to their level of expression.
[0297] [0297] Non-interactive allele information may also include expression of the gene of origin of the peptide (for example, as measured by RNA-seq or mass spectrometry). Possible optimizations include adjusting the measured expression to explain the presence of stromal cells and tumor infiltrating lymphocytes in the tumor sample. Peptides from more highly expressed genes are more likely to be presented. Gene peptides with undetectable levels of expression can be excluded from consideration.
[0298] [0298] Non-interactive allele information may also include the probability that the mRNA of the origin of the peptide encoded by neoantigen is subject to meaningless mediated decay, as predicted by a meaningless mediated decay model, for example, the Rivas et al., Science 2015.
[0299] [0299] Non-interactive allele information may also include tissue-specific expression typical of the peptide's origin gene during various stages of the cell cycle. Genes that are expressed at a low general level (as measured by RNA-seq proteomics or mass spectrometry), but which are known to be expressed at a high level during specific stages of the cell cycle, are likely to produce more highly displayed peptides than genes which are stably expressed at very low levels.
[0300] [0300] Non-interactive allele information may also include a comprehensive catalog of the characteristic of the source protein, as provided in, for example, uniProt or PDB http: // www. resb. org / pdb / home / home. of. These characteristics may include, among others: secondary and tertiary structures of the protein, subcellular location 11, terms of genetic ontology (GO). Specifically, this information may contain annotations that act at the protein level, for example, 5 'RTU length and annotations that act at the level of specific residues, for example, helix motif between residues 300 and 310. These characteristics may also include motifs of turns, motifs of leaves and disordered residues.
[0301] [0301] Non-interactive allele information can also include characteristics that describe the properties of the source protein domain that contains the peptide, for example: secondary or tertiary structure (for example, alpha helix vs beta leaf); Alternative junction.
[0302] [0302] Non-interactive allele information can also include characteristics that describe the presence or absence of a presentation hotspot at the position of the peptide in the protein of origin of the peptide.
[0303] [0303] Non-interactive allele information may also include the probability of presenting peptides of the protein of origin of the peptide in question in other individuals (after adjusting for the level of expression of the source protein in these individuals and the influence of different types of these individuals).
[0304] [0304] Non-interactive allele information may also include the likelihood that the peptide will not be detected or over-represented by mass spectrometry due to technical bias.
[0305] [0305] The expression of several genetic modules / pathways, measured by a gene expression assay such as RNASeg, microarray (s), targeted panel (s) like Nanostring or representatives of one or more genes of genetic modules measured by assays as RT-PCR (which does not need to contain the protein of origin of the peptide) which is informative about the state of tumor cells, stroma or tumor infiltrating lymphocytes (TILs).
[0306] [0306] Non-interactive allele information can also include the number of copies of the peptide's originating gene in tumor cells. For example, peptides from genes that are subject to homozygous deletion in tumor cells may receive a probability of presenting zero.
[0307] [0307] Non-interactive allele information may also include the likelihood that the peptide will bind to TAP or the measured or predicted binding affinity of the peptide to TAP. Peptides that are more likely to bind to TAP or peptides that bind to TAP with greater affinity are more likely to be presented by MHC-LI.
[0308] [0308] Non-interactive allele information may also include the level of TAP expression in tumor cells (which can be measured by RNA-sec, proteome mass spectrometry, immunohistochemistry). For MHC-I, higher levels of TAP expression increase the probability of presenting all peptides.
[0309] [0309] Non-interactive allele information may also include the presence or absence of tumor mutations, including, but not limited to:
[0310] [0310] Presence or absence of functional germline polymorphisms, including, among others: i. In the genes that encode the proteins involved in the antigen presenting machines (for example, B2M, HLA-A, HLA-B, HLA-C, TAP-I, TAP-2, TAPBP, CALR, CNX, ERP57, HLA-DM , HLA-DMA, HLA-DMB, HLA-DO, HLA-DOA, HLA-DOBHLA-DP, HLA-DPAI, HLA-DPB1, HLA-DQ, HLA-DQAI, HLA-DQA 2, HLA-DQB1, HLA -DQB2, HLA-DR, HLA-DRA, HLA-DRB1, HLA-DRB3, HLA-DRBA4, HLA-DRB5 or any of the genes encoding components of the proteasome or immunoproteasome)
[0311] [0311] Non-interactive allele information can also include the type of tumor (eg, NSCLC, melanoma).
[0312] [0312] Non-interactive allele information may also include features known to HLA alleles, as reflected by, for example, HLA allele suffixes. For example, the suffix N in the name of the HLA-A * 24: 09N allele indicates a null allele that is not expressed and is therefore unlikely to have epitopes; the complete nomenclature of the HLA allele suffix is described at https: // www. ebi. B.C. uk / ipd / imgt / hla / nomenclature / suffixes. html.
[0313] [0313] Non-interactive allele information may also include clinical tumor subtypes (eg, squamous versus non-squamous lung cancer).
[0314] [0314] Non-interactive allele information may also include a smoking history.
[0315] [0315] Non-interactive allele information may also include a history of sunburn, sun exposure or exposure to other mutagens.
[0316] [0316] Non-interactive allele information may also include the typical expression of the gene of origin of the peptide in the relevant tumor type or clinical subtype, optionally stratified by driver mutation. Genes that are typically expressed at high levels in the relevant tumor type are more likely to be presented.
[0317] [0317] Non-interactive allele information may also include the frequency of mutation in all tumors, or tumors of the same type, or in tumors of individuals with at least one shared MHC allele or in tumors of the same type in individuals with at least least one shared MHC allele.
[0318] [0318] In the case of a tumor-specific peptide mutated, the list of characteristics used to predict a probability of presentation may also include annotation of the mutation (eg, missense, reading, frame shift, fusion, etc.) or is predicted it is believed that the mutation will result in meaningless mediated decay (NMD). For example, peptides from protein segments that are not translated into tumor cells due to homozygous mutations of early arrest may receive a probability of presenting zero. NMD results in less translation of the mMRNA, which decreases the likelihood of presentation.
[0319] [0319] FIG. 3 is a high-level block diagram illustrating the computer's logical components of the presentation identification system 160, according to one embodiment. In this example embodiment, the presentation identification system 160 includes a data management module 312, an encoding module 314, a training module 316 and a prediction module 320. The presentation identification system 160 is also composed of a training data store 170 and a presentation template store 175. Some modalities of the model management system 160 have different modules than those described here. Likewise, functions can be distributed between modules in a different way than described here.
[0320] [0320] Data management module 312 generates training data sets 170 from presentation information 165. Each training data set contains a plurality of data instances, in which each data instance i contains a set of independent variables z 'that include at least one peptide sequence displayed or not shown p', one or more MHC alleles associated with 'associated with peptide sequence p', and a dependent variable y 'that represents information that the presentation identification system 160 is interested in predicting new values for independent variables.
[0321] [0321] In a specific implementation referred to throughout the rest of the specification, the dependent variable y 'is a binary label indicating whether the peptide p' was presented by one or more MHC alleles associated with '. However, it is appreciated that in other implementations, the dependent variable y 'can represent any other type of information that the presentation identification system 160 is interested in predicting dependent on the independent variables zi. For example, in another implementation, the dependent variable y 'can also be a numerical value indicating the ion current of the mass spectrometry identified for the data instance.
[0322] [0322] The peptide sequence p 'for data instance i is an amino acid sequence k ;, in which k; can vary between instances of data i within a range. For example, this range can be 8 to 15 for MHC class I or 6 to 30 for MHC class II. In a specific implementation of system 160, all p 'peptide sequences in a training data set can be the same length, for example, 9. The number of amino acids in a peptide sequence can vary depending on the type of MHC allele (for example, MHC alleles in humans, etc.). The MHC alleles ai for data instance i indicate which MHC alleles were present in association with the corresponding peptide sequence p '.
[0323] [0323] The data management module 312 may also include additional interactive allele variables, such as b 'binding affinity and s' stability predictions in conjunction with the p' peptide sequences and associated MHC alleles contained in the training data 170. For example, training data 170 may contain b 'binding affinity predictions between a p' peptide and each of the associated MHC molecules indicated in a '. As another example, training data 170 may contain stability predictions for each of the MHC alleles indicated in a.
[0324] [0324] The data management module 312 may also include non-interactive variables of wi allele, such as C-terminal flanking sequences and MRNA quantification measurements in conjunction with the p 'peptide sequences.
[0325] [0325] The data management module 312 also identifies peptide sequences that are not presented by the doeMHC alleles to generate the 170 training data. This usually involves identifying the “longer” sequences of the source protein that include presented peptide sequences before the presentation. When presentation information contains engineered cell lines, the data management module 312 identifies a series of peptide sequences in the synthetic protein to which cells have been exposed and which have not been shown in the MHC alleles of the cells. When the presentation information contains tissue samples, the data management module 312 identifies source proteins from which the presented peptide sequences originated and identifies a series of peptide sequences in the source protein that were not shown in the MHC alleles of the cells tissue sample.
[0326] [0326] The data management module 312 can also artificially generate peptides with random amino acid sequences and identify the sequences generated as peptides not shown in the MHC alleles. This can be achieved by randomly generating peptide sequences, allowing the 312 data management module to easily generate large amounts of synthetic data for peptides not shown in the MHC alleles. Since, in reality, a small percentage of peptide sequences are presented by the MHC alleles, it is highly likely that the synthetically generated peptide sequences were not presented by the MHC alleles, even though they were included in the proteins processed by the cells.
[0327] [0327] FIG. 4 illustrates an example of training data set 170A, according to an embodiment. Specifically, the first 3 instances of data in the 170A training data indicate peptide presentation information from a single allele cell line involving the HLA-C * 01: 03 allele and 3 peptide sequences QCEIOWAREFLKEIGJ (SEQ ID NO: 10), FIEUHFWI (SEQ ID NC: 11) and FEWRHRITRUJR (SEQ ID NO: 12). The fourth instance of data in the 170A training data indicates peptide information from a multiple allele cell line involving the HLA-B * 07: 02, HLA-C * 01: 03, HLA-A * 01: 01 alleles and a sequence of QIEJOEIJE peptides (SEQ ID NO: 13). The first instance of data indicates that the QCEIOWARE peptide sequence (SEQ ID NO: 10) was not presented by the HLA-DRB3: 01: 01 allele. As discussed in the previous two paragraphs, the negatively labeled peptide sequences can be generated randomly by the data management 312 or identified from the source protein of the peptides presented. The 170A training data also includes a 1,000nM binding affinity prediction and a 1h half-life stability prediction for the peptide-allele sequence pair. The 170A training data also includes non-interactive allele variables, such as the C-terminal flanking sequence of the FIELFISBOSJFIE peptide (SEQ ID NO: 14) and a MRNA quantification measurement of 10º TPM. The fourth instance of data indicates that the QIEJOEIJE peptide sequence (SEQ ID NO: 13) was presented by one of the HLA-B * 07: 02, HLA-C * 01: 03, or HLA-A * 01 alleles:
[0328] [0328] The coding module 314 encodes the information contained in the training data 170 in a numerical representation that can be used to generate the one or more presentation models. In one implementation, the one-hot code strings from coding module 314 encode a hot sequence (for example, C-terminal peptide sequences or flanking sequences) over a predetermined 20-letter amino acid alphabet. Specifically, a p 'peptide sequence with k amino acids; is represented as a line vector of elements 20-k ;, where a single element between pi2o.6-1) +1, P'20-6-1) 12, - .., P'20 that corresponds to the alphabet of the amino acid in the jth position of the peptide sequence it has a value of 1. Otherwise, the remaining elements will have a value of 0. As an example, for a given alphabet (A, C, D, E, E, G,, IL K, L , M, N, P, Q, R, S, T, V, W, Yj, the 3-amino acid EAF peptide sequence for data instance i can be represented by the 60 element line vector p '= [0 0 0 1000000000000000010000000000000000000
[0329] [0329] When training data 170 contains sequences of different lengths of amino acids, coding module 314 can further encode peptides into vectors of equal length by adding a PAD character to extend the predetermined alphabet. For example, this can be accomplished by filling in the left peptide sequences with the character PAD until the length of the peptide sequence reaches the peptide sequence with the longest length in the training data 170. Thus, when the peptide sequence with the longest length has kmax amino acids , the coding module 314 numerically represents each sequence as a line vector of elements (20 + 1): kmax. As an example, for the extended alphabet (PAD, A, C, D, E, F, G, 1, LK, L, M, N, PR, Q, RS, T, V, W, Y j and a maximum length of amino acids of kmax = 5, the same example of the 3-amino acid EAF peptide sequence can be represented by the 105-element line vector pi = [10000000000000000000010000000000000000000000 100000000000000000100000000000000000000000010000 00000000000]. The C-terminal flanking sequence ci or other sequence data can be encoded from similarly as described above, so each independent variable or column in the peptide sequence p 'or c' represents the presence of a specific amino acid at a specific position in the sequence.
[0330] [0330] Although the above method for encoding sequence data has been described with reference to sequences that have amino acid sequences, the method can be extended in a similar way to other types of sequence data, such as DNA or RNA sequence data, and the like.
[0331] [0331] The coding module 314 also encodes the one or more alleles of MHC ai for data instance i as a line vector of elements m, in which each element h = 1, 2, ..., m corresponds to a unique identified MHC allele. The elements corresponding to the MHC alleles identified for data instance i have a value of 1. Otherwise, the remaining elements will have a value of 0. As an example, the alleles HLA-B * 07: 02 and HLA-DRBI1 * 10: 01 for a data instance 1 corresponding to a cell line of multiple alleles between m = 4 types of alleles of
[0332] [0332] The encoding module 314 also encodes the y mark; for each instance of data i as a binary variable with values from the set of (0, 13, in which a value of | indicates that the peptide x 'was presented by one of the MHC alleles associated with', and a value O indicates that peptide x 'was not presented by any of the MHC alleles associated with'. When the dependent variable y; represents the ion current of mass spectrometry, the 314 coding module can additionally scale the values using various functions, such as the function of log that has a range of (-2o, oo) for ion current values between [0, co).
[0333] [0333] The coding module 314 can represent a pair of xn 'allele interactive variables for the pi peptide; and an MHC allele associated with h as a line vector in which numerical representations of interactive allele variables are concatenated one after the other. For example, coding module 314 can represent xn 'as a line vector equal to ['], [ 'br], [' sn], or [ 'br' sh], where bn! is the prediction of binding affinity for the p peptide; and MHC allele associated with h, and likewise for sr 'for stability. Alternatively, one or more combinations of interactive allele variables can be stored individually (for example, as vectors or individual matrices).
[0334] [0334] In one example, the 314 encoding module represents binding affinity information incorporating measured or predicted values for binding affinity in allele xx interactive variables.
[0335] [0335] In one example, the 314 encoding module represents binding stability information incorporating measured or predicted values for binding stability in the xn 'allele interactive variables,
[0336] [0336] In one example, the coding module 314 represents information on binding rate incorporating measured or predicted values for binding rate on interactive xr allele variables.
[0337] [0337] In one example, for peptides presented by MHC class 1 molecules, the coding module 314 represents the length of the peptide as a vector Ti = [1 (L = 8) 1 (L = 29) 1 (L = 10) 1 (L = 11) 1 (Le12) 1 (Le13) 1 (L214) 1 (L215)] where 1 is the indicator function and Lx represents the length of the px peptide. The Tr vector can be included in the interactive xr allele variables. In another example, for peptides presented by MHC class II molecules, coding module 314 represents the length of the peptide as a vector Ti = [1 (Lx = 6) 1 (Lx = 7) 1 (Lx = 8) 1 (Lx = 9) 1 (L = 10) 1 (L = 11) 101212) 102813) 1028214) 10:15) 10216) 100217) 101218) 1 (L = 19) 101720) 1 (L = 21) 1 (L222) 1 (L223) 1 (L24) 1 (L 25) 1 (Lx226) 1 (Lx227) 1 (Lx = 28) 1 (Lx = 29) 1 (L «x = 30)] where 1 is the function of the indicator and Lx represents the length of the px peptide. The vector Tx can be included in the interactive variables of allele xa.
[0338] [0338] In one example, the 314 coding module represents RNA expression information from MHC alleles incorporating RNA-seq-based expression levels of MHC alleles into the xr allele interactive variables.
[0339] [0339] Likewise, coding module 314 can represent non-interactive allele variables wW "as a line vector in which numerical representations of non-interactive allele variables are concatenated one after the other. For example, w 'can be a line vector equal to [0 '] or [c' m 'w] where w' is a line vector representing any other non-interactive allele variables, in addition to the C-terminal flanking sequence of peptide p 'e the measurement of quantification of MBRNA m 'associated with the peptide Alternatively, one or more combinations of non-interactive allele variables can be stored individually (for example, as vectors or individual matrices).
[0340] [0340] In one example, coding module 314 represents the turnover rate of the source protein for a peptide sequence, incorporating the turnover rate or half-life in the non-interactive variables of the w allele.
[0341] [0341] In one example, the coding module 314 represents the length of the protein or isoform of origin, incorporating the length of the protein in the non-interactive variables of the w allele.
[0342] [0342] In one example, coding module 314 represents the activation of the immunoproteasome, incorporating the mean expression of the specific proteasome subunits for the immunoproteasome, including the B1 ;, B2 ;, B5 subunits; in the non-interactive variables of w allele.
[0343] [0343] In one example, coding module 314 represents the RNA-seq abundance of the source protein of the peptide or gene or transcription of a peptide (quantified in units of FPKM, TPM by techniques such as RSEM) may be incorporating the abundance of source protein in the non-interactive variables of wi allele.
[0344] [0344] In one example, coding module 314 represents the probability that the source transcript of a peptide will undergo nonsense mediated decay (NMD), as estimated by the model in, for example, Rivas et. al. Science, 2015, incorporating this probability into the non-interactive variables of wi allele.
[0345] [0345] In one example, the 314 coding module represents the activation status of a module or gene pathway evaluated via RNA-sec, for example, by quantifying the expression of genes in the pathway in TPM units using, for example, RSEM for each of the genes in the pathway then calculates a summary statistic, for example, the average, through the genes in the pathway. The mean can be incorporated into the non-interactive allele w "variables.
[0346] [0346] In one example, the coding module 314 represents the copy number of the source gene, incorporating the copy number in the non-interactive variables of the wi allele.
[0347] [0347] In one example, coding module 314 represents TAP binding affinity including measured or predicted TAP binding affinity (for example, in nanomolar units) in the non-interactive w allele variables.
[0348] [0348] In one example, coding module 314 represents the levels of TAP expression including the levels of TAP expression measured by RNA-seq (and quantified in TPM units by, for example, RSEM) in the non-interactive variables of allele w ".
[0349] [0349] In one example, coding module 314 represents tumor mutations as a vector of indicator variables (ie, d * = 1 if the p * peptide comes from a sample with a KRAS GI12D and O otherwise) mutation in the variables non-interactive allele wi.
[0350] [0350] In one example, the 314 coding module represents germline polymorphisms in antigen presenting genes as a vector of indicator variables (ie d * = 1 if the p * peptide comes from a sample with a specific polymorphism germ line in TAP). These indicator variables can be included in the non-interactive variables of allele w '.
[0351] [0351] In one example, the encoding module 314 represents the tumor type as a vector encoded by one-hot of length one over the alphabet of tumor types (for example, NSCLC, melanoma, colorectal cancer, etc.). These one-hot coded variables can be included in the wi non-interactive allele variables.
[0352] [0352] In one example, the 314 encoding module represents suffixes from the MHC allele treating 4-digit HLA alleles with different suffixes. For example, HLA-A * 24: 09N is considered a different allele than HLA-A * 24: 09 for the purpose of the model. Alternatively, the likelihood of presentation by an MHC allele with an N suffix can be set to zero for all peptides, because the HLA alleles ending with the N suffix are not expressed.
[0353] [0353] In one example, coding module 314 represents the tumor subtype as a one-hot vector encoded in length one on the tumor subtype alphabet (e.g., lung adenocarcinoma, squamous cell carcinoma of the lung, etc). These one-hot coded variables can be included in the non-interactive allele w "variables.
[0354] [0354] In one example, the coding module 314 represents the smoking history as a binary indicator variable (d * = 1 if the patient has a smoking history and O otherwise), which can be included in the non-interactive variables of allele w. Alternatively, the smoking history can be encoded as a one-hot coded variable of length one over an alphabet with the severity of smoking. For example, smoking status can be rated on a scale of 1 to 5, where 1 indicates non-smokers and 5 indicates current heavy smokers. As the smoking history is mainly relevant for lung tumors, when training a model in various types of tumors, this variable can also be set to equal 1 if the patient has a smoking history and the tumor type is lung tumors and zero otherwise.
[0355] [0355] In one example, coding module 314 represents the history of sunburn as a binary indicator variable (d “= 1 if the patient has a history of severe sunburn and O otherwise), which can be included in the non-interactive allele variables w. Since severe sunburn is particularly relevant for melanomas, when training a model on various types of tumors, this variable can also be set to equal 1 if the patient has a history of severe sunburn and the type of tumor is melanoma and zero, otherwise.
[0356] [0356] In one example, coding module 314 represents the distribution of the expression levels of a specific gene or transcript for each gene or transcription in the human genome as summary statistics (for example, mean, median) of the distribution of the expression levels using reference databases like TCGA. Specifically, for a p * peptide in a sample with tumor-like melanoma, we can include not only the measured level of expression of the gene or transcription of the gene or transcription of the origin of the p * peptide in the non-interactive variables of the w 'allele, but also the average and / or median expression of the gene or transcription of the gene or transcription of the origin of the p * peptide in melanomas measured by the TCGA.
[0357] [0357] In one example, coding module 314 represents the mutation type as a one-hot coded variable of length one over the alphabet of mutation types (eg missense, frame shift, NMD induction, etc. .). These one-hot coded variables can be included in the wi non-interactive allele variables.
[0358] [0358] In one example, the 314 coding module represents protein characteristics at the protein level as the annotation value (eg 5 'RTU length) of the source protein in the non-interactive w allele variables. In another For example, coding module 314 represents residue-level annotations from the source protein to the peptide pi including an indicator variable, which is equal to 1 if peptide p 'overlaps with a helix motif and O, otherwise, or is equal to 1 if the p 'peptide is completely contained within a helix motif in the non-interactive variables of allele W. In another example, a characteristic that represents the proportion of residues in the pi peptide that are contained in a motif annotation helix can be included in the non-interactive allele variables w ..
[0359] [0359] In one example, the coding module 314 represents the type of proteins or isoforms in the human proteome as an indicator vector 0º that has a length equal to the number of proteins or isoforms in the human proteome and the corresponding element o; is 1 if peptide p * comes from protein i and 0, otherwise.
[0360] [0360] In one example, coding module 3 14 represents the source gene G = gene (p)) of the peptide p 'as a categorical variable with L possible categories, where L represents the upper limit of the number of originating genes indexed 1,2, ..., L.
[0361] [0361] Coding module 314 can also represent the general set of variables z 'for peptide p' and an associated MHC allele h as a line vector in which numerical representations of the interactive variables of allele x 'and the variables not allele w 'interactions are concatenated one after the other. For example, the coding module 314 can represent zh as a line vector equal to [xn 'wW] or [wi xn].
[0362] [0362] Training module 316 builds one or more presentation models that generate probabilities of whether the peptide sequences will be presented by the MHC alleles associated with the peptide sequences. cSpecifically, given a p * peptide sequence and a set of MHC af alleles associated with the peptide sequence ”, each presentation model generates an ur estimate indicating a probability that the p * peptide sequence will be presented by one or more associated MHC alleles The*.
[0363] [0363] Training module 316 builds another presentation model based on the training data sets stored in reserve 170 generated from presentation information stored in 165. Generally, regardless of the specific type of presentation model, all presentation models capture the dependency between independent variables and dependent variables in the training data 170, so that a loss function is minimized. Specifically, the loss function f (yies, uies; O) represents discrepancies between values of dependent variables yies for one or more instances of data S in training data 170 and the estimated probabilities uies for instances of data S generated by the model of presentation. In a specific implementation referred to throughout the rest of the specification, the loss function (yies, uies; 0) is the negative log probability function given by equation (1a) as follows: parties: 0) = X Or logui + A y0log (1 = ud). (a) ies However, in practice, another loss function can be used. For example, when predictions are made for the ion current of mass spectrometry, the loss function is the mean quadratic loss given by equation 1b as follows: fO'es. mes: 0) = D Ulyi = ul). (ab) ies
[0364] [0364] The presentation model can be a parametric model in which one or more 0 parameters mathematically specify the dependence between the independent variables and the dependent variables. Typically, various parameters of parametric presentation models that minimize the loss function (yies, uies; 0) are determined using gradient-based numerical optimization algorithms, such as batch gradient algorithms, stochastic gradient algorithms and the like . Alternatively, the presentation model can be a non-parametric model in which the structure of the model is determined from training data 170 and is not strictly based on a fixed set of parameters. VIII. B. Models by allele
[0365] [0365] Training module 316 can build presentation models to predict the probabilities of presenting peptides on an allele basis. In this case, training module 316 can train presentation models based on instances of S data in training data 170 generated from cells expressing unique MHC alleles.
[0366] [0366] In one implementation, training module 316 models the estimated likelihood of presentation ux for peptide p * for a specific allele h: uk = Pr (ptpresented; MHCalleleh) = f (gnGxÉ: 9n)), 2)
[0367] [0367] The output of the gn dependency function (xn5; Ohn) represents a dependency score for the MHC h allele indicating whether the MHC h allele will present the corresponding neoantigen based on at least the interactive characteristics of the xx “allele, and , in particular, based on the amino acid positions of the peptide sequence of the p * peptide. For example, the dependency score for the MHC h allele can be high if the MHC h allele is likely to have the p1 peptide, and it can be low if the presentation is not likely. The transformation function f -) transforms the input and, more specifically, transforms the dependency score generated by gn (xn '; On) in this case, with an appropriate value to indicate the probability that the p * peptide will be presented by a MHC allele.
[0368] [0368] In a specific implementation referred to throughout the rest of the specification, f (-) is a function that has the range within [0, 1] for an appropriate domain range. In one example, f (-) is the output function given by: exp (2) fo = Irem (D '(4) As another example, f (-) can also be the hyperbolic tangent function given by: f (2 ) = tanh (z) (5) when the values for the z domain are equal to or greater than 0. Alternatively, when predictions are made for the ion current of mass spectrometry that have values outside the range [0, 1], f ()), can be any function, such as the identity function, the exponential function, the log function and the like.
[0369] [0369] Thus, the probability per allele that a p * peptide sequence will be presented by an MHC h allele can be generated by applying the gn (-) dependency function for the MHC h allele to the coded version of the p * peptide sequence to generate the corresponding dependency score. The dependency score can be transformed by the transformation function f (-) to generate a probability by allele that the peptide sequence p * will be presented by the allele of MHC h. VIIL.B.1 Dependency Functions for Interactive Allele Variables
[0370] [0370] In a specific implementation referred to throughout the specification, the dependency function gn (-) is a related function provided by: gnlxh; 97) = xh "On. (6) which linearly combines each xn * allele interactive variable with a corresponding parameter in the On parameter set determined for the associated MHC allele h.
[0371] [0371] In another specific implementation mentioned throughout the specification, the dependency function gn (*) is a network function provided by: gnlxi; 9%) = NN, (x); 9%). O) represented by a network model NNh (-) having a series of nodes arranged in one or more layers. One node can be connected to other nodes through connections, each with an associated parameter in the hn parameter set. A value on a specific node can be represented as a sum of the values of the nodes connected to the specific node weighted by the associated parameter mapped by an activation function associated with the specific node. In contrast to the related function, network models are advantageous because the presentation model can incorporate non-linearity and process data with different lengths of amino acid sequences. Specifically, through nonlinear modeling, network models can capture the interaction between amino acids at different positions in a peptide sequence and how that interaction affects the presentation of the peptide.
[0372] [0372] In general, NNn (-) network models can be structured as feed-forward networks, such as artificial neural networks (ANN), convolutional neural networks (CNN), deep neural networks (DNN) and / or recurrent networks, such as long short-term memory networks (LSTM), bidirectional recurring networks, deep bidirectional recurring networks and the like.
[0373] [0373] In an example referred to throughout the rest of the specification, each MHC allele at h = 1,2, ..., m is associated with a separate network model, and NNh (-) represents the results of a network model associated with the MHC allele h.
[0374] [0374] FIG. 5 illustrates an example of an NN3 (*) network model in association with an arbitrary MHC h = 3 allele. As shown in FIG. 5, the NN3 (-) network model for the MHC allele h = 3 includes three input nodes at layer 1 = 1, four nodes at layer | = 2, two nodes in layer 1 = 3 and an output node in layer 1 = 4. The NN3 (-) network model is associated with a set of ten parameters 63 (1), 83 (2), ..., 8: (10). The NN3 network model (-) receives input values (individual data instances, including encoded polypeptide sequence data and any other training data used) for three interactive allele variables x3 (1), x3 “(2), and x3 "(3) for the MHC allele h = 3 and generates the NN3 value (x3º). The network function can also include one or more network models, each having different interactive allele variables as input.
[0375] [0375] In another example, the MHC alleles identified h = 1, 2, ..., m are associated with a single network model NNn (-), and NNn (*) represents one or more outputs of the network model unique associated with the MHC h allele. In this case, the set of parameters On may correspond to a set of parameters for the single network model and, therefore, the set of parameters On may be shared by all MHC alleles.
[0376] [0376] FIG. 6A illustrates an example of a network model NNn () shared by MHC alleles h = 1,2, ..., m. As shown in FIG. 6A, the NNn (:) network model includes output nodes m, each corresponding to an MHC allele. The NN3 (-) network model receives the interactive variables of allele x3 * for the MHC allele h = 3 and generates m values including the NN3 (x3) value corresponding to the MHC allele h = 3.
[0377] [0377] In another instance, the single network model NNnH () can be a network model that generates a dependency score, given the xn * allele interactive variables and the dn encoded protein sequence of an MHC h allele. In this case, the set of parameters On can again correspond to a set of parameters for the single network model and, therefore, the set of parameters hn can be shared by all MHC alleles. Thus, in such a case, NNn (-) can represent the output of the inputs given from the single network model NNu (-) [xn "dh] to the single network model. This network model is advantageous because the probabilities of presenting peptides for MHC alleles that were unknown in the training data can be predicted only by identifying their protein sequence.
[0378] [0378] FIG. 6B illustrates an example of an NNn () network model shared by MHC alleles. As shown in FIG. 6B, the NNn () network model receives the allele interactive variables and the protein sequence of the MHC allele h = 3 as input and generates an NN3 (x3º) dependency score corresponding to the MHC allele h = 3.
[0379] [0379] In yet another example, the gn (-) dependency function can be expressed as: gInlxK; On) = 9'n (K; 07) + OR where g'n (xn; 0 "n) is the affine function with a set of 86ºn parameters, the network function or similar, with an Ohº bias parameter in the set of parameters for interactive allele variables for the MHC allele that represents a baseline probability of presentation for the MHC allele h.
[0380] [0380] In another implementation, the 8hº bias parameter can be shared according to the MHC h allele gene family. That is, the bias parameter 614º for the MHC h allele can be equal to Ogenem) º, where gene (h) is the family of genes of the MHC h allele. For example, the MHC class I alleles HLA-A * 02: 01, HLA-A * 02: 02, and HLA-A * 02: 03 can be assigned to the “HLA-A” gene family and the bias parameter Ohnº for each of these MHC alleles can be shared. As another example, the MHC class IT alleles HLA-DRBI1: 10: 01, HLA-DRBI: 11: 01, and HLA-DRB3: 01: 01 can be assigned to the “HLA-DRB” gene family and the Onº bias for each of these MHC alleles can be shared.
[0381] [0381] Going back to equation (2), as an example, the probability that the peptide p * will be presented by the MHC allele h = 3, between m = 4 different MHC alleles identified using the affinity dependence function gn (-) , can be generated by: ut = f (x5-03),
[0382] [0382] As another example, the probability that the p * peptide will be presented by the MHC allele h = 3, between m = 4 different identified MHC alleles, using separate gn (-) network transformation functions, can be generated by: ut = F (NNs3GE; 93)), where x3 * are the interactive allele variables identified for the MHC allele h = 3 and 03 are the set of parameters determined for the NN3 (-) network model associated with the allele MHC h = 3.
[0383] [0383] FIG. 7 illustrates the generation of a presentation probability for the peptide p * in association with the MHC allele h = 3 using an example network model NN; (:).). As shown in FIG. 7, the network model NN3 (-) receives the interactive variables of allele x3 “for the MHC allele h = 3 and generates the output NN3 (x35). The output is mapped by the function f (-) to generate the estimated probability of presentation ux.
[0384] [0384] In one implementation, training module 316 incorporates non-interactive allele variables and models the estimated likelihood of presentation ux for peptide p * by: uk = Pr (ptpresented) = f (gu CW "Ow) + gnlxih; The%)) (8) where wt represents the non-interactive allele variables encoded for the pÍ peptide, gw (-) is a function for the non-interactive allele variables w * based on a set of parameters 0, determined for the non-interactive allele variables Specifically, the values for the On parameter set for each MHC allele h and the O parameter set for interactive allele variables can be determined by minimizing the loss function with respect to On and Ow, where i is each subset S instance of training data 170 generated from cells expressing unique MHC alleles.
[0385] [0385] The output of the gw (w "; Ow) dependency function represents a dependency score for non-interactive allele variables, indicating whether the p * peptide will be presented by one or more MHC alleles based on the impact of variables for example, the dependency score for non-interactive allele variables can be high if the p * peptide is associated with a C-terminal flanking sequence that is known to positively impact the presentation of the p peptide ", And may have a low value if the p * peptide is associated with a C-terminal flanking sequence that is known to negatively impact the presentation of the p peptide".
[0386] [0386] According to equation (8), the probability per allele that a peptide sequence p * will be presented by an MHC h allele can be generated by applying the function gn (-) for the MHC h allele to the coded version of the p * peptide sequence to generate the corresponding dependency score for interactive allele variables. The gw (-) function for non-interactive allele variables is also applied to the coded version of non-interactive allele variables to generate the dependency score for non-interactive allele variables. Both scores are combined, and the combined score is transformed by the transformation function f (-) to generate a probability by allele that the peptide sequence p * will be presented by the MHC allele h.
[0387] [0387] Alternatively, training module 316 can include non-interactive variables of allele w * in the prediction by adding non-interactive variables of allele w "to non-interactive variables of allele xr" in equation (2). Thus, the probability of presentation can be given by: uk = Pr (ptapresented; allele) = f (gn (xkw "]; O). (9) VIII. B. 3 Dependency Functions for non-interactive allele variables
[0388] [0388] Just like the gn (-) dependency function for allele interactive variables, the gw (-) dependency function for non-interactive allele variables can be an affine function or a network function in which a model of separate network is associated with non-interactive w allele variables.
[0389] [0389] Specifically, the gw (-) dependency function is an affine function given by: In (W "5; 8,) = wc Oy. Which linearly combines the non-interactive allele variables in w * with a corresponding parameter in set of Ow parameters.
[0390] [0390] The dependency function gw (-) can also be a network function provided by: gn (w "; 8,) = NN, (W"; 0,). represented by an NNw (") network model with an associated parameter in the O, parameter set. The network function can also include one or more network models, each having different non-interactive allele variables as input.
[0391] [0391] In another case, the gw (") dependency function for non-interactive allele variables can be given by: Iw (W" 5 9) = 9 OW "8,) + hm"; 65), (10) where g'w (w "; 0'w) is the affine function, the network function with the set of non-interactive parameters of allele 0ºw, or similar, m * is the measure of quantification of MRNA for the p peptide ", h (-) is a function that transforms the measurement of quantification and 0," is a parameter in the set of parameters for non-interactive allele variables that is combined with the measurement of quantification of mMRNA to generate a score of dependence for measuring MRNA quantification. In a specific modality referred to throughout the rest of the specification, h (-) is the log function, however, in practice, h () can be any of a variety of different functions.
[0392] [0392] In yet another example, the gyw (-) dependency function for non-interactive allele variables can be given by: Iw WE 9) = 958) +08: 0 ", (655) where g'w (w "; 80'w) is the affine function, the network function with the set of non-interactive parameters of allele 9º, or similar, oº is the indicator vector described in Section VII. C. 2, which represents proteins and isoforms in the human proteome for p “peptides, and 04º is a set of parameters in the set of parameters for non-interactive allele variables that is combined with the indicator vector. In a variation, when the dimensionality of ot and the set of parameters 8º are significantly high, a parameter regularization term, such as À + | J9Z |, where | - | represents the L1 norm, the L2 norm, a combination or the like, can be added to the loss function when determining the value of the parameters. The optimal value of hyperparameter X can be determined using appropriate methods.
[0393] [0393] In yet another example, the gw () dependency function for non-interactive allele variables can be given by:
[0394] [0394] In practice, the additional terms of any of equations (10), (11) and (12) can be combined to generate the gw (-) dependency function for non-interactive allele variables. For example, the term h (-) indicating measurement of mMRNA quantification in equation (10) and the term indicating antigenicity of the source gene in equation (12) can be added together with any other related or network function to generate the function of dependence for non-interactive allele variables.
[0395] [0395] Going back to equation (8), as an example, the probability that the peptide p * will be presented by the MHC allele h = 3, between m = 4 different MHC alleles identified using the related transformation functions gn (*) , g & w ("), can be generated by: uz = f (w" -a, + x1- 03), where wt are the non-interactive allele variables identified for the p “peptide, and Ow are the set of parameters determined for non-interactive allele variables.
[0396] [0396] As another example, the probability of peptide p * will be presented by the MHC allele h = 3, between m = 4 different identified MHC alleles, using network transformation functions gn (), gw (), can be generated by: uz = FINN, GW "; 9,) + NN3 (5; 93)) where w" are the interactive allele variables identified for the p “peptide, and 8w are the set of parameters determined for the non-interactive variables of allele.
[0397] [0397] FIG. 8 illustrates the generation of a presentation probability for the p * peptide in association with the MHC allele h = 3 using example network models
[0398] [0398] Training module 316 can also build presentation models to predict the probabilities of peptide presentation in a multiple allele scenario, where two or more MHC alleles are present. In this case, the training module 316 can train the presentation models based on the S data instances in the training data 170 generated from cells that express single MHC alleles, cells that express multiple MHC alleles or a combination of them . VIIL.C.1. Example 1: Maximum Models Per Allele
[0399] [0399] In one implementation, training module 316 models the estimated presentation probability ux for the p * peptide in association with a set of multiple MHC H alleles as a function of the u and presentation probabilities determined for each of the MHC alleles h in set H determined based on cells expressing unique alleles, as described above in conjunction with equations (2) - (11). Specifically, the probability of presentation ux can be any function of eu ". In an implementation, as shown in equation (12), the function is the maximum function and the probability of presentation ur can be determined as the maximum probability of presentation for each MHC allele h in the set H. ur = Pr (represented p; H allele) = max (ur ""). VIII. C. 2. Example 2. 1: Sum Function Models
[0400] [0400] In one implementation, training module 316 models the estimated likelihood of presentation ux for peptide p “by: m ur, = Pr (p" shown) => ak. Gnlxk; on). (13) h = 1 where the elements ar "are 1 for the multiple MHC H alleles associated with the peptide sequence p * and xn * represents the interactive allele variables encoded for the p * peptide and the corresponding MHC alleles. The values for the parameter set On for each MHC allele h can be determined by minimizing the loss function in relation to Gn, where i is each instance of subset S of training data 170 generated from cells that express unique MHC alleles and / or cells that express multiple MHC alleles The gn dependency function can be in the form of any of the gn dependency functions introduced above in sections VIII. B. 1.
[0401] [0401] According to equation (13), the probability of presenting that a peptide sequence p * will be presented by one or more alleles of MHC h can be generated by applying the dependency function gn (-) for the coded version of p * peptide sequence for each of the MHC H alleles to generate the corresponding score for the interactive allele variables. The scores for each MHC h allele are combined and transformed by the f (-) transformation function to generate the probability of presentation that the p * peptide sequence will be presented by the MHC H allele set.
[0402] [0402] The model for presenting equation (13) is different from the allele model of equation (2), in which the number of associated alleles for each p * peptide can be greater than 1. In other words, more than one element in an “can have values of 1 for the multiple MHC H alleles associated with the p * peptide sequence.
[0403] [0403] As an example, the probability that the p * peptide will be presented by the MHC alleles h = 2, h = 3, between m = 4 different identified MHC alleles, using the related gn (-) transformation functions, can be generated by: ur = f (x3-02 + x5- 83), where x2º, x3 * are the interactive allele variables identified for MHC alleles h = 2, h = 3 and 07.03 are the set of parameters determined for MHC alleles h = 2, h =
[0404] [0404] As another example, the probability that peptide p * will be presented by the MHC allele h = 2, h = 3, between m = 4 different identified MHC alleles, using network transformation functions gn (-), gw (), can be generated by: u, = F (NN2 (X5; 92) + NN3 (X5; 93)),
[0405] [0405] FIG. 9 illustrates the generation of a presentation probability for the p * peptide in association with MHC alleles h = 2, h = 3 using sample network models NN2 (-) and NN3 (:). As shown in FIG. 9, the network model NN2 (-) receives the interactive variables of allele x2 * for the MHC allele h = 2 and generates the output NNo (x29 and the network model NN3 (-) receives the interactive variables of allele x3 * for the MHC allele h = 3 and generates the NN3 output (x3 *) .The outputs are combined and mapped by the function f (-) to generate the estimated probability of presentation ux.
[0406] [0406] In one implementation, training module 316 incorporates non-interactive allele variables and models the estimated likelihood of presentation ux for peptide p * by: mu, = Pr (ptpresented) = 1 (sn: 9.) +> af: gn (xk; 2) (14) h = 1 where w represents the non-interactive allele variables encoded for the peptide pt. Specifically, the values for the On parameter set for each MHC allele h and the O parameter set, for interactive allele variables can be determined by minimizing the loss function with respect to On and Ow, where i is each instance of the S subset of the data training 170 generated from cells expressing single MHC alleles and / or cells expressing multiple MHC alleles. The gw dependency function can be in the form of any of the gw dependency functions introduced above in sections VIII. B. 3.
[0407] [0407] Thus, according to equation (14), the probability of presenting that a p * peptide sequence will be presented by one or more MHC H alleles can be generated by applying the gn (*) function for the coded version of p * peptide sequence for each of the MHC H alleles to generate the corresponding dependency score for the interactive allele variables for each MHC h allele. The gvw () function for non-interactive allele variables is also applied to the coded version of non-interactive allele variables to generate the dependency score for non-interactive allele variables. The scores are combined and the combined score is transformed by the transformation function f (-) to generate the probability of presentation that the peptide sequence p * will be presented by the MHC H alleles.
[0408] [0408] In the equation (14) presentation model, the number of associated alleles for each p * peptide can be greater than 1. In other words, more than one element in an “can have values of 1 for the multiple alleles of MHC H associated with the pº peptide sequence.
[0409] [0409] As an example, the probability that the p * peptide will be presented by the MHC alleles h = 2, h = 3, between m = 4 different identified MHC alleles, using the related transformation functions gn (*), gw (-), can be generated by: uz = f (wt-0, + x5-0, + x5: 83), where w "are the non-interactive allele variables identified for the p“ peptide, and Ow are the set of parameters determined for non-interactive allele variables.
[0410] [0410] As another example, the probability that peptide p * will be presented by the MHC allele h = 2, h = 3, between m = 4 different identified MHC alleles, using network transformation functions gn (), gw ( "), can be generated by: ur = FINN, (W" 59,) + NN2 (X7; 02) + NN3 (x3; 03)) where wi are the interactive allele variables identified for the p * peptide, and Ow they are the set of parameters determined for non-interactive allele variables.
[0411] [0411] FIG. 10 illustrates the generation of a presentation probability for the p * peptide in association with MHC alleles h = 2, h = 3 using sample network models NN2 (), NN :()), and NNu ()). As shown in FIG. 10, the network model NN2 () receives the interactive variables of allele x2 * for the MHC allele h = 2 and generates the output NN2 (x29). The red model NN3 (-) receives the interactive variables of allele x3 * for the MHC allele h = 3 and generates the output NN3 (x35). The NNv (-) network model receives the non-interactive variables of the w "allele for the p * peptide and generates the NNw (w") output. The outputs are combined and mapped by the function f (-) to generate the estimated probability of presentation ux.
[0412] [0412] Alternatively, training module 316 can include non-interactive variables of allele w * in the prediction by adding non-interactive variables of allele w "to non-interactive variables of allele xn * in equation (15). Thus, the probability of presentation can be given by:
[0413] [0413] In another implementation, training module 316 models the estimated probability of presentation ux for peptide p * by: ur = Pr (pYpresented) = r (st = | at -ut (0) .. ak uF) )), (16) where the elements an "are 1 for the multiple alleles of MHC h EH associated with the peptide sequence p", u'x "is an implicit probability of presentation per allele for the allele of MHC h, vector v is a vector in which the element vn corresponds to an “* uk, s (-) is a function that maps the elements to see () is a fixation function that fixes the value of the input in a given interval. As described in more detail below, s (-) can be the sum function or the second order function, but it is appreciated that in other modalities, s (.) Can be any function, such as the maximum function. The values for the set of parameters O for the implied probabilities per allele can be determined by minimizing the loss function with respect to 0, where i is each instance of the subset S of the training data 170 generated from cells expressing unique alleles of MHC and / or cells that express multiple MHC alleles.
[0414] [0414] The presentation probability in the presentation model of equation (17) is modeled as a function of the implied presentation probabilities by allele u'Yº that each corresponds to the probability that the p * peptide will be presented by an individual allele of MHC h. The implied probability by allele is different from the probability of presentation by allele in section VIII. B, insofar as the parameters for the implied probability per allele can be learned in several allele configurations, in which the direct association between a presented peptide and the corresponding MHC allele is unknown, in addition to the single allele configurations. Thus, in a scenario of multiple alleles, the presentation model can estimate not only whether the p * peptide will be presented by a set of MHC H alleles as a whole, but also provide individual u'v "E! Probabilities that indicate which allele of MHC h probably presented the p * peptide. One advantage of this is that the presentation model can generate the implied probabilities without training data for cells expressing unique MHC alleles.
[0415] [0415] In a specific implementation referred to throughout the rest of the specification, r (*) is a function with the range [0, 1]. For example, r (-) can be the fixing function: r (z) = min (max (z, 0), 1), where the minimum value between z and | is chosen as the probability of ur presentation. In another implementation, r (-) is the hyperbolic tangent function given by: r (z) = tanh (z) when the values for the z domain are equal to or greater than O. VIII. C. 5. Example 3. 2: Sum of Functions Model
[0416] [0416] In a specific implementation, s (-) is a sum function and the probability of presentation is given by the sum of the implicit probabilities of presentation by allele: m u, = Pr (represented) = r> af. uk 9) A7) h = 1
[0417] [0417] In an implementation, the implied probability of presentation per allele for the MHC h allele is generated by: uj "= f (gn (xE 01), (as) so that the presentation probability is estimated by: m ur = Pr (pYpresented) = r> ak f (gnlxk;% v)) as) h = 1
[0418] [0418] According to equation (19), the probability of presenting that a pt peptide sequence will be presented by one or more MHC H alleles can be generated by applying the gn (-) function to the coded version of the p peptide sequence * for each of the MHC H alleles to generate the corresponding dependency score for interactive allele variables. Each dependency score is first transformed by the function f (-) to generate implicit probabilities of presentation by allele uh. The probabilities by u'x allele are combined and the fixation function can be applied to the combined probabilities to fix the values in an interval [0, 1] to generate the probability of presentation that the peptide sequence p * will be presented by the set of MHC H alleles. The gn dependency function can be in the form of any of the gn dependency functions introduced above in sections VIII. B. 1.
[0419] [0419] As an example, the probability that the p * peptide will be presented by the MHC alleles h = 2, h = 3, between m = 4 different identified MHC alleles, using the related gn (-) transformation functions, can be generated by: ur = r (f (xh- 62) + f (x -03)), where x2 ", x3º are the allele interactive variables identified for MHC alleles h = 2, h = 3 and 07.03 are the set of parameters determined for MHC alleles h = 2, h =
[0420] [0420] As another example, the probability that peptide p * will be presented by the MHC allele h = 2, h = 3, between m = 4 different identified MHC alleles, using network transformation functions gn (), gw ( "), can be generated by: ur = r (f (NN2 (5; 92)) + F (NN3G5; 03))), where NN2 (-), NN3 (-) are the network models identified for alleles of MHC h = 2, h = 3 and 07.03 are the set of parameters determined for MHC alleles h = 2, h = 3.
[0421] [0421] FIG. 11 illustrates the generation of a presentation probability for the p * peptide in association with MHC alleles h = 2, h = 3 using sample network models NN2 (-) and NN :( :). As shown in FIG. 9, the network model NN2 (-) receives the interactive variables of allele x2 * for the MHC allele h = 2 and generates the output NNo (x29 and the network model NN3 (-) receives the interactive variables of allele x3 * for the MHC allele h = 3 and generates the NN3 output (x35), each output is mapped by the function f (-) and combined to generate the estimated probability of presentation ux.
[0422] [0422] In another implementation, when predictions are made for the ion current log of mass spectrometry, r (-) is the logarithmic function and f () is the exponential function.
[0423] [0423] In an implementation, the implied probability of presentation by allele for the MHC h allele is generated by:
[0424] [0424] According to equation (21), the probability of presenting that a p * peptide sequence will be presented by one or more MHC H alleles can be generated by applying the gn (-) function to the coded version of the peptide sequence p * for each of the MHC H alleles to generate the corresponding dependency score for the interactive allele variables for each MHC h allele. The gw (-) function for non-interactive allele variables is also applied to the coded version of non-interactive allele variables to generate the dependency score for non-interactive allele variables. The score for non-interactive allele variables is combined with each of the dependency scores for interactive allele variables. Each of the combined scores is transformed by the function f (-) to generate the implicit probabilities of presentation by allele. The implied probabilities are combined and the fixation function can be applied to the combined outputs to fix the values in an interval [0.1] to generate the probability of presentation that the peptide sequence p * will be presented by the MHC H alleles. gw dependency function can be in the form of any of the gw dependency functions introduced above in sections VIII. B. 3.
[0425] [0425] As an example, the probability that the peptide "p" will be presented by the MHC alleles h = 2, h = 3, between m = 4 different identified MHC alleles, using the related transformation functions gn (*), gw (-), can be generated by: u, = r (fw "O, + x5- 802) + f (W" C0, + xb- 93), where w * are the non-interactive allele variables identified for the peptide p “, and 9 are the set of parameters determined for non-interactive allele variables.
[0426] [0426] As another example, the probability that peptide p * will be presented by the MHC allele h = 2, h = 3, between m = 4 different identified MHC alleles, using gn (-), gw network transformation functions ("), can be generated by: up = r (FONN, (W 0,) + NN2 (X8; 02)) + F (NN, (WS 0,) + NN3 (É; 03)))
[0427] [0427] FIG. 12 illustrates the generation of a presentation probability for the p * peptide in association with MHC alleles h = 2, h = 3 using sample network models NN2 (), NN :()), and NNu ()). As shown in FIG. 12, the network model NN2 () receives the interactive variables of allele x2 * for the MHC allele h = 2 and generates the output NN2 (x25). The NNyv (-) network model receives the non-interactive variables of the w * allele for the p * peptide and generates the NNyw (w ") output. The outputs are combined and mapped by the f () function. The NN3 ( -) receives the non-interactive variables of allele x3 * for the MHC allele h = 3 and generates the NN3 output (x3), which is again combined with the NNw (wW ") output of the same NNyw (-) network model and mapped by the f () function. Both outputs are combined to generate the estimated probability of ux presentation.
[0428] [0428] In another implementation, the implied probability of presentation by allele for the MHC h allele is generated by: uj "= f (gn (liw'1: 01)). 22) so that the probability of presentation is generated by: m ur = Pr (pYpresented) = r> ak f (gn (lxkw "]; ow) h = 1 VIII. C. 7. Example 4: Second Order Models
[0429] [0429] In an implementation, s (-) is a second order function and the estimated probability of presentation ux for peptide p * is given by: u, = Pr (ptpresented) mm = X atufio) - ND abatuf (o ) uf (o) (23) h = 1 h = 1 j <h where the elements u'Yº are the implicit probability of presentation by allele for the MHC h allele. The values for the set of parameters 0 for the implied probabilities per allele can be determined by minimizing the loss function with respect to 90, where i is each instance of the subset S of the training data 170 generated from cells that express unique alleles of MHC and / or cells that express multiple MHC alleles. The implicit probabilities of presentation by allele can be in any form shown in equations (18), (20) and (22) described above.
[0430] [0430] In one aspect, the model in equation (23) may imply that there is a possibility that the p * peptide is presented by two MHC alleles simultaneously, in which the presentation by two HLA alleles is statistically independent.
[0431] [0431] According to equation (23), the probability of presentation that a p * peptide sequence will be presented by one or more MHC H alleles can be generated by combining the implicit probabilities of presentation by allele and subtracting the probability that each pair of MHC alleles will simultaneously display the summation p * peptide to generate the probability of presentation that the p * peptide sequence will be presented by the MHC H alleles.
[0432] [0432] As an example, the probability that the p * peptide will be presented by the HLA alleles h = 2, h = 3, between m = 4 different identified HLA alleles, using the related gn (*) transformation functions, can be generated by: ur = f (x5-02) + f (x5-03) -— f (x5-02) - f (x5-03), where x2º, x3 * are the interactive variables of alleles identified for alleles of HLA h = 2, h = 3 and 07.03 are the set of parameters determined for HLA alleles h = 2, h =
[0433] [0433] As another example, the probability that p * peptide will be presented by the HLA allele h = 2, h = 3, between m = 4 different identified HLA alleles, using gn () network transformation functions, Sw ( "), can be generated by: ur = F (NN2 (x3; 02)) + F (NN3 (X5; 03)) - F (NN2 (X3; 92)) - F (NN3 (X3; 03)), where NN2 (), NN; (-) are the network models identified for HLA alleles h = 2, h = 3 and 07.03 are the set of parameters determined for HLA alleles h = 2, h = 3. IX Example 5: Prediction Module
[0434] [0434] Prediction module 320 receives sequence data and selects candidate neoantigens in the sequence data using the presentation models. Specifically, the sequence data can be DNA sequences, RNA sequences and / or protein sequences extracted from tumor tissue cells from patients. Prediction module 320 processes the sequence data into a plurality of p * peptide sequences with 8-15 amino acids for MHC-I or 6-30 amino acids for MHC-II. For example, the prediction module 320 can process the given sequence “IEFROEIFJEF (SEQ ID NO: 15) into three peptide sequences with 9 amino acids“ IEFROEIFJ (SEQ ID NO: 16), ”“ EFROEIFJE (SEQ ID NO: 17), ”And“ FROEIFJEF (SEQ ID NO: 18). “In one embodiment, prediction module 320 can identify candidate neoantigens that are mutated peptide sequences by comparing sequence data extracted from a patient's normal tissue cells with the sequence data extracted from the patient's tumor tissue cells to identify portions that contain one or more mutations.
[0435] [0435] Presentation module 320 applies one or more of the presentation models to the peptide sequences processed to estimate the probabilities of presentation of the peptide sequences. Specifically, the prediction module 320 can select one or more candidate neoantigen peptide sequences that are likely to be presented in tumor HLA molecules by applying the presentation models to the candidate neoantigens. In one implementation, presentation module 320 selects candidate neoantigen sequences that have an estimated likelihood of presentation above a predetermined threshold. In another implementation, the presentation model selects the sequences of neoantigens candidates for N that have the highest estimated presentation probabilities (where N is generally the maximum number of epitopes that can be delivered in a vaccine). A vaccine including the candidate neoantigens selected for a given patient can be injected into the patient to induce immune responses.
[0436] [0436] The validity of the various presentation models described above was tested on T test data that were subsets of training data 170 that were not used to train the presentation models or a separate data set from the training data 170 that have variables and data structures similar to those of training data 170.
[0437] [0437] A relevant metric indicative of the performance of a presentation model is: Positive Predictive Value (PPV) = P (Yjer = 1 | uer 2 0) = the one that indicates the ratio between the number of peptide instances that were correctly predicted for be presented in the associated HLA alleles and the number of peptide instances that were predicted to be presented in the HLA alleles. In one implementation, it was predicted that a p 'peptide in the T test data would be presented in one or more associated HLA alleles if the corresponding probability estimate u; is greater than or equal to a certain threshold value t. Another relevant metric indicative of the performance of presentation models is: Lier 101 = Lu>) Recall = P (ujer 2 t | yjer = 1) = OT. = D which indicates the ratio between the number of peptide instances that were correctly predicted to be presented in the associated HLA alleles and the number of peptide instances that were known to be presented in the HLA alleles. Another relevant metric indicative of the performance of the presentation models is the area under curve (AUC) of the receiver's operating characteristic (ROC). The ROC represents the callback against the false positive rate (FPR), which is provided by: Lier 101 = 0, u; 21) FPR = P (Ujer 2 t | Yjer = 0) = EI, = 0 X. A. Performance of the Presentation Model in Mass Spectrometry Data X.A.1. Example 1
[0438] [0438] FIG. 13A is a histogram of peptide lengths eluted from MHC class II alleles in human tumor cells and tumor infiltrating lymphocytes (TIL) using mass spectrometry. Specifically, mass spectrometry peptidomics was performed on HLA-DRB1 * 12: 01 homozygous alleles (“Dataset 1”) and HLA-DRB1 * 12: 01, HLA-DRB1 * 10: 01 (“multiple allele samples” (“ Data set 2 ”). The results show that peptide lengths eluted from MHC class II alleles range from 6-30 amino acids. At the frequency distribution shown in FIG. 13A is similar to that of peptide lengths eluted from class II MHC alleles using cutting-edge mass spectrometry techniques, as shown in FIG. 1C of reference 69.
[0439] [0439] FIG. 13B illustrates the dependence between MRNA quantification and the peptides presented per residue for data set 1 and data set 2. The results show that there is a strong dependence between MRNA expression and the presentation of peptides for MHC alleles class II.
[0440] [0440] Specifically, the horizontal axis in FIG. 13B indicates mRNA expression in terms of log transcripts per million (TPM). The vertical axis in FIG. 13B indicates the presentation of the peptide per residue as a multiple of that of the lowest compartment corresponding to the expression of MRNA between 10 <logioTPM <
[0441] [0441] The results indicate that the performance of the presentation model can be greatly improved by incorporating MRNA quantification measures, as these measures are strongly predictive of peptide presentation.
[0442] [0442] FIG. 13C compares performance results, for example, presentation templates trained and tested using Data Set 1 and Data Set 2. For each set of model characteristics of the example presentation templates, FIG. 13C represents a PPV value with 10% recovery when the characteristics in the model feature set are classified as interactive allele features and, alternatively, when features in the model feature set are classified as non-interactive allele feature variables. As seen in FIG. 13C, for each set of model features in the sample presentation models, a PPV value with 10% recall that was identified when the features in the set of model features were classified as interactive allele features is shown on the left and a PPV value on recall of 10% that was identified when the features in the model feature set were classified as non-interactive allele features are shown on the right side. Note that the peptide sequence characteristic has always been classified as an interactive allele characteristic for the purposes of FIG. 13C. The results showed that the presentation models reached a PPV value at 10% of recall, ranging from 14% to 29%, which are significantly (approximately 500 times) higher than PPV for a random forecast.
[0443] [0443] Peptide sequences of lengths 9-20 were considered for this experiment. The data were divided into training, validation and test sets. Peptide blocks of 50 residue blocks from data set 1 and data set 2 were assigned to training and test sets. Duplicate peptides in any part of the proteome were removed, ensuring that no peptide sequences appeared in the training and test set. The prevalence of presentation of peptides in the training and test set was increased by 50 times with the removal of non-presented peptides. This is because the Dataset | and Data Set 2 are from human tumor samples in which only a fraction of the cells are HLA class II alleles, resulting in peptide yields approximately 10 times less than in pure HLA class II allele samples, the which is still underestimated due to the imperfect sensitivity of mass spectrometry. The training set contained 1,064 peptides presented and 3,810,070 peptides not shown. The test set contained 314 and 807,400 peptides not shown.
[0444] [0444] The example model 1 was the sum of functions model in equation (22) using a network dependency function gn (-), the output function f (-) and the identity function r (-). The gn (*) network dependency function was structured as a multilayer perceptron (MLP) with 256 hidden nodes and activations per rectified linear unit (ReLU). In addition to the peptide sequence, the allele w interactive variables contained the C-terminal and N-terminal flanking sequence encoded by one-hot, a categorical variable indicating the index of the origin gene G = gene (p ") of the p'e peptide a variable indicating MRNA quantification measurement. Example 2 model was identical to example 1 model, except that the C-terminal and N-terminal flanking sequence was omitted from the allele interactive variables. Example 3 model was identical to the example model 1, except that the index of the source gene was omitted from the allele interactive variables The example model 4 was identical to the example model 1, except that the MRNA quantification measurement was omitted from the allele interactive variables .
[0445] [0445] The example model 5 was the sum of functions model in equation (20) with a network dependency function gn (), the output function f (-), the identity function f (-), the identity function r (-) and the dependency function gw (-) of equation (12). The gw (") dependency function also included a network model that takes MRNA quantification measures as input, structured as an MLP with 16 hidden nodes and ReLU activations, and a network model that takes the C flanking sequence as input , structured as an MLP with 32 hidden nodes and ReLU activations The gn (-) network dependency function was structured as a multilayer perceptron with 256 hidden nodes and rectified linear unit (ReLU) activations. 6 was identical to the example model 5, except that the network model for the C-terminal and N-terminal flanking sequence was omitted.The example model 7 was identical to the example model 5, except that the index of the gene for origin was omitted from the non-interactive allele variables The example model 8 was identical to the example model 5, except that the network model for measuring MRNA quantification was omitted.
[0446] [0446] The prevalence of peptides presented in the test set was approximately 1/2400 and therefore the PPV for a random prediction would also be approximately 1/2400 = 0.00042. As shown in FIG. 13C, the best performing presentation model achieved a PPV value of approximately 29%, approximately 500 times better than the PPV value of a random prediction.
[0447] [0447] FIG. 13D is a histogram that describes the amount of peptides sequenced using mass spectrometry for each sample out of a total of 39 samples comprising HLA class II molecules. In addition, for each sample of the plurality of samples, the histogram shown in FIG. 13D represents the amount of peptides sequenced using mass spectrometry at different thresholds of q value. Specifically, for each sample of the plurality of samples, FIG. 13D represents the amount of peptides sequenced using mass spectrometry with a q value less than 0.01, with a q value less than 0.05 and with a q value less than 0.2.
[0448] [0448] As noted above, each sample of the 39 samples in FIG. 13D comprised class II HLA molecules. More specifically, each sample of the 39 samples in FIG. 13D comprised HLA-DR molecules. The HLA-DR molecule is a type of HLA class II molecule. Even more specifically, each sample of the 39 samples in FIG. 13D comprised HLA-DRB1 molecules, HLA-DRB3 molecules, HLA-DRB4 molecules and / or HLA-DRB5 molecules. The HLA-DRB1 molecule, the HLA-DRB3 molecule, the HLA-DRB4 molecule and the HLA-DRB5 molecule are types of the HLA-DR molecule.
[0449] [0449] Although this particular experiment was carried out using samples that comprise HLA-DR molecules and, in particular, HLA-DRB1 molecules, HLA-DRB3 molecules, HLA-DRB4 molecules and HLA-DRB5 molecules, in alternative modalities, this experiment can be performed using samples comprising one or more of any type (s) of HLA class II molecules. For example, in alternative embodiments, identical experiments can be performed using samples comprising molecules of HLA-DP and / or HLA-DQ. This ability to model any type of MHC class II molecules using the same techniques, and still achieve reliable results, is well known to those skilled in the art. For example, Jensen, Kamilla Kjaergaard, et al. ** are an example of a recent scientific article that uses identical methods to model the binding affinity for HLA-DR molecules, as well as for HLA-DQ and HLA-DP molecules. Therefore, one skilled in the art would understand that the experiments and models described here can be used to model separately or simultaneously not only HLA-DR molecules, but any other class II MHC molecule, while still producing reliable results.
[0450] [0450] To sequence the peptides of each sample of the 39 total samples, mass spectrometry was performed for each sample. The resulting mass spectrum for the sample was then searched with Comet and tagged with Percolator to sequence the peptides. Then, the amount of peptides sequenced in the sample was identified for a plurality of different Percolator q value thresholds. Specifically, for the sample, the amount of sequenced peptides with a Percolator q value less than 0.01, with a Percolator q value less than 0.05 and with a Percolator q value less than 0.2 was determined.
[0451] [0451] For each of the 39 samples, the amount of peptides sequenced at each of the different thresholds of the Percolator q value is represented in FIG. 13D. For example, as seen in FIG. 13D, for the first sample, approximately 4000 peptides with a q value less than 0.2 were sequenced using mass spectrometry, approximately 2800 peptides with a q value less than 0.05 were sequenced using mass spectrometry and approximately 2300 peptides with a q value less than 0.01 was sequenced using mass spectrometry.
[0452] [0452] In general, FIG. 13D demonstrates the ability to use mass spectrometry to sequence a large amount of peptides from samples containing MHC class II molecules, with low q values. In other words, the data represented in FIG. 13D demonstrate the ability to reliably sequence peptides that can be presented by MHC class II molecules, using mass spectrometry.
[0453] [0453] FIG. 13E is a histogram that describes the number of samples in which a specific MHC class II molecule allele has been identified. More specifically, for the 39 total samples comprising class II HLA molecules, FIG. 13E represents the number of samples in which certain alleles of the MHC class II molecule have been identified.
[0454] [0454] As discussed above with reference to FIG. 13D, each sample of the 39 samples of FIG. 13D comprised HLA-DRB1 molecules, HLA-DRB3 molecules, HLA-DRB4 molecules and / or HLA-DRB5 molecules. Therefore, FIG. 13E represents the number of samples in which certain alleles for the HLA-DRB1, HLA-DRB3, HLA-DRB4 and HLA-DRB5 molecules have been identified. To identify the HLA alleles present in a sample, DR HLA class II typing is performed for the sample. Then, to identify the number of samples in which a specific HLA allele has been identified, the number of samples in which the HLA allele has been identified using HLA class II DR typing is simply added together. For example, as shown in FIG. 13E, 19 samples from the 39 total samples contained the HLA class II HLA-DRB4 * 01: 03 allele. In other words, 19 samples from the 39 total samples contained the HLA-DRB4 * 01: 03 allele for the HLA molecule. -DRB4. In general, FIG. 13E represents the ability to identify a wide range of class II HLA molecule alleles from the 39 samples comprising class II HLA molecules.
[0455] [0455] FIG. 13F is a histogram that describes the proportion of peptides presented by MHC class II molecules in the 39 total samples, for each peptide length in a range of peptide lengths. To determine the length of each peptide in each sample of the 39 total samples, each peptide was sequenced using mass spectrometry as discussed above with respect to FIG. 13D and then the number of residues in the sequenced peptide was simply quantified.
[0456] [0456] As noted above, MHC class II molecules typically have peptides with lengths between 9 to 20 amino acids. Therefore, FIG. 13F represents the proportion of peptides presented by MHC class II molecules in the 39 samples for each peptide length between 9-20 amino acids, inclusive. For example, as shown in FIG. 13F, approximately 22% of the peptides presented by MHC class II molecules in the 39 samples comprise a length of 14 amino acids.
[0457] [0457] Based on the data represented in FIG. 13F, the modal lengths of the peptides presented by the MHC class II molecules in the 39 samples were identified as being 14 and 15 amino acids in length. These modal lengths identified for the peptides presented by the MHC class II molecules in the 39 samples are consistent with previous reports of modal lengths of the peptides presented by the MHC class II molecules. In addition, as well as consistent with previous reports, the data in FIG. 13F indicate that more than 60% of the peptides presented by the MHC class II molecules of the 39 samples comprise different lengths of 14 and 15 amino acids. In other words, FIG. 13F indicates that, although the peptides presented by MHC class II molecules are most often 14 or 15 amino acids in length, a large proportion of the peptides presented by MHC class II molecules are not 14 or 15 amino acids in length. Therefore, it is a poor assumption to assume that peptides of all lengths are equally likely to be presented by MHC class II molecules or that only peptides that are 14 or 15 amino acids long are presented by MHC class TI molecules. As discussed in detail below with reference to FIG. 13J, these defective assumptions are currently used in many leading models to predict the presentation of peptides by MHC class II molecules and, therefore, the presentation probabilities predicted by these models are generally unreliable.
[0458] [0458] FIG. 13G is a line graph that represents the relationship between gene expression and the prevalence of presentation of the gene expression product by a class II MHC molecule, for genes present in the 39 samples. More specifically, FIG. 13G represents the relationship between gene expression and the proportion of residues resulting from gene expression that forms the N-terminal of a peptide presented by a class II MHC molecule. To quantify gene expression in each sample of the 39 total samples, RNA sequencing is performed on the RNA included in each sample. In FIG. 13G, gene expression is measured by RNA sequencing in transcription units per million (TPM). To identify the prevalence of the presentation of gene expression products for each sample of the 39 samples, the identification of DR HLA class TI peptidomic data for each sample was performed.
[0459] [0459] As shown in FIG. 13G, for the 39 samples, there is a strong correlation between the level of gene expression and the presentation of residues of the gene product expressed by a class II MHC molecule. Specifically, as shown in FIG. 13G, peptides resulting from the expression of less expressed genes are more than 100 times less likely to be presented by a MHC class 1 molecule than peptides resulting from the expression of the most expressed genes. In simpler terms, the most highly expressed gene products are most often presented by MHC class II molecules.
[0460] [0460] FIGS. 13H-J are line graphs that compare the performance of various presentation models in predicting the likelihood that peptides in a peptide test dataset will be presented by at least one of the class II MHC molecules present in the dataset of test. As shown in FIGS. 13H-J, the performance of a model in predicting the probability that a peptide will be presented by at least one of the class II MHC molecules present in the test data set is determined by identifying a ratio between a true positive rate and a false positive rate for each prediction made by the model. These relationships identified for a given model can be viewed as a ROC curve (characteristic of the receiving operator), in a line graph with a false positive rate for quantification of the x axis and a true positive rate for quantification of the y axis. An area under the curve (AUC) is used to quantify the model's performance. Specifically, a model with a higher AUC has a higher performance (that is, greater accuracy) than a model with a lower AUC. In FIGS. 13H-J, the blackened dashed line with a slope of 1 (i.e., a ratio between the true positive rate and the false positive rate of 1) represents the expected curve for random peptide presentation probabilities. The dashed line AUC is 0.5. The ROC curves and the AUC metric are discussed in detail in relation to the upper portion of Section X. above.
[0461] [0461] FIG. 13H is a line graph that compares the performance of five sample presentation models in predicting the likelihood that peptides in a peptide test data set will be presented by a class II MHC molecule, different data sets of interactive variables allele and non-interactive allele variables. In other words, FIG. 13H quantifies the relative importance of various interactive allele variables and non-interactive allele variables to predict the likelihood that a peptide will be presented by a class MHC molecule
[0462] [0462] The model architecture of each sample presentation model of the five sample presentation models used to generate the line graph ROC curves of FIG. 13H, comprised a set of five sigmoid sum models. Each sigmoid sum model in the set was configured to model the presentation of peptides for up to four unique HLA-DR alleles per sample. In addition, each sigmoid sum model in the set was configured to make predictions of the probability of presenting peptides based on the following interactive and non-interactive allele variables: peptide sequence, flanking sequence, RNA expression in TPM units, identifier gene, and sample identifier. The allele interactive component of each sigmoid sum model in the set was a hidden layer MLP with ReLu activations such as 256 hidden units.
[0463] [0463] Before using the sample models to predict the likelihood that the peptides in a peptide test data set will be presented by a class II MHC molecule, the sample models were trained and validated. To train, validate and finally test the sample models, the data described above for the 39 samples was divided into training, validation and test data sets.
[0464] [0464] To ensure that no peptides appear in more than one of the training, validation and test data sets, the following procedure was performed. First, all peptides from the 39 total samples that appeared in more than one location on the proteome were removed. Then, the peptides from the 39 total samples were partitioned into blocks of 10 adjacent peptides. Each peptide block of the 39 total samples was assigned exclusively to the training data set, the validation data set or the test data set. In this way, no peptides appeared in more than one data set from the training, validation and test data sets.
[0465] [0465] Of the 28,081 944 peptides in the 39 total samples, the training data set comprised 21,077 peptides presented by the MHC class II molecules of 38 of the 39 total samples. The 21,077 peptides included in the training data set were between the lengths of 9 and 20 amino acids, inclusive. The example models used to generate the ROC curves in FIG. 13H were trained on the training data set using the ADAM optimizer and early stop.
[0466] [0466] The validation data set consisted of 2,346 peptides presented by MHC class IT molecules from the same 38 samples used in the training data set. The validation set was used only for early stops.
[0467] [0467] The test data set comprised peptides presented by MHC class II molecules that were identified from a tumor sample using mass spectrometry. Specifically, the test data set comprised 203 peptides presented by MHC class II molecules - specifically moleculesHLA-DRB1 * 07: 01, HLA-DRB1 * 15: 01, HLA-DRB4 * 01: 03 - that were identified from of the tumor sample. The peptides included in the test data set were kept outside the training data set described above.
[0468] [0468] As noted above, FIG. 13H quantifies the relative importance of several interactive allele variables and non-interactive allele variables to predict the likelihood that a peptide will be presented by a class II MHC molecule. As also noted above, the example models used to generate the ROC curves of the line graph of FIG. 13H were configured to make predictions of the probability of presenting peptides based on the following interactive and non-interactive allele variables: peptide sequence, flanking sequence, RNA expression in TPM units, gene identifier and sample identifier. To quantify the relative importance of four of these five variables (peptide sequence, flanking sequence, RNA expression and gene identifier) to predict the likelihood that a peptide will be presented by a class II MHC molecule, each sample model of the five The sample models described above were tested using test data set data, with a different combination of the four variables. Specifically, for each peptide in the test data set, an example 1 model generated predictions of the likelihood of presenting peptides based on a peptide sequence, a flanking sequence, a gene identifier and a sample identifier, but not on expression of RNA. Likewise, for each peptide in the test data set, an example model 2 generated predictions of the likelihood of presenting peptides based on a peptide sequence, RNA expression, a gene identifier and a sample identifier, but not in a flanking sequence. Likewise, for each peptide in the test data set, an example 3 model generated predictions of the likelihood of peptide presentation based on a flanking sequence, RNA expression, a gene identifier and a sample identifier, but not in a peptide sequence. Likewise, for each peptide in the test data set, an example model 4 generated predictions of the likelihood of presenting peptides based on a flanking sequence, RNA expression, a peptide sequence and a sample identifier, but not in a gene identifier. Finally, for each peptide in the test data set, a sample model 5 generated predictions of the likelihood of presenting peptides based on all five flanking sequence variables, RNA expression, peptide sequence, sample identifier and sample identifier. gene.
[0469] [0469] The performance of each of these five example models is represented in the line graph of FIG. 13H. Specifically, each of the five example models is associated with a ROC curve that represents a ratio of a true positive rate to a false positive rate for each prediction made by the model. For example, FIG. 13H represents a curve for the example 1 model that generated predictions of the likelihood of presenting peptides based on a peptide sequence, a flanking sequence, a gene identifier and a sample identifier, but not on RNA expression. FIG. 13H represents a curve for the example model 2 that generated predictions of the likelihood of presenting peptides based on a peptide sequence, RNA expression, a gene identifier and a sample identifier, but not a flanking sequence. FIG. 13H also represents a curve for the example model 3 that generated predictions of the likelihood of peptide presentation based on a flanking sequence, RNA expression, a gene identifier and a sample identifier, but not a peptide sequence. FIG. 13H also represents a curve for example model 4 that generated predictions of the likelihood of peptide presentation based on a flanking sequence, RNA expression, a peptide sequence and a sample identifier, but not a gene identifier. And finally FIG. 13H represents a curve for the example model 5 that generated predictions of peptide presentation probability based on all five variables of flanking sequence, RNA expression, peptide sequence, sample identifier and gene identifier.
[0470] [0470] As noted above, the performance of a model in predicting the likelihood that a peptide will be presented by a class II MHC molecule is quantified by identifying an AUC for a ROC curve that represents a ratio between a positive rate true and a false positive rate for each prediction made by the model. A model with a higher AUC has a higher performance (that is, greater accuracy) than a model with a lower AUC. As shown in FIG. 13H, the curve for example model 5 that generated predictions for the likelihood of peptide presentation based on all five variables of flanking sequence, RNA expression, peptide sequence, sample identifier and gene identifier, achieved the most AUC high of 0.98. Therefore, the sample model that used all five variables to generate peptide presentation predictions achieved the best performance. The curve for example model 2 that generated predictions of the likelihood of peptide presentation based on a peptide sequence, RNA expression, a gene identifier and a sample identifier, but not a flanking sequence, reached the second AUC highest of 0.97. Therefore, the flanking sequence can be identified as the least important variable to predict the likelihood that a peptide will be presented by a class II MHC molecule. The curve for example model 4 generated predictions of the likelihood of peptide presentation based on a flanking sequence, RNA expression, a peptide sequence and a sample identifier, but not a gene identifier, achieved the third most AUC high of 0.96. Therefore, the gene identifier can be identified as the second least important variable to predict the likelihood that a peptide will be presented by a class II MHC molecule. The curve for example model 3 that generated predictions of the likelihood of peptide presentation based on a flanking sequence, RNA expression, a gene identifier and a sample identifier, but not a peptide sequence, achieved the most AUC low of 0.88. Therefore, the peptide sequence can be identified as the most important variable to predict the likelihood that a peptide will be presented by a class II MHC molecule. The curve for example model 1 that generated predictions of the likelihood of peptide presentation based on a peptide sequence, a flanking sequence, a gene identifier and a sample identifier, but not on RNA expression, reached the second Lowest AUC of 0.95. Therefore, RNA expression can be identified as the second most important variable to predict the likelihood that a peptide will be presented by a class II MHC molecule.
[0471] [0471] FIG. 131 is a line graph that compares the performance of four different presentation models in predicting the likelihood that peptides in a peptide test data set will be presented by a class II MHC molecule.
[0472] [0472] The first model tested in FIG. 131 is referred to here as a “complete non-interactive model”. The complete non-interactive model is a modality of the presentation models described above in which non-interactive allele variables w and interactive variables of allele xr are inserted in separate dependency functions, such as, for example, a neural network, and the outputs of these functions of separate dependencies are added. Specifically, the complete non-interactive model is a modality of the presentation models described above, in which the non-interactive variables of allele w "are inserted in a dependency function gy, interactive variables of allele x"
[0473] [0473] The second model tested in FIG. 131 is referred to here as a “complete interactive model”. The full interactive model is a modality of the presentation models described above, in which the non-interactive variables of the allele w are directly concatenated to the interaction variables of the allele xn * before being inserted into a dependency function, such as, for example, a neural network. Therefore, in some modalities, the complete interactive model determines the probability of presenting the peptide using equation 9, as shown above. In addition, modalities of the complete interactive model in which the w * allele interactive variables are concatenated with xn * allele interactive variables before the variables are inserted into a dependency function are discussed in detail above with respect to the lower portion of Section VIIL .B.2., The lower portion of Section VIII.C 2. and the lower portion of Section VIIL.C.5.
[0474] [0474] The third model tested in FIG. 131 is referred to herein as a “CNN model”. The CNN model comprises a convolutional neural network and is similar to the complete non-interactive model described above. However, the layers of the convolutional neural network of the CNN model differ from the layers of the neural network of the complete non-interactive model. Specifically, the entry layer of the CNN model convolutional neural network accepts a 20 m peptide strip and subsequently incorporates the 20 m peptide strip as a tensor (n, 20, 21). The next layers of the CNN model convolutional neural network comprise a size 5 1-D convolutional core layer with a pass of 1, a global maximum pool layer, an abandonment layer with p = 0.2 and finally a dense layer 34 knots with a ReLu activation.
[0475] [0475] The fourth and last model tested in FIG. 131 is referred to herein as an “LSTM model”. The LSTM model comprises a short-term long-memory neural network. The input layer of the LSTM model's short-term long-memory neural network accepts a 20 mers peptide strip and subsequently incorporates the 20 mers peptide strip as a tensor (n, 20, 21). The next layers of the LSTM model short-term neural network comprise a 128-node short-term long-memory layer, an abandonment layer with p = 0.2 and, finally, a 34-node dense layer with a ReLu activation.
[0476] [0476] Before using each of the four models in FIG. 131 to predict the probability that the peptides in the peptide test dataset will be presented by a class II MHC molecule, the models were trained using the 38 sample training dataset described above and validated using the validation data described above. After this training and validation of the models, each of the four models was tested using the 39th extended sample test data set described above. Specifically, for each of the four models, each peptide from the test data set was inserted into the model and the model subsequently produces a presentation probability for the peptide.
[0477] [0477] The performance of each of the four models is represented in the line graph of FIG. 131. Specifically, each of the four example models is associated with a ROC curve that represents a ratio of a true positive rate to a false positive rate for each prediction made by the model. For example, FIG. 131 represents a ROC curve for the CNN model, a ROC curve for the complete interactive model, a ROC curve for the LSTM model and a ROC curve for the complete non-interactive model.
[0478] [0478] As noted above, the performance of a model in predicting the likelihood that a peptide will be presented by a class 1 MHC molecule is quantified by identifying an AUC for a ROC curve that represents a ratio between a positive rate true and a false positive rate for each prediction made by the model. A model with a higher AUC has a higher performance (that is, greater accuracy) than a model with a lower AUC. As shown in FIG.
[0479] [0479] FIG. 13J is a line graph that compares the performance of two examples of the best prior art models, with two different criteria, and two example presentation models, with two different sets of interactive allele variables and non-interactive allele variables, for predict the likelihood that peptides in a peptide test data set will be presented by a class II MHC molecule. Specifically, FIG. 13J is a line graph that compares the performance of an example of the best prior art model that uses the minimum predicted binding affinity of NetMHCII 2.3 as a criterion for generating predictions (example 1 model), an example of the best prior art model which uses the predicted minimum binding rating of NetMHCII 2.3 as a criterion for generating predictions (example model 2), an example presentation model that generates predictions of the probability of presenting peptides based on the type of MHC molecule and class II and sequence peptide (example model 4) and a sample presentation model that generates predictions of the likelihood of presenting that peptides based on the type of MHC class II molecule, peptide sequence, RNA expression, gene identifier and flanking sequence (example template 3).
[0480] [0480] The best prior art model used as example model 1 and example model 2 in FIG. 13J is the NetMHCII 2.3 model. The NetMHCII model
[0481] [0481] As noted above, the NetMHCII 2.3 model was tested according to two different criteria. Specifically, the sample model of model 1 generated predictions of probability of presenting peptides according to the minimum predicted binding affinity of NetMHCII 2.3 and example model 2 generated predictions of probability of presenting peptides according to the minimum classification of predicted NetMHCII 2.3 connection.
[0482] [0482] The presentation model used as example model 3 and example model 4 is a modality of the presentation model disclosed here that is trained using data obtained by mass spectrometry. As noted above, the presentation model generated predictions of the likelihood of peptide presentation based on two different sets of interactive allele variables and non-interactive allele variables. Specifically, example model 4 generated predictions for the likelihood of presenting peptides based on the type of molecule and the MHC class II peptide sequence (the same variable used by the NetMHCII 2.3 model) and model 3 generated generated probability predictions. of presentation of peptides based on the type of MHC class II molecule, peptide sequence, RNA expression, gene identifier and flanking sequence.
[0483] [0483] Before using the example models in FIG. 13J to predict the probability that the peptides in the peptide test data set will be presented by a class II MHC molecule, the models have been trained and validated. The NetMHCII 2.3 model (example model 1 and example model 2) has been trained and validated using its own training and validation data sets based on HLA peptide binding affinity assays deposited in the immune epitope database (IEDB, www. iedb. org). The training data set used to train the NetMHCII 2.3 model is known to comprise almost exclusively 15 mer peptides. On the other hand, example models 3 and 4 were trained using the training data set described above in relation to FIG. 13H and validated and using the validation data set described above in relation to FIG. 13H.
[0484] [0484] After training and validating the models, each model was tested using a set of test data. As noted above, the model
[0485] [0485] To test the sample models using the test data set, for each of the sample models, for each peptide out of the 933 peptides in the test data set, the model generated a prediction of the probability of presentation of the peptide. Specifically, for each peptide in the test data set, example model 1 generated a presentation score for the peptide by MHC class II molecules using types of MHC class II molecules and peptide sequence, classifying the peptide by affinity of minimum binding predicted by Net MHCII 2.3 in the four HLA class II DR alleles in the test data set. Likewise, for each peptide in the test data set, the model in example 2 generated a presentation score for the peptide by MHC class II molecules using types of MHC class II molecules and peptide sequence, classifying the peptide by the minimum binding rating predicted by NetMHCII 2.3 (ie, normalized quantile binding affinity) in the four HLA class II DR alleles in the test data set. For each peptide in the test data set, the model in example 4 generated a probability of presentation for the peptide by MHC class II molecules based on the type of molecule and the MHC class II peptide sequence. Likewise, for each peptide in the test data set, example model 3 generated a probability of presentation for the peptide by MHC class II molecules based on the types of MHC class II molecules, peptide sequence, expression of RNA, gene identifier and flanking sequence.
[0486] [0486] The performance of each of the four example models is represented in the line graph of FIG. 13J. Specifically, each of the four example models is associated with a ROC curve that represents a ratio of a true positive rate to a false positive rate for each prediction made by the model. For example, FIG. 13J represents a ROC curve for example model 1 that used the minimum predicted binding affinity of NetMHCII 2.3 to generate predictions, a ROC curve for example model 2 that used the minimum predicted binding classification of NetMHCII 2.3 to generate predictions, a ROC curve for the example model 4 that generated probabilities of presentation of peptides based on the type of MHC class II molecule and peptide sequence and a ROC curve for the model of example 3 that generated probabilities of presentation of peptides based on the class II MHC molecule type, peptide sequence, RNA expression, gene identifier and flanking sequence.
[0487] [0487] As noted above, the performance of a model in predicting the likelihood that a peptide will be presented by a class 1 MHC molecule is quantified by identifying an AUC for a ROC curve that represents a ratio between a positive rate true and a false positive rate for each prediction made by the model. A model with a higher AUC has a higher performance (that is, greater accuracy) than a model with a lower AUC. As shown in FIG. 13J, the curve of the model in example 3 that generated probabilities of peptide presentation based on the type of MHC class II molecule, peptide sequence, RNA expression, gene identifier and flanking sequence, reached the highest AUC of 0 , 95. Therefore, the model in Example 3 that generated probabilities of presenting peptides based on the type of MHC class II molecule, peptide sequence, RNA expression, gene identifier and flanking sequence achieved the best performance. The curve for the model in example 4 that generated probabilities of presenting peptides based on the type of MHC class II molecule and the peptide sequence reached the second highest AUC of 0.91. Therefore, the model in example 4, which generated probabilities of presenting peptides based on the type of MHC class II molecule and the peptide sequence, achieved the second best performance. The model curve of example 1 that used the minimum predicted binding affinity of NetMHCII 2.3 to generate predictions reached the lowest AUC of 0.75.
[0488] [0488] As shown in FIG. 135, the performance discrepancy between the sample models | and 2 and the example models 3 and 4 is large. Specifically, the performance of the NetMHCII 2.3 model (which uses a minimum expected link affinity criterion for NetMHCII 2.3 or minimum expected connection rating for NetMHCII 2.3) is almost 25% less than the performance of the presentation model disclosed here (which generates probabilities presentation of peptides based on the type of MHC class II molecule and peptide sequence, or the type of MHC class II molecule, peptide sequence, RNA expression, gene identifier and flanking sequence). Therefore, FIG. 13J demonstrates that the presentation models disclosed here are capable of obtaining significantly more accurate presentation predictions than the current best-in-class model, the NetMHCII 2.3 model.
[0489] [0489] Furthermore, as discussed above, the NetMHCII 2.3 model is trained on a training data set comprising almost exclusively 15 mers peptides. As a result, the NetMHCII 2.3 model is not trained to learn which lengths of peptides are most likely to be presented by MHC class II molecules. Therefore, the NetMHCII 2.3 model does not weigh its predictions of probability of presentation of peptides by MHC class II molecules according to the length of the peptide. In other words, the NetMHCII 2.3 model does not modify its predictions for the likelihood of peptide presentation by MHC class II molecules for peptides that have modal peptide lengths outside of 15 amino acids. As a result, the NetMHCII 2.3 model overapproves the probability of presenting peptides with lengths greater than or less than 15 amino acids.
[0490] [0490] On the other hand, the presentation models disclosed here are trained using peptide data obtained by mass spectrometry and, therefore, can be trained on training data sets that comprise peptides of different lengths. As a result, the presentation models disclosed here are able to learn which lengths of peptides are most likely to be presented by MHC class II molecules. Therefore, the presentation models disclosed herein can weight predictions of probability of presentation of peptides by MHC class II molecules according to the length of the peptide. In other words, the presentation models disclosed here are capable of modifying their predictions of probability of presentation of peptides by MHC class II molecules to peptides that have lengths outside the modal length of 15 amino acids. As a result, the presentation models disclosed here are able to obtain significantly more accurate presentation predictions for peptides of lengths greater or less than 15 amino acids, than the current model of the best technique in the category, the NetMHCII 2 model. an advantage of using the presentation models disclosed here to predict the likelihood of peptide presentation by MHC class II molecules. X.B. Example of Determined Parameters for the MHC Allele
[0491] [0491] The following shows a set of parameters determined for a variation of the multiple allele presentation model (equation (16)), generating implicit probabilities of presentation by allele for MHC class II alleles HLA-DRB1 * 12: 01 and HLA-DRB1 * 10: 01: u = expit (relu (XW! + B!) - W + B ), Where relu (-) is the function of the rectified linear unit (RELU), W! ', B! , W , eb are the set of parameters 6 determined for the model. The allele X interactive variables are contained in a 1 x 399 matrix) consisting of 1 line of peptide sequences encoded by one-hot and filled in the middle by input peptide. The dimensions of W 'are (399 x 256), the dimensions of b! (1 x 256), the dimensions of W are (256 x 2), and b are (1 x 2). The first output column indicates the implied probability per presentation allele for the peptide sequence for the HLA-DRB1 * 12: 01 allele, and the second output column indicates the implant by allele for the peptide sequence for the HLA-DRB1 * 10 allele: 01. For demonstration purposes, the values for b !, b , W !, and W are listed below.
[0492] [0492] b ":
[0493] [0493] 2: -3. 017294168472290039 € +00 -3. 030839920043945312e + 00
[0494] [0494] W !:
[0495] [0495] 4, 2335607111454010010-02 8. 138803392648696899c-02 A. 656502604484558105e-02 7. 230042666196823120c-02 8. 9775562286376953120-02 5. 444654822349548340e-02 -8. 769565075635910034e-02 4. 059319198131561279c-02 -
[0496] [0496] 3. 085911273956298828e-02 9, 068728983402252197e-02 -2. 633090876042842865e-02 3. 693536296486854553e-02 3. 310681506991386414e-02 2. 388063445687294006 € -02 5. 356629937887191772e-02 9. 871018119156360626e-03 -
[0497] [0497] 1. 427586823701858521e-01 -7. 231150567531585693e-02 4. 904205910861492157e-03 2. 783779799938201904e-01 5. 328441038727760315e-02 -
[0498] [0498] 1. 804834045469760895ec-02 -6. 538107246160507202e-02 7. 933633774518966675e-02 -4. 392367973923683167e-02 -3. 913663700222969055e-02
[0499] [0499] -l. 115055941045284271e-02 -8. 712556958198547363e-02 -2. 373490296304225922e-02 6. 936468929052352905e-02 1. 036548404954373837e-03 1. 435988303273916245e-02 -2. 718369476497173309ec-02 -4. 1708622127771377560-02 -
[0500] [0500] 4. 365295637398958206e0-03 4. 656402021646499634e-02 7. 395512610673904419e-02 1. 371310502290725708e-01 4. 901177063584327698e-02 -
[0501] [0501] -8. 564729243516921997e0-03 -9. 529667347669601440e0-02 5. 183426290750503540 € -02 -7. 491327077150344849e-02 -1. 094889193773269653e-01
[0502] [0502] 1. 305916160345077515e-01 -1. 078678742051124573e-01 8. 201939985156059265e-03 1. 308720260858535767e-01 6. 592839211225509644e-02 2. 427846193313598633e-02 8. 769390337169170380e-03 -3. 1494028866291046] 14e-02 1. 112432777881622314e-01 8. 340127766132354736 € -02 -2. 892854856327176094e-03 2. 407504916191101074e-01 2. 471237070858478546 € -02 1. 445051878690719604e-01 7. 777267694473266602e-02 4. 292483255267143250e-02 1. 565658897161483765107686 642441496253013611e-02 1. 218502223491668701e-01 3. 208779543638229370e-02 -4. 194468259811401367e-02 6. 188327819108963013e-02 -8. 4593981 50444030762e-02 8. 905439823865890503e-02 -
[0503] [0503] 2. 204831456765532494e-03 -7. 4844397604465484620-02 -6. 789403408765792847e-02 -6. 098340451717376709e-02 3. 783712908625602722e-02 -
[0504] [0504] 8. 810643106698989868e-02 -4. 699551686644554138e-02 4. 421688243746757507e-02 1. 3503634929656982420-01 8. 4028482437133789060-02 1. 045651081949472427e-02 3. 054236061871051788e-02 9. 082502871751785278e-02 - 02
[0505] [0505] 1. 240125745534896851e-01 -8. 903886377811431885e-02 5. 741309374570846558e-02 1. 175503954291343689e-01 2. 617571689188480377e-02 4. 322101920843124390e-02 -4. 287067800760269165e-02 1. 094607114791870117e-01 1. 429035235196352005e-02 6. 520090159028768539e-03 -8. 093666285276412964e-03 2. 186897583305835724e-02 6. 340958923101425171e-02 4. 423595219850540161e-02 1. 3634607 19585418701e-01 4. 699309542775154114e-02 1.33697569370267797263703 397313460707664490e-02 5. 725258588790893555e-02 -1. 897132024168968201c-02 -
[0506] [0506] 7. 315602898597717285e-02 -2. 503681369125843048e-02 1. 930215768516063690e-02 1. 159798204898834229e-01 9. 557762183248996735e-03 1. 958409138023853302e-02 -1.527437102049589157e-02 6. 881920620799064 266111582517623901e-01 9. 435189887881278992e-03 6. 954447180032730103e-02 3. 640574216842651367e-02 2. 5696953758597373960-02 8. 274929225444793701e-02 -
[0507] [0507] -7. 551768422126770020e-02 -4. 545046761631965637e-02 4. 243563860654830933e-02 -8. 221449702978134155e-02 -1. 778533123433589935e-02
[0508] [0508] 1. 106902137398719788e-01 -2. 3764410987496376040-02 5. 217778310179710388e-02 2. 117201536893844604e-01 3. 342163190245628357e-02 -
[0509] [0509] 1. 269571632146835327e-01 5. 814104899764060974e-02 -5. 080512724816799164e-03 6. 960341334342956543e-02 8. 499863743782043457e-02 6.
[0510] [0510] 4. 796088859438896179e-02 5. 510679632425308228e-02 -5. 555555596947669983 € -02 1. 337369233369827271e-01 -2. 225189283490180969e-02 -
[0511] [0511] 5. T785515997558832169c-03 6. 700182706117630005e-02 -5. 314164236187934875e-02 -7. 213075645267963409e-03 5. 538868904113769531e-02 -
[0512] [0512] 2. 655223011970520020e-02 1. 038572285324335098e-02 -L. 140253692865371704e-01 3. 937279805541038513e-02 1. 692912355065345764e-02 -
[0513] [0513] 1. 360329054296016693e-02 -7. 211557030677795410c0-02 4. 195940215140581131c-03 2. 151363901793956757e-02 -6. 054398790001869202e-02 -
[0514] [0514] 1. 002630889415740967e-01 -6. 658916175365447998e-02 5. 456307157874107361ec-02 4. 031676650047302246 € -01 -5. 502983182668685913ec-02 5. 856209620833396912e-02 -3. 336066752672195435e-02 1. 108936890959739685e-01 -
[0515] [0515] 4. 195605218410491943e-02 -8. 119218796491622925e-02 -9, 011746197938919067e-02 6. 637725234031677246e-02 -3. 276840224862098694e-02 -
[0516] [0516] -7. 209850847721099854c-02 6. 884892284870147705e-02 7. 339402288198471069e-02 -6. 556650251150131226 € -02 -2. 0287502557039260860-02
[0517] [0517] -l. 6257485374808311460-02 -1. 075620874762535095e-01 3. 860892029479146004e-03 -6. 137971952557563782e-02 -4. 463825374841690063e-02 -
[0518] [0518] 6. 8599581718444824220-02 3. 646993776783347130ec-03 -6. 508616730570793152e-03 3. 349136114120483398e-01 1. 153858229517936707e-01 -
[0519] [0519] 5. 873806402087211609e-02 -5. 362452939152717590e-03 8. 574135601520538330e-02 1. 017753183841705322e-01 6. 802455987781286240c-03 -
[0520] [0520] 1. 261337548494338989c-01 -6. 616394966840744019e-02 -6. 865292042493820190e-02 3. 023189306259155273e-02 3. 431053832173347473ec-02 -
[0521] [0521] 7. 312644273042678833e-02 5. 2466459572315216060-02 -2. 8357343748211860660-02 3. 110809624195098877e-01 -5. 444446951 150894165e-02 -
[0522] [0522] -l. 3688452541828155520-02 -1. 556199323385953903e-02 |. 083098649978637695e-01 -6. 177869066596031189e-02 -1. 392558962106704712e-01 9, 404736757278442383e-02 3. 392037749290466309e-02 -6. 108184531331062317e-02
[0523] [0523] 2. 6677191257476806640-02 5. 6987721472978591920-02 6. 529657635837793350e-03 1. S48548638820648193e-01 2. 727760374546051025e-02 6. 441962718963623047e-02 2. 107882173753
[0524] [0524] 5. 449816212058067322e-02 1. 892925985157489777e-02 -2. 0466551 18465423584e-02 3. 059766627848148346 € -02 9. 7631394863 12866211e-02 1. 744062453508377075e-02 4. 576938599348068237e-02 3. 169437125325202942e-02 2. 67004370689397238388393138338386343 01 5.127409100532531738e-02 -8. 102724701166152954e-02 1. 087911203503608704e-01 -8. 453795313835144043e-02 1. 034669694490730762e0-03 -
[0525] [0525] 4. 908315092325210571e-02 4. 687566682696342468e-02 -1. 1137639731 16874695e-01 1. 555493921041488647e-01 -4. 038666188716888428e-02 -
[0526] [0526] 8. 540467917919158936c-02 -1. 027349606156349] 82e0-01 3. 604605793952941895e-02 2. 090096920728683472e-01 -4. 070349410176277161e-02 1. 235357392579317093e-02 -7. 529452443122863770e-02 9. 358168393373489380e-02 4. 3679177761077880860-03 3. 604184836149215698e-03 2. 957904897630214691e-02 3. 3503554761409759520-02 -3. 603352978825569153e-02 1. 304543316364288330e-01 6. 840901821851730347e-02 -5. 167580768465995789e-02 6. 964340806007385254e-02 1. 458993554115295410e-02 5. 786298587918281555e-02 -1. 511506922543048859e-02 4. 911896213889122009e-02 8. 389481902122497559e-02 3. 632528707385063171e-02 -
[0527] [0527] 1. 069942265748977661e-01 2. 931387722492218018e-02 -1. 065118703991174698e-02 1. 835492998361587524e-01 9. 332873672246932983e-02 -
[0528] [0528] -8. 558973670005798340c-02 -4. 310989286750555038e-03 3. 261769562959671021e-02 -9. 008675813674926758e-02 6. 356125324964523315e-02 2. 347143180668354034e-02 -9. 064750373363494873e-02 -8. O84800839424133301e-02
[0529] [0529] -7. 566627115011215210e-02 4. 4703036546707153320-02 -1. 519178040325641632e-02 -1. 009783893823623657e-01 9. 287630021572113037e-02 -
[0530] [0530] -7. 561279088258743286e0-02 8. 163869380950927734e-02 -8. 082308620214462280e-02 -1. 175542846322059631e-01 4. 294025897979736328e-02 4. 769440740346908569e-02 -1. 070467904210090637e-01 1. 082515064626932144e-02 7. 111620903015136719e-02 -5. 354597046971321106 € -02 5. S86164444684982300e-02 -
[0531] [0531] 1. 413376629352569580e-01 -6. 831099838018417358e-02 |. 211901754140853882e-02 2. 5S48536360263824463e-01 -5. 107542872428894043e-02 -
[0532] [0532] 5. 914099141955375671e-02 -3. 887363895773887634e-02 2. 924362197518348694e-02 9. 470376186072826385e-03 -9. 961505234241485596 € -02 6. 057203747332096100e-03 -4. 279591515660285950e-02 -1. 231343578547239304c-02
[0533] [0533] -4. 901335760951042175e-02 6. 970176100730895996e-02 3. 363081021234393120ec-03 1. 082899272441864014e-01 -1. 163023174740374088e-03 -
[0534] [0534] -1. 379721891134977341e-02 7. 332481443881988525e-02 3. 163217380642890930e-02 8. 646478503942489624e-02 1. 195115689188241959e-02 5.
[0535] [0535] 1. 888449788093566895e-01 -8. 8155388832092285160-02 -1. 058763936161994934e-01 3. 079998195171356201e-01 2. 353592403233051300e-02 -
[0536] [0536] 3. 7393655627965927120-02 -9, 651449322700500488e-02 5. 334780365228652954e-02 6. 040098 1456041336060-02 -3. 789112903177738190e-03 2. 785195223987102509e-02 8. 896316401660442352e-03 4. 46775481 1048507690e-02 1. 845969259738922119e-02 -6. 566123664379119873e-02 6. 3884302973747253420-02 1. 122175380587577820e-01 5. 359574034810066223e-02 1. 043849065899848938e-01 -
[0537] [0537] -2. 542039193212985992ec-02 -8. 218019455671310425e-02 8. 785786107182502747e-03 -8. 189122378826141357e-02 -8. 719720691442489624e-02 -
[0538] [0538] -1. 152848731726408005e-02 -1. 4743625652045011520-03 -3. 714315220713615417e-02 -4. 2737916111946105960c-02 -3. 484169393777847290ec-02
[0539] [0539] 9. 706448763608932495e-02 -4. 7722723335027694700-02 7. S8414110168814659] 1e-03 2. 439693659543991089e-01 1. 342512369155883789e-01 -
[0540] [0540] -4. 175704345107078552e-02 -9. 995118528604507446 € -02 -2. 000405080616474152e-02 -3. 313992172479629517e-02 -3. 649541642516851425e-03
[0541] [0541] -4. 9412198364734649660-02 4. 080281779170036316c-02 -8. 132989704608917236 € -02 8. 651164174079895020e-02 -5. 472549051046371460e-02 -
[0542] [0542] 8. 383004367351531982e-02 -7. 621740549802780151e-02 4. 762198368553072214 € -04 5. 896569043397903442e-02 -3. 473737090826034546e € -02 -
[0543] [0543] 5. 209536477923393250ec-02 -9, 628586471080780029e-02 7. 630670815706253052e-02 1. 800257340073585510e-02 -6. 500766426324844360e-023. 392039611935615540e-02 -8. 376629650592803955e-02 -5. 44235855340957641 60-02 -
[0544] [0544] 1. 248651146888732910ec-01 -6. 352679338306188583e-04 4. 137500002980232239e-02 1. 453564167022705078e-01 5. 8143369853496551510-02 -
[0545] [0545] -3. 361240774393081665e-02 -1. 065570861101150513e-01 -7. 006123661994934082e-02 2. 339961938560009003e-02 -1. 057655885815620422e-01 8. 364885300397872925e-02 -1. 124694272875785828e-01 -5. 360668525099754333e-02 -
[0546] [0546] 5. 5976796895265579220-02 -7. 2964534163475036620-02 -7. 847391068935394287e-02 4. 862646386027336121e-02 3. 875120356678962708e-02 3. 2539684325456619260-03 2. 520192228257656097e-02 8 &. 436141908168792725e-02 -
[0547] [0547] 6. 224662065505981445e-02 -3. 9436656981706619260-02 -3. 944940492510795593e-02 1. 657070815563201904e-01 4. 565249010920524597e-02 -
[0548] [0548] 1. 030491292476654053e-01 4. 819731414318084717e-02 -9. 275247156620025635e-02 6. 494653970003128052e-02 -5. 229672789573669434c-02 -
[0549] [0549] -2. 018752880394458771e-02 -3. 872476285323500633e-03 -5. 067412555217742920 € -02 -4. 951309412717819214e-02 5. 858522281050682068e-02 -
[0550] [0550] 4. 614380374550819397e-02 5. 7827718555927276610-02 -3. 473401069641113281ec-02 1. 040196642279624939e-01 4. 412507265806198120e-02 -
[0551] [0551] 8. 836930617690086365e-03 -8. 266210556030273438e-02 -1. 970501244068145752e-02 -3. 117918968200683594e-02 8. 020016551017761230e-02 -
[0552] [0552] 9. 527189284563064575e-02 -5. 150014907121658325e-02 1. 491536572575569153e-02 1. 022936478257179260e-01 7. 200413942337036133e-02 -
[0553] [0553] 6. 134003400802612305e-03 2. 957489714026451111e-02 1. 042789965867996216e € -02 5. 949835106730461121 € -02 -8. 1183187663555145260-023.
[0554] [0554] 7. 299335300922393799ec-02 -4. 204149171710014343e-02 -1. 085489168763160706 € -01 1. 746872998774051666e-02 -2. 373399026691913605e-023. 882227838039398193e-02 1. 5797540545463562010-02 -1. 809378899633884430e-02 4. 776366800069808960e-02 -2. 524876035749912262e-02 -6. 227674335241317749e-02 -
[0555] [0555] 9. 015270322561264038e-02 5. 519023910164833069c-02 -6. 583470851182937622e-02 -4. 965527728199958801e-02 -1. 023016944527626038e-01
[0556] [0556] 1. 2398601323366165160-01 5. 0322249531745910640-02 -2. 721861191093921661e-02 3. 240483999252319336e-01 1. 870138570666313171e-02 -
[0557] [0557] -9, 585490077733993530ec-02 -2. 940851822495460510ec-02 -4. 217590030748397112e-04 1. 78667961 1541330814e-03 -5. 478074401617050171e-023. 508562222123146057e-02 1. 961958408355712891e-02 7. 386039197444915771e-02 -
[0558] [0558] -7. 142431288957595825e-02 5. 323882214725017548e-04 -7. 817948609590530396e-02 -9. 212539531290531158e-03 -2. 406965941 190719604e-02 -
[0559] [0559] -4 ,. 726774524897336960e-03 6. 097768247127532959e-02 -8. 490203320980072021e-02 7. 521948963403701782e-02 -4. 301571473479270935e-02 4. 527747631072998047e-02 -3. 590170294046401978e-02 -6. 1356253921985626220-02
[0560] [0560] 9. 109907597303390503e-02 -7. 599887996912002563e-02 -5. 794495344161987305e-02 2. 422440201044082642e-01 -6. 013576872646808624e-03 -
[0561] [0561] 7. 043635100126266479c-02 -3. 085738047957420349ec-02 4. 995672404766082764e-02 1. 441784054040908813e-01 7. 644854485988616943ec-03 -
[0562] [0562] 6. 151535734534263611e6-02 -6. 8390473723411560060-02 -1. 497240271419286728e-02 2. 290117926895618439e-02 -1. 5S04544075578451157e-02 -
[0563] [0563] -4, 529131576418876648e-03 -3. 166146576404571533e-02 -7. 925513386726379395e-02 2. 971722185611724854e-02 6. 576989591121673584c-02 -
[0564] [0564] 9. 765602648258209229c-02 2. 8740130364894866940-02 7. 375375833362340927e-03 1. 620792150497436523e-01 1. 860560290515422821e-02 -
[0565] [0565] 3. 020260855555534363e-02 -4. 261482879519462585e-02 -9, 696666151285171509e-03 2. 610794007778167725e-01 5. 237784236669540405e-02 -
[0566] [0566] -6. 438077241182327271e-02 -5. 624282360076904297e-02 -1. 236270368099212646 € -01 2. 891128882765769958e-03 2. 071599289774894714e-02 -
[0567] [0567] 1. 238262653350830078e-02 3. 054698742926120758e-02 1. 157098542898893356e-02 3. 707341 104745864868e-02 2. 762936614453792572e-02 4. 094767197966575623e-02 -7. 498073391616344452e-04 1. 485649198293685913e-01 8.
[0568] [0568] 6. 749743223190307617e-02 -9. 576640278100967407e-02 -8. 900643 140077590942e-02 1. 242433562874794006 € -01 1. 291496679186820984e-02 -
[0569] [0569] 8. 800630271434783936ec-02 -6. 649091839790344238e-02 1. 850103493779897690e-03 1. 349143683910369873e-01 8. 689543604850769043e-02 2. 685183100402355194e-02 5. 982894822955131531e-02 7. 923950452594757080e-0241 201184689998626709e-01 -3. 650053963065147400e-02 1. 985712200403213501e-01 1. 220422834157943726 € -01 -9. 129960834980010986e-02 1. 014058887958526611e-01 -
[0570] [0570] -l. 446852809749543667e-03 1. 202998962253332138e-02 -9. 746513515710830688e-02 6. 614078884012997150e-04 9. 196152538061141968e-02 3. 069162741303443909e-02 2. 9703568667173385620-02 -5. 259514786303043365e-03 -
[0571] [0571] 8. 067373186349868774e-02 -7. 681368291378021240c-02 -8. 455868065357208252e-02 1. 426276266574859619e-01 -7. 180812954902648926e-02 -
[0572] [0572] -6. 705752760171890259e-02 3. 985116258263587952e0-03 -L. 342842914164066315e-02 1. 076473519206047058e-O01 -1. 137044057250022888e-01 2. 966187894344329834e-02 -4. 759540036320686340e-02 -4. 786271601915359497e-02
[0573] [0573] 8. 869623392820358276ec-02 4. 053802043199539185e-02 -9. 619213640689849854e-02 2. 316712141036987305e-01 5. 215295031666755676 € -02 -
[0574] [0574] 3. 673189505934715271e-02 -1. 116116493940353394e-01 -7. 136083394289016724e-02 1. 917887330055236816e-01 -1. 134073510766029358e-01 -
[0575] [0575] 9. 559250622987747192e-02 -7. 9597353935241699220-02 -9, 198858588933944702e0-02 -7. 598009705543518066 € -02 -7. 769442349672317505e-02 -
[0576] [0576] -4. 738411307334899902ec-02 -7. 464324682950973511c-02 -2. 145619690418243408e-02 -1.939605921506881714e-02 -3. 852173686027526855e-02 -
[0577] [0577] 1. 251333802938461304e0-01 1. 7111351713538169860-02 -6. 621837615966796875e-02 1. 742709130048751831e-01 1. 267599612474441528e-01 -
[0578] [0578] 4. 801169782876968384c-02 -2. 1656237542629241940-02 2. 336861798539757729e-03 -4. 761352017521858215e-02 -1.709268242120742798e-01 -
[0579] [0579] 1. 235087886452674866e-01 4. 510502144694328308e-02 -6. 849319487810134888e-02 1. 622317284345626831e-01 -2. 307317219674587250e-02 -
[0580] [0580] 5. 218897131271660328e-04 3. 362040594220161438e-02 -7. 485469430685043335e-02 6. 652375310659408569e-02 7. 509334385395050049e-02 4. 475534334778785706 € -02 -7. 961992919445037842e-02 -7. 417756132781505585e-03 -
[0581] [0581] 6. 550812721252441406 € 0-02 -2. 457906864583492279e-02 1. 473423372954130173e-02 2. 350640445947647095e-01 4. 311989620327949524e-02 -
[0582] [0582] 7. 583407312631607056e-02 5. 8839362114667892460-02 6. 493310211226344109e-04 1. 971419006586074829e-01 7. 102779299020767212e-02 3. 261463716626167297e-02 -3. 967470303177833557e-02 1. 788157671689987183e-01 2. 145568840205669403e-02 1. 741301454603672028e-02 5. 528895929455757141e-02 7. 309927046298980713e-02 -9. 212677925825119019e-02 -2. 044030837714672089c-02
[0583] [0583] 1. 042592599987983704e-01 -4. 190626740455627441 € -02 -6. 503871828317642212e-02 1. 043057069182395935e-01 8. 489029109477996826e-02 1. 045334897935390472e-02 -1. 161475898697972298e-03 8. 924265205860137939e-02 1. 019122675061225891e-01 -6. 539917737245559692e-02 3. 6688387393951416020-02 1. 004384011030197144e-01 -5. 178530514240264893e-02 1.527451910078525543e-02 1. 092369854450225830e-01 1. 408441457897424698e-02 6. 843966245651245117e-02 -
[0584] [0584] 7. 363800704479217529c-02 -3. 8667567074298858640-02 -2. 001976221799850464e-02 6. 322881579399108887e-02 6. 912150979042053223e-02 -
[0585] [0585] 1. 080348789691925049e-01 -6. 623218953609466553e-02 3. 600075840950012207e-02 1.284252852201461792e-01 -3. 356804326176643372e-029., 263959527015686035e-02 -8. 257491 886615753174ec-02 6. 327088922262191772e-02 7. 886464707553386688e-03 -1. 121233589947223663e-02 1. 059004198759794235e-02 -
[0586] [0586] 1. 073024421930313110e-01 7. 107981294393539429e-02 -2. 021136134862899780e-02 2. 566446959972381592e-01 8. 707801997661590576e-02 2. 006691321730613708e-02 4.312979057431221008e-02 -1. 104042585939168930e-023.
[0587] [0587] -8. 656753599643707275e-02 1. 018067449331283569e0-01 -2. 614619210362434387e-02 -9. 455595910549163818e-02 -1.633726246654987335e-02 -
[0588] [0588] 1. 7252614721655845640-02 6. 662292778491973877e-02 -2. 930153533816337585e-02 9. 8263353109359741210-02 5. 6623466312885284420-02 -
[0589] [0589] -5. 4516810923814773560-03 2. 291524782776832581e-02 |. 425514370203018188e-02 -2. 14970503002405 1666 € -02 7. 958848029375076294e-02 2. 434893324971199036 € -02 -4. 137597605586051941c-02 -1. 168355718255043030e-02 -
[0590] [0590] 1. 569024845957756042c-02 -7. 961544394493103027e-02 5. 635188147425651550e-02 5. 424044094979763031e-03 -4. 304505512118339539e-02 1. 132261473685503006 € -02 3. 339326009154319763e-02 6. 642406433820724487e-02 6. 284601241350173950e-02 3. 125973790884017944e-02 1. 13606005907058715835-O24 125018373131752014e-01 3. 000873699784278870e-02 4. 713540151715278625e-02 3. 896022215485572815e-02 6. 665597856044769287e-02 4. 346607998013496399e-02 4. 07105609774589538633803803803803 . 200433224439620972e-01 -
[0591] [0591] 6. 410059332847595215e-02 -5. 952033028006553650e-02 -9, 761091321706771851e-02 6. 631826609373092651e-02 1. 014157906174659729e0-01 -
[0592] [0592] 1. 548477858304977417e-01 -1. 4255004003643989560-02 3. 877667337656021118e-02 2. 237478345632553101e-01 -1. 605893857777118683e-025. 969559028744697571e-02 -2. 583224512636661530e-02 3. 610627725720405579e-02 6. 700634211301803589e-02 -1. 541895885020494461e-02 6. 525963544845581055e-02 1. 048436388373374939e-02 -8. 1767819821834564210-02 2. 330836802721023560e-01 5. 1721684634685516360-02 -5.270288512110710144e-02 9. 044290333986282349e-02 6. 489664316177368164e-02 1. 571478694677352905e-01
[0593] [0593] 8. 0051898956298828120-02 6. 1449766159057617190-02 -5. 113256350159645081e-02 -2. 6093356311321258540-02 7. 398383226245641708e-03 -
[0594] [0594] 4. 734942689538002014c-02 -5. 4834738373756408690-02 A. 672101512551307678e-02 1. 084081605076789856e-01 8. 236353099346160889e-02 -
[0595] [0595] 5. 169199779629707336c-02 -6. 593367457389831543e-02 -9, 588603675365447998e-02 9. 783164598047733307e-03 8. 1079334020614624020-02 5. 546050146222114563e-02 -6. 9509632885456085210-02 2. 7366543188691139220-02 -
[0596] [0596] 3. 841124102473258972e-02 2. 131873928010463715e-02 6. 979493051767349243e-02 3. 7366751 58143043518e-02 -1. 914822496473789215e-02 -
[0597] [0597] 2. 8044722974300384520-02 -8. 168220520019531250ec-02 -9. 641241282224655151e-02 -7. 929468899965286255e-02 -1. 514580249786376953e-01 -
[0598] [0598] 9. 700497239828109741e-02 -4. 696951806545257568e-02 4. 669659771025180817e-03 1. 510975956916809082e-01 5. 072569847106933594e-02 -
[0599] [0599] 7. 630144059658050537e-02 -9, 491692297160625458e-03 -4. 504972323775291443 € -02 3. 233884274959564209e-02 8. 812027052044868469e-03 -
[0600] [0600] 1. 299909651279449463e-01 -2. 524033747613430023e-02 -2. 599593950435519218e-03 1. 275358945131301880e-01 6. 2992289662361145020-02 -
[0601] [0601] -5. 808926001191139221e-02 6. S518135964870452881e-02 3. 278494626283645630e-02 -1. 400414184900000691e-04 7. 512911409139633179e-024. 255261272192001343e-02 3. 084849193692207336e-02 5. 6438479572534561 160-02 3. 860083594918251038e-02 -2. 324638515710830688e-02 1.297881710343062878e-03 5. 209069326519966125e-02 -2. 134633809328079224e-02 1. 188031136989593506 € -01 -
[0602] [0602] 1. 946093440055847168e-01 6. O014484167098999023e-02 2. 385893464088439941e-02 4. 579062759876251221e-01 1. 081116423010826111e-01 3. 956614062190055847e-02 2. 72622387856 . 816705897450447083e-02 7. 3563754558563232420-02 4. 426819086074829102e-02 -1. 030797213315963745e-01 9. 078910201787948608e-02 1. 307001709938049316e € -01 -7. 997239381074905396e-02 3. 013371825218200684e-01 1. 692214049398899078e-02 8. 071990311145782471e-02 6. 452175974845886230e-02 -
[0603] [0603] -1. 002706121653318405e-02 6. 506037712097167969e-02 6. 586498022079467773e-02 4. 155382141470909119e-02 -1.809219270944595337e-02 8. 314642123878002167e-03 -5. 680640041828155518e-02 3. 005709033459424973e-03 -
[0604] [0604] 3. 322812728583812714e-03 -2. 5529934093356132510-02 -9. 882508218288421631e-02 1. 003189831972122192e-01 9. 989196062088012695e-02 -
[0605] [0605] 9. 217460453510284424e € -02 -1. 165473181754350662 € 0-02 4. 278072714805603027e-02 1. 682393550872802734e-01 -3. 166247904300689697e-02 3. 716272488236427307e-02 -6. 042947992682456970e-02 1. 615374237298965454e-01 -
[0606] [0606] -3. 8741290569305419920-02 1. 244647428393363953e-02 -6. 359701603651046753e-02 4. 465615376830101013e-02 -4. 253571107983589172e-02 -
[0607] [0607] 1. 591212749481201172e-01 1. 051592268049716949e-02 -1. 098365634679794312e-01 1. 926321685314178467e-01 4. 645079001784324646 € -02 -
[0608] [0608] -6. 660576164722442627e-02 8. 156086504459381104e-02 3. 107530437409877777e-02 -7. 114225625991821289e-02 8. 393819630146026611e-02 -
[0609] [0609] -5. 588078498840332031c-02 2. 170677110552787781e-02 4. 077598825097084045e-02 1. 586921662092208862e-01 7. 307216525077819824c-02 -
[0610] [0610] 3. 863072022795677185e-02 1. 2101710587739944460-01 -5. 690059065818786621e-02 2. 843152545392513275e-02 -5. 9845041 48364067078e-02 1. 008045077323913574e-01 -1. 6940686851739883420-02 -8. 647721260786056519c-02
[0611] [0611] -1. 128495391458272934e-02 6. 658480316400527954e-02 -8. 003281801939010620e-02 9. 821167588233947754e-02 -5. 701660737395286560e-02 -
[0612] [0612] -2. 365992777049541473e-02 -9. 851229935884475708e-02 -5. 916777625679969788e-02 5. 805382132530212402e-02 -4. 505416750907897949e-02 1. 406664494425058365e-02 -2. 251774258911609650e-02 -5. 610410124063491821e-03
[0613] [0613] -l. 6405818983912467960-02 2. 667169831693172455e-02 2. 013008855283260345e-02 1. 611980646848678589e-01 9. 317573904991149902e-02 -
[0614] [0614] -5. 1254425197839736940-03 1. 506499480456113815e-02 -2. 558741904795169830e-02 1. 462682485580444336e-01 -9. 878762811422348022e-02 -
[0615] [0615] 3. 5035077482461929320-02 -1. O01865882426500320c-02 3. 559166938066482544e-02 1. 915801167488098145e-01 9. 055941551923751831e-02 -
[0616] [0616] 7. 949082180857658386ec-03 -8. 485835790634155273e-02 -5. 56620657444000244] 1e-02 8. 004585653543472290e-02 2. 482327446341514587e-02 -
[0617] [0617] 1. 366416737437248230c-02 -1. 924183592200279236 € -02 -1. 960823498666286469e-02 1. 043053418397903442e-01 -6. 781245768070220947e-02 -
[0618] [0618] 3. 104852139949798584c-02 -5. 838524922728538513e-02 6. 267432868480682373e-02 -1. 264576762914657593e-01 -3. 062127158045768738e-02 -
[0619] [0619] 1. 435246616601943970e-01 -5. 318130180239677429ec-02 4. 026373475790023804e-02 6. 740230321884155273e-02 8. 847470581531524658e-02 -
[0620] [0620] 1. 396159455180168152e-02 3. 792616352438926697e-02 -9, 046936780214309692e-02 8. 054792881011962891e-03 -1.042952761054039001e-01 8. 577302843332290649ec-02 -1. 137505993247032166 € -01 4. 098758846521377563e-02 -
[0621] [0621] -3. 319167345762252808e-02 3. 3396475017070770260-02 8. 598500862717628479e-03 5. 736464634537696838e-02 1. 492953300476074219e-02 -
[0622] [0622] 2. 159089595079421997e-02 -8. 8114507496356964110-02 8. 473745733499526978e-02 -7. 631704211235046387e-02 -9. 931492060422897339e-02
[0623] [0623] 7. 802469283342361450ec-02 -5. 6262083351612091060-02 4. 7793641686439514160-02 3. 923513889312744141e-01 8. 162643760442733765e-02 6. 096827238798141479e-02 5. 701934918761253357e-02 1. 206197291612625122e-02. 4975391 10481739044e-02 1.712065786123275757e-015. 450910329818725586e-02 -6. 056093424558639526e-02 6. 055634841322898865e-02 9. 083182364702224731e-02 -1. 494379248470067978e-03 2. 484335154294967651e-01 -
[0624] [0624] 7. 5566887855529785160-02 -3. 337087109684944153e-02 -3. 045510873198509216 € -02 -1. 968364790081977844e-02 -1. 116561703383922577e-02 -
[0625] [0625] 6. 058611907064914703e-04 -3. 103320114314556122e-02 2. 656221948564052582e-02 1. 079306006431579590e-01 -3. 107092343270778656e-02 -
[0626] [0626] 1. 001213937997817993e-01 -1. 229761168360710144e-02 4. 109898209571838379e-02 4. 498318955302238464e-02 -1. 095769405364990234e-01 -
[0627] [0627] 1. 295266002416610718e-01 -8. 622201532125473022e-02 3. 102386184036731720e-02 1. 131484732031822205e-01 -5. 471796914935112000e-02 -
[0628] [0628] 8. 675645291805267334e-02 -5. 2561856806278228760-02 -4. 265268146991729736 € -02 1. 478209495544433594e-01 1. 0209505259990692] 14e-01 6. 695374101400375366e-02 -8. 521109074354171753e-02 1. 940848231315612793e-01 6. 287954002618789673e-02 -1.023123562335968018e-01 4. 783285036683082581e-02 1. 044344380497932434 € -01 5. 7773686945438385024-02 227406233549118042e-02 1. 078769713640213013e-01 -
[0629] [0629] 5. 429123528301715851e-03 -6. 943853199481964111e0-02 -2. 902946725953370333e-04 -4. 248057026416063309e-03 6. 811551749706268311e-024. 208919033408164978e-02 -1. 388525664806365967e-01 -2. 576255239546298981e-02 -
[0630] [0630] 1. 564420759677886963e-01 -5. 513710156083106995e-02 1. 071530859917402267e-02 1. 713315099477767944e-01 2. 634437382221221924 € -02 5.
[0631] [0631] 1. 030635759234428406 € -01 -3. 223283914849162102e-03 -L. 0665047 16873168945e-01 5. 582571402192115784e-02 -9. 947061538696289062e-02 -
[0632] [0632] 1. 342199295759201050e-01 3. 441995009779930115e-02 -9. 9906653 16581726074e-02 1. 280367970466613770e-01 7. 272629439830780029e-02 6. 233214959502220154e-02 3. 122546523809432983e-02 3. 256123885512351990e-02 6. 708292663097381592e-02 -4. 362672939896583557e-02 5. 790472030639648438e-02 1. 102650091052055359e-01 -3. 886394947767257690e-02 1. 760566383600234985e-01 1. 014340743422508240e-01 -6. 293843686580657959e-02 1. 691797226667404175e-01 -
[0633] [0633] -1. 672632247209548950ec-01 -1. 047450974583625793e-01 -6. 506739556789398193e-02 -9. 538714587688446045e-02 -1. 677507758140563965e-01 -
[0634] [0634] 2. 19372715801.0005951e-02 3. 9397023618221282960-02 -6. 559856235980987549e-02 5. 877260863780975342e-02 -1. 2080452591 18080139e-01 -
[0635] [0635] 7. 533901929855346680e-02 2. 642023190855979919c-02 -9, 743896871805191040e-02 8. 770319819450378418e-02 1. 891937665641307831e-02 4. 641436040401458740e-02 -3. 524236846715211868e-03 4. 178847372531890869ec-02 -
[0636] [0636] 7. 959018647670745850e-02 4. 169445112347602844e-02 -6. 460010260343551636 € -02 1. 554234623908996582e-01 -7. 290890067815780640e-02 3. 489479422569274902e-02 -9. 850801527500152588e-02 -1. 026795525103807449e-02 -
[0637] [0637] 3. 192191198468208313e-02 -6. 4420863986015319820-02 8. 119575679302215576e € -03 2. 871187869459390640ec-03 -3. 587030991911888123e-02 -
[0638] [0638] -1. 451991870999336243c-02 -8. O18681406974792480e-02 -2. 602921612560749054e-02 -4. 110972210764884949ec-02 2. 222760207951068878e-02 5. 361549183726310730e-02 3. 39202880859375.0000e-02 -4. 800372198224067688e-02
[0639] [0639] 9. 547582268714904785e-02 -1. 113114058971405029e-01 -5. 739761143922805786e-02 -1.673143170773983002e-02 2. 204282395541667938e-02 7. T49646436423063278e-03 3. 449622541666030884e-02 2. 769745700061321259e-02 1. 12108312547206876 . 198964491 486549377e-01 5. 921980738639831543e-02 -
[0640] [0640] 1. 340604573488235474e-01 -4. 260551556944847107e-02 -4. 348991438746452332e-02 4. 856496751308441162e-01 2. 293084934353828430e-02 -
[0641] [0641] 7. 919710129499435425e-02 6. 295375525951385498e-02 -3. 962960839271545410e-02 1. 212137416005134583e-01 7. 765172421932220459e-02 -
[0642] [0642] -4. 3753227218985557560-03 7. 131621241569519043e-02 &. 360573090612888336e-03 7. 869817316532135010e-02 7. 877551019191741943e-02 -
[0643] [0643] 8. 428846299648284912e0-02 -3. 651213273406028748e-02 -4. 156636074185371399e-02 1. 192366331815719604e-01 3. 010158799588680267e-02 1. 614195480942726135e-02 1. 547606661915779114e-02 7. 073902338743209839e-02 3. 007609955966472626. 483487248420715332e-02 5. 117449164390563965e-02 1.
[0644] [0644] 1. 226420849561691284e-01 -2. 452150359749794006 € -02 -1. 083258688449859619e-01 3. 438846170902252197e-01 -5. 619533732533454895e-023. 477991 744875907898e-02 -6. 055117025971412659e-02 1. 042058169841766357e-01 1. 069127395749092102e-01 5. 739085376262664795e-02 8. 653866499662399292e-02 1. 160802841186523438e-01 -6. 480787415057420731c-03 -3. 996894136071205139c-02
[0645] [0645] 8. 253529667854309082e-02 -9, 485039860010147095e-02 3. 994449973106384277e-02 4. 1987970471382141116c-02 8. 269973844289779663e-02 9. 255757555365562439e-03 1. 3763904571533 14869808405637741 1e-02 8. 684209734201431274e-02 7. 286737114191055298e-02 1. 172644868493080139e-01 -1. 470092218369245529e-02 -
[0646] [0646] 7. T788150757551193237e-02 -3. 450091183185577393e-02 -L1. 238269545137882233e-02 2. 326644808053970337e-01 3. 634910797700285912e-03 -
[0647] [0647] 2. 194721065461635590e-02 -8. 713315427303314209e0-02 -5. 573138594627380371e-02 2. 125017344951629639e-01 9. 038160741329193115e-02 1. 275798713322728872e-04 -1. 114714071 15459442] 1e-01 1. 354486793279647827e-01 2. 855063416063785553e-02 -8. 064810186624526978e-02 3. 022887744009494781e-03 5. 090716108679771423e-02 -7. 0056922733783721920-02 1. 198583748191595078e-02 1. 221261322498321533e-01 -9. 968652576208114624e-02 2. 094873189926147461e-01 2. 072165720164775848e-02 1. 753466725349426270e-01 2. 057920396327972412e-01 -
[0648] [0648] 1. 453088223934173584e-01 -5. 449109245091676712e-03 5. 018474906682968140e-02 7. 028742134571075439e-02 -5. 345717072486877441e-025. 088078603 148460388e-02 -8. 291936479508876801ec-03 3. 690007328987121582e-024.
[0649] [0649] 1. 0725400596857070920-01 -1. 516697183251380920c-02 -7. 685118913650512695e-02 1. 048609092831611633e-01 8. 317070454359054565e-02 -
[0650] [0650] 8. 5697837173938751220-02 -6. 456696242094039917e-02 5. 581008270382881165e-02 8. 831838518381118774e-02 -9. 745343029499053955e-02 -
[0651] [0651] 8. 235121518373489380c-02 -6. 103688105940818787e-02 5. 268296226859092712e-02 3. 574078679084777832e-01 -2. 792743965983390808e-02 -
[0652] [0652] 4, 419054090976715088e-02 -5. 999341234564781189c-02 4. 708605632185935974e-02 1. 2955576181411743166-01 2. 330743148922920227e-02 6. 779911369085311890e-02 -7. 442694902420043945e-02 8. 111421018838882446e-02 1. 083712503314018250e-01 -8. 835159242153167725e-02 3. 288286179304122925e-024. 891288653016090393e-02 -5. 482118949294090271e-02 -2. 781130187213420868e-02
[0653] [0653] 1. 151493564248085022e0-01 -8. 854280412197113037e-02 7. 444263435900211334e-03 4. 409644305706024170e-01 7. 589400559663772583e-02 -
[0654] [0654] -9, 584771841764450073e-02 1. 0524783283472061160-03 -L. 148735638707876205e-02 -1. 014407724142074585e-01 -6. 162538379430770874e-02 -
[0655] [0655] -2. 004251815378665924e-02 -1. 227143593132495880e-02 -3. 250823169946670532e-02 1. 290509849786758423e-01 -1. 894178986549377441e-02 -
[0656] [0656] 5. 371837317943572998e-02 1. 984719187021255493c-02 3. 4888476133346557620-02 2. 288290709257125854e-01 7. 448182255029678345e-02 -
[0657] [0657] 1. 8309038877487182620-01 2. 9713731259107589720-02 -5. 59561401605606079] 1e-02 2. 122479528188705444e-01 1. 042133793234825134e-01 -
[0658] [0658] -2. 215764671564102173e-02 -1. 4305577613413333890-02 3. 561086580157279968e-02 1. 048702225089073181e-01 2. 319722063839435577e-02 9. 408680349588394165e-02 2. 020253613591194153e-02 -3. 528528660535812378e-02 1. 140273213386535645e-01 -7. 566191256046295166 € -03 -1. 197381317615509033e-02
[0659] [0659] 3. 205818682909011841c-02 -2. 063759602606296539c-02 6. 190691143274307251e-02 1. 104095280170440674e-01 1. 307551562786102295e-02 -
[0660] [0660] 8. 3220593631267547610-02 8 2588382065296173100-03 7. 613412290811538696 € -02 2. 549467794597148895e-02 -1. 116014420986175537e-01 2. 256155945360660553e-02 -4. 514684900641441345e-02 5. 5417947471 141815190-02 6. 225881725549697876e-02 1. 067192628979682922e-01 3. 177972510457038879e-02 1. 105888858437538147e-01 -5. 666538327932357788e-02 2. 396272681653499603e-02 -
[0661] [0661] 1. 612239927053451538e-01 -7. 047510147094726562e-02 -3. 360250219702720642e-02 3. 273741900920867920e-01 -3. 806498274207115173e-02 -
[0662] [0662] 1. 276615858078002930e-01 -5. 822714790701866150ec-02 3. 593113645911216736 € -02 1. 315664350986480713e-01 9. 617247432470321655e-02 1. 469378825277090073e-02 -4. 565301910042762756e-02 1. 958173066377639771e-01 8. 658401668071746826e-03 -1. 114058196544647217e-01 6. 890913099050521851e-02 6.
[0663] [0663] 1. 340765302302315831e-04 -5. 800437182188034058e-02 -4. 660554975271224976 € -02 -2. 404395304620265961e-02 -6. 305926293 134689331e-02 -
[0664] [0664] 1. 396090090274810791e-01 -7. 593826204538345337e-02 5. 149894952774047852e-02 3. 943829238414764404e-01 -3. 074555285274982452e-02 2. 078612707555294037e-02 -3. 2323624938726425 17-02 6. 142423674464225769e-02 -
[0665] [0665] 8. 499234169721603394e-02 5. 0176575779914855960-02 5. 342638865113258362e-02 2. 376019656658172607e-01 -6. 274843961.000442505e-02
[0666] [0666] 1. 3015779852867126460-01 -9. 600681252777576447e-03 -4. 28121909499] 1683960e-02 1. 799509525299072266 € -01 -6. 838180124759674072e-02 1. 4666629023849964] 14e-02 6. 014643982052803040e-02 1. 642630845308303833e-01 8. 843940496444702148e-02 1. 909289695322513580e-02 5. 347952246665954590
[0667] [0667] -9. 336955845355987549e-03 2. 113926969468593597e-02 -8. 190452307462692261e-02 1. 895388513803482056e-01 -1. 199507117271423340e-01 -
[0668] [0668] 6. 122607365250587463e-02 -8. 7279155850410461430-02 -8. 473993092775344849e-02 2. 793295681476593018e-01 -6. 156541407108306885e-02 9. 193333983421325684e-02 6. 325929611921310425e-02 1. 415537949651479721e-02 5. 998788774013519287e-02 -8. 655156940221786499e-02 5. 095520615577697754e-02 -
[0669] [0669] 1. 019959524273872375e-01 -8. 219117671251296997e-02 -1. 046140342950820923e-01 2. 064564377069473267e-01 2. 889804542064666748e-02 -
[0670] [0670] 2. 762375399470329285e-02 7. 1680605411529541020-02 |. 150116324424743652e0-02 9. 395629167556762695ec-02 -2. 276694215834140778e-02 -
[0671] [0671] -2. 585052046924829483c-03 -2. O08803561329841614c-02 4. 133007675409317017e-02 7. 173193246126174927e-02 7. 8430110588669776920-03 -
[0672] [0672] 6. 245182827115058899ec-02 6. 307031959295272827e-03 -9. 128043800592422485e-02 3. 415592014789581299e-01 2. 625980973243713379e-02 6. 590692698955535889e-02 -4. 4295471 16160392761e0-02 1.319349706172943115e-013. 780712187290191650e-02 1. 047811377793550491e-02 9. 692079573869705200e-02 8. 737251162528991699e-02 -5. 097502470016479492e-02 1.421207212843000889e-03 1. 387032270431518555e-01 -7. 888031750917434692e-02 2. 261597067117691040e-01 -
[0673] [0673] 2. 3456815630197525020-02 -2. 8732386417686939240-03 4. 675823822617530823e-02 6. 843237578868865967e-02 -6. 677216850221157074e-03 -
[0674] [0674] 1. 028850376605987549c-01 -8. 359704166650772095e-02 -2. 737238258123397827e-02 3. 021255433559417725e-01 6. 078061088919639587e-02 6. 618043780326843262e-02 1. 798987761139869690e-02 1. 356722861528396606 -01 01 21506921947002410910 468316748738288879e-02 2. 790845744311809540e-02 1. 028544306755065918e-01 2. 185394801199436188e-02 2. 124027609825134277e-01 -
[0675] [0675] 2. 997473813593387604e-02 -7. 856106758117675781e-02 5. 796376615762710571e-02 3. 399249166250228882e-02 -1. 870334334671497345e-02 -
[0676] [0676] -5. 179492384195327759e-02 6. 2788434326648712160-02 -3. 202200680971145630e-02 9. 530647844076156616 € -02 5. 449059978127479553e-02 -
[0677] [0677] 6. 075738742947578430e-02 4. 142096638679504395e-02 -6. 926563382148742676 € -02 1. 437340974807739258e-01 7. 6385110616683959960-02 -
[0678] [0678] 1. 776244342327117920e-01 7. 343861460685729980e-02 2. 689079754054546356 € -02 2. 788483500480651855e-O1 -4. 818086698651313782e-024. 526561 126112937927e-02 3. 323608636856079102e-02 3. 423786163330078125e-02 1.
[0679] [0679] -4. 439295828342437744e-02 -1. 1290637403726577760-01 4. 578836262226104736 € -02 7. 092023640871047974e-02 -1.328152716159820557e-01 5. 653227120637893677e-02 2. 657009847462177277e-02 5. 689545348286628723e-02 7.4079746-02 073094010353088379e-01 4. 281089641 153812408e-03 3. 583551570773124695e-02 6. 569854170083999634e-02 6. 393197923898696899e-03 -4. 6378735452890396120-023. 398638591170310974e-02 2. 2969622 16496467590e-02 6. 360525637865066528e-02 -
[0680] [0680] 9. 414638578891754150e-02 1. 411648374050855637e-02 -4. 442369099706411362e-03 1. 779196858406066895e-01 -5. 089190229773521423e-025. 580550804734230042e-02 3. 810810670256614685e-02 3. 522202000021934509e-02 2. 001058496534824371e-02 -2. 408905513584613800c-02 -1. 145134679973125458e-02
[0681] [0681] -5. 729510262608528137e-02 -5. 044311285018920898e-02 -1. 915148459374904633e-02 3. 758484125137329102e-02 -2. 628199197351932526e-02 -
[0682] [0682] 9. 029290825128555298e-02 -4. 2339906096458435060-02 5. 16447 1268653869629e-02 2. 600964307785034180e-01 -7. 634288165718317032e-03 -
[0683] [0683] 1. 573756486177444458e-01 -8. 467203378677368164e0-02 -5. 257217586040496826 € -02 2. 340736389160156250e-01 -1. 857174560427665710e-02 -
[0684] [0684] 2. 246714197099208832c-02 8 587028831243515015e-03 5. 566048435866832733e-03 -4. 164227470755577087e-02 3. 192691504955291748e-02 -
[0685] [0685] 6. 678205728530883789ec-02 -1. 434825733304023743e-02 1. 282991841435432434e-02 1. 334389895200729370e-01 -2. 323003485798835754e-03 -
[0686] [0686] 9. 400551021099090576e-02 7. 054000347852706909e-02 2. 346634678542613983e-02 1. 143360808491706848e-01 -5. 086684226989746094e-02 -
[0687] [0687] 1. 603710651397705078e-01 -9 ,. 2725396156311035160-02 6. 054875254631042480e-02 2. 265041321516036987e-01 4. 417140036821365356 € -02 1. 116480212658643723e-02 1. 296758651733398438e-02 1. 234373375773429834323
[0688] [0688] 1. 407949775457382202e-01 -1. 669958978891372681e-02 4. 338072612881660461e-02 8. 328308165073394775e-02 6. 284748017787933350e-02 6. 034487485885620117e-02 -8. 39890539646 14868 16ec-03 9. 709134697914123535e-02 2. 278674161061644554e-03 5. 275671835988759995e-03 2. 804535254836082458e-02 1. 229700595140457153e-01 5. 125224590301513672e-02. 626304239034652710e-02 1. 034147590398788452e-01 -5. 7260893285274505620-02 1.217947453260421753e-01 4. 615801945328712463e-02 1. 078893095254898071e-01 -3. 032624907791614532e-02 -
[0689] [0689] 1. 781539320945739746e0-01 -1. 829775422811508179c-02 -9, 171881526708602905e-02 1. 512420177459716797e-01 1. 097893621772527695e-02 -
[0690] [0690] 1. 160941421985626221e-01 -7. 6707936823368072510-03 -1. 666222885251045227e-02 2. 801343500614166260e-01 1. 496211811900138855e-02 3. 029396012425422668e-02 1. 583362743258476257e-02 1. 126550063490867615e-01 4. 582400619983673096 3. 476507589221.000671e-02 -8. 739483356475830078e-02
[0691] [0691] 2. 830021083354949951e-03 2. 210569009184837341e-02 2. 851567417383193970e-02 1. 8657457828521728520-01 2. 244805544614791870c-02 -
[0692] [0692] 1. 276761293411254883e-01 2. 580226026475429535e-02 3. 472609817981719971e-02 1. 363623738288879395e-01 -6. 762163341045379639e-02 -
[0693] [0693] 1. 248862445354461670ec-01 -3. 417943045496940613e-02 -1. 620046049356460571e-02 2. 219332605600357056e-01 2. 840258926153182983e-02 1. 337046828120946884ec-02 3. 480778634548187256e-02 1. 092524826526641846 € -01 9. 112270921468734741e-01. 111085489392280579e-01 7. 652253657579421997e-025. 671722069382667542e-02 -5. 7216256856918334960-02 -2. 509053051471710205e-02
[0694] [0694] 5. 163992196321487427e-02 -5. 1368262618780136110-02 6. 3675954937934875490-02 2. 16704234480857849] 1e-01 -1.299477415159344673e-03 1. 646113954484462738e-02 -2. 9858678579330444340-02 3. 4344568848609924320-023. 065217286348342896e-02 -9. 2557944357395172120-02 6. 617461 144924163818e-02 1. 031540483236312866 € -01 -1. 603079959750175476e-02 -3. 223963081836700439c-02
[0695] [0695] 1. 113368198275566101e-01 -6. 4796390943229198460-03 -8. 404289186000823975e-02 1. 930690705776214600e-01 6. 288302130997180939e-03 -
[0696] [0696] 4. 6913184225559234620-02 2. 950111590325832367e-02 -7. 082570344209671021e-02 -1. 559339091 181755066 € -02 3. 943201154470443726e-02 -
[0697] [0697] -1. 371251512318849564e-02 -4. 759683832526206970e-02 |. 775156520307064056 € -02 1. 215628832578659058e-O01 -2.298295125365257263e-023. 013341920450329781e-04 -7.972018793225288391ec-03 9. 156360290944576263e-03 5. 860886722803115845e-02 5. 938365310430526733e-02 1. 232472360134124756e € -01 1. 34724603700 055794447660446167e-02 -5. 110701173543930054e-02
[0698] [0698] 6. 697405874729156494e-02 3. 439865633845329285e-02 -7. 102423161268234253e-02 2. 445770353078842163e-01 1. 344645768404006958e-02 -
[0699] [0699] 3. 123677894473075867e-02 2. 794985845685005188e-02 7. 282314449548721313e-02 2. 032366245985031128e-01 7. 351617515087127686e-02 -
[0700] [0700] 8. 736465126276016235e-02 -3. 287238627672195435e-02 1. 337689720094203949e-02 2. 526172399520874023e-01 -2. 399312704801559448e-02 -
[0701] [0701] 8. 062319457530975342c-02 4. 239271860569715500e0-03 6. 023700907826423645e-02 5. 887019634246826172e-02 6. 904093921184539795e-02 -
[0702] [0702] -3. 899135440587997437e-02 -4. 4905211776494979860-02 5. 401611700654029846e € -02 -1. 115692686289548874e-02 -6. 833747774362564087e-02 -
[0703] [0703] 9. 111319482326507568e-02 -3. 988735750317573547e-02 -5. 259400978684425354e-02 2. 517829537391662598e-01 -4.627230390906333923e-02 3. 611111268401145935e-02 -1. 029011905193328857e-01 9. 317332506179809570e-02 -
[0704] [0704] 6. 556469947099685669c-02 -2. 555911988019943237e-02 6. 471869349479675293e-02 2. 062644213438034058e-01 6. 806883960962295532e-02 -
[0705] [0705] 1. 128311827778816223e-01 -6. 210741400718688965e-02 4. 086093604564666748e-02 5. 636319518089294434e-02 7. 3058858513832092290-02 -
[0706] [0706] 3. 751291334629058838e-02 -4. 764186218380928040c-02 -1. 867624558508396149e-02 6. 914572417736053467e-02 -2. 5637247017584741120-04 -
[0707] [0707] 1. 018291935324668884e-01 -8. 813277631998062134c-02 -9. 266402572393417358e-03 3. 241009712219238281e-01 5. 752012506127357483e-02 -
[0708] [0708] -5. 337104201316833496c-02 -1l. 014496684074401855e-01 6. 4248844981 19354248e-02 1. 454829890280961990e-02 -8. 9790351688861846920-02 2. 087452076375484467e-02 3. 7092618644237518310-02 9. 245935827493667603e-02 4. 071256145834922791e-02 -3. 541174158453941345e-02 1.2314202636480331420-01 6. 019562855362892151e-02 -3.203960135579109192e-02 3. 809405863285064697e-02 2. 165871858596801758e-02 -2. 736903959885239601e-03 -6. 034485250711441040ec-02 -
[0709] [0709] 1. 367841362953186035e-01 -1. 07744358479976654] e-01 -L1. 009812429547309875e-01 1. 66273936629295349] e-01 -4. 994849860668182373e-02 -
[0710] [0710] 7. 894173264503479004ec-02 -4 ,. 4192567467689514160-02 6. 487940996885299683e-02 1. 5157483518123626710-01 1. 125410571694374084e-01 -
[0711] [0711] 1. 065739616751670837e-01 7. 415091991424560547e-02 -7. 251530885696411133e-02 1. 449302881956100464e-01 -6. 252230377867817879e-04 -
[0712] [0712] 1. 543376594781875610e-01 -6. 160500645637512207e-02 -4. 616063460707664490e-02 3. 124710023403167725e-01 5. 980124231427907944e-03 2. 240843139588832855e-03 1. 573733426630496979e-02 3. 962474688887596130e-02 8.
[0713] [0713] -1l. 458687242120504379e-03 -1. 003080531954765320e-01 -6. 827123463153839111c-02 8. 321045339107513428e-02 2. 100150100886821747e-02 -
[0714] [0714] 2. 811610139906406403e-02 7. 4017859995365142820-02 4. 966225847601890564e-02 1. 516226530075073242e-01 2. 603001147508621216 € -02 4. 367440566420555115e-02 -3. 419990837574005127e-02 1. 285144686698913574e-01 -
[0715] [0715] 9. 688420593738555908e-02 -7. 8202456235885620120-02 -2. 762696146965026855e-02 1. 368550807237625122e-01 5. 598516389727592468e-02 5. 022092908620834351e-02 1.573096960783004761e-02 -5. 7553388178348541260-02 1. 809531077742576599e-02 1. 350062433630228043e-02 1.312227845191955566 € -01 3. 039092943072319031e-02 5. 852652713656425476e € -02 1. 5956154093146324160-02 -
[0716] [0716] 8. 361519873142242432c-02 4. 485816881060600281e-02 -8 &. 889878541231155396e-02 2. 928794622421264648e-01 5. 994814634323120117e-02 -
[0717] [0717] -4. 1535429656505584720-02 -6. 964607536792755127e-02 -L1. 092122402042150497e-02 -3. 010655939579010010e-02 5. 2844382822513580320-02 5. 340316146612167358e-02 -4. 560313001275062561e-02 3. 0591606628 14974785e-03 1. 189081892371177673e-01 8. 432922512292861938e-02 1. 382349133491516113e-01 2. 4924218654632568360-02 -7. 152648270130157471e-02 3. 921335563063621521c-02 -
[0718] [0718] -2. 220422402024269104e-02 -1. 692947186529636383e-02 -1. 039435490965843201 € -01 4. 583523422479629517e-02 1. 571316272020339966e-03 -
[0719] [0719] 2. 915714867413043976e-02 -2. 355684712529182434e-02 3. 882803022861480713e-02 1. 41329243779182434] 1e-01 1. 031245067715644836e € -01 -
[0720] [0720] 7. 334270328283309937e-02 -9, 6441455185413360600-02 3. 807788714766502380e-02 2. 856694757938385010e-01 1. 717728748917579651e-02 7. 869792729616165161e-02 6. 4.3533340941 026 4.353 3.392 -1. 025068014860153198e-01 1. 5S43858647346496582e0-01 -
[0721] [0721] 4. 299519211053848267e0-02 -7. 1824856102466583250-02 -3. 948076069355010986e € -02 7. 927553355693817139e-02 9. 608397632837295532e-02 8. 226821571588516235e-02 -3. 198806941509246826e-02 -9. 380115196108818054e-03 -
[0722] [0722] -6. 233932822942733765e-02 4. 701514169573783875e-02 -A4. 348801448941230774e-02 3. O80898895859718323e-02 5. 424711853265762329e-02 -
[0723] [0723] 6. 436422467231750488e-02 -1. 456975378096103668e-02 -9, 898027032613754272e-02 2. 176572680473327637e-01 4. 329755529761314392e-02 -
[0724] [0724] 1. 896340548992156982e-01 -4. 606541618704795837e-02 A4. 737361147999763489e-02 2. 570518851280212402e-01 3. 972181119024753571e-03 1. 888511143624782562e-02 -1. 628066040575504303e-02 1. 614133417606353760e-01 8.
[0725] [0725] 1. 371608227491378784e-01 5. 27363494038581 | 84810-02 4. 667719453573226929e-02 1. 234965845942497253e-01 -5. 918418243527412415e-02 -
[0726] [0726] 2. 893635630607604980e-02 -7. 828740030527114868e-02 -2. 689831517636775970e-02 3. 863329440355300903e-02 -2. 760453335940837860e-02 -
[0727] [0727] -8. 065184205770492554e-02 -6. 817293912172317505e-02 -7. 998153567314147949e-02 -7. 401459664106369019e-02 -5. 9774234890937805 18e-02 -
[0728] [0728] 1. 104459613561630249c-01 4. 076564684510231018ec-02 -4. 592858999967575073e-02 3. 506980836391448975e-01 7. 887510955333709717e-02 -
[0729] [0729] 2. 4084964767098426820-02 -1. 197756901383399963e-01 -1. 54091 8648242950439e-02 -6. 2681294977664947510-02 2. 578669041395187378e-02 -
[0730] [0730] 1. 005508750677108765e-01 -2. 592546120285987854e-02 -1. 177651286125183105e-01 1. 661281436681747437e-01 4. 365719854831695557e-02 -
[0731] [0731] 1. 740490943193435669e-01 1. 885643973946571350c-02 -8. 239676803350448608e-02 2. 358230799436569214e-01 5. 378657579421997070e-02 -
[0732] [0732] 8. 352466672658920288e-02 -2. 047137916088104248e-02 -7. 666187733411788940e-02 3. 832509741187095642e-03 8 &. 279196918010711670e-02 -
[0733] [0733] 8. 307155966758728027e-02 2. 2927330806851387020-02 -6. 182302162051200867e-02 6. 202301755547523499e-02 -6. 648864597082138062e-02 2. 331844344735145569e-02 -5. 587451159954071045e-02 8. 268135040998458862e-02 1. 320039331912994385e-01 -4. 858083277940750122e-02 9. 169662371277809143e-03 -
[0734] [0734] 9. 718924015760421753e-02 -7. 300005853176116943e-02 4. 439882561 5644454960-02 4. 232090711593627930e-02 -5. 3393807262182235720-025. 807453393936157227e-02 -2. 839324763044714928e-03 -8. 02646651 8640518188e-02 -
[0735] [0735] 3. 524347022175788879e0-02 2. 887188829481601715c-02 -8. 717320114374160767e-02 8. 825802803039550781e-02 8. 867309987545013428e-02 -
[0736] [0736] 3. 924714773893356323e-02 5. 087722092866897583e-02 -5. 516155064105987549e-02 8. 732569962739944458e-02 -6. 779053807258605957e-02 -
[0737] [0737] 2. 385249733924865723c-02 1. 5269871801137924190-02 4. 279951378703117371e-02 2. 192755341529846191e-01 -2. 972893789410591125e-02 -
[0738] [0738] 3. 812470287084579468e-02 -6. 550129503011703491e0-02 9, 989806264638900757e-02 -8. 0468162894248962400-02 -7. 357944548130035400e-02
[0739] [0739] -1. 433822792023420334c-02 -9. 684432297945022583e-02 -7. 060860097408294678e-02 -4. 704597592353820801e-02 4. 656309727579355240e-03 5. 253298580646514893e-02 -5. 769508890807628632e-03 -6. 859393417835235596e-02
[0740] [0740] 9. 071070700883865356e-02 -3. 261385858058929443e-02 -1. 676329039037227631e-02 8. 059843629598617554e-02 7. 099431008100509644e-02 -
[0741] [0741] 3. 9344415999948978420-03 6. 387247890233993530ec-02 -3. 847029060125350952e-02 1. 476632952690124512e-01 -1. 312008872628211975e-02 -
[0742] [0742] -8. 167605847120285034e-02 -6. 908456981182098389e-02 6. 801045034080743790e-03 1. 125824754126369953e-03 -9. 935430437326431274e-029. 948171675205230713e-02 -1. 636562496423721313e-02 -3. 657168010249733925e-03 -
[0743] [0743] -9. 877742826938629150ec-02 2. 942228317260742188e-02 |. 680879155173897743e-03 4. 085459746420383453e-03 1. 091316863894462585e-01 5. 646877363324165344e-02 -6. 892484426498413086e-02 -2. 822694927453994751e-02 -
[0744] [0744] 5. 181561037898063660e-02 -6. 442663818597793579ec-02 -4. 845743998885154724e-03 1. 211084946990013123e-01 9. 317204356193542480e-03 -
[0745] [0745] 1. 123085394501686096c-01 5. 317306146025657654e-02 -3. 326736390590667725e-02 3. 201057314872741699e-01 9. 3249998986721038820-02 -
[0746] [0746] 1. 716225780546665192e0-02 -1. 179818995296955109e-02 -1. 779097877442836761e-02 1. 603112220764160156e-01 -8. 539583534002304077e-02 -
[0747] [0747] -l. 455068704672157764e-03 7. 376357167959213257e-02 -2. 617021650075912476e € -02 5. 394358187913894653e-02 8. 710808306932449341] 1e-02 -
[0748] [0748] -1. 2182071805.00030518e-01 -2. 448439411818981171e-02 -L. 327233463525772095e-01 -3. 427767753601074219e-02 -7. 678963989019393921e-02 -
[0749] [0749] 2. 021701484918594360e-01 -1. 021339222788810730e-01 2. 725620754063129425e-02 4. 071777760982513428e-01 8. 595822751522064209e-02 -
[0750] [0750] 3. 036807058379054070ec-03 -1. 128899678587913513e-01 -2. 650267817080020905e-02 7. 4588559567928314210-02 -1. 058233156800270081e-02 -
[0751] [0751] 4. 0367547422647476200-02 3. 044705651700496674e-02 -5. 818421021103858948e-02 1. 409350484609603882e-01 -6. 319366395473480225e-02 -
[0752] [0752] 1. 830411553382873535e-01 1. 5314253978431224820-02 3. 127394989132881165e-02 2. 903991043567657471e-01 7. 895445823669433594e-02 -
[0753] [0753] 4. 739399626851081848ec-02 -5. 9373859316110610960-02 2. 202751860022544861e-02 2. 115803398191928864e-02 8. 448009938001632690e-02 1. 647618599236011505e-02 -5. 982827767729759216 € -02 3. 594325482845306396 € -02 5. 789876729249954224e-02 -8. 632205426692962646 € -02 5. 934059619903564453e-02 -
[0754] [0754] -2. 694814279675483704e-02 -1. 311984006315469742e0-02 -7. 988302409648895264e-02 1. 381005793809890747e-01 1. 7775010317564010620-02 -
[0755] [0755] 1. 054984927177429199e-01 -5. 4958067834377288820-02 -9, 810092300176620483e-02 7. 891445606946945 190e-02 -8. 348174393177032471e-02 1. 956523768603801727e-02 -7. 418994605541229248e-02 2. 141497284173965454e-02 6. 814574450254440308e-02 8. 450059592723846436 € -02 -1. 192546449601650238e-02 9. 307193756103515625e-02 3. 951892256736755371e-02 6. 773950159549713135e-02 1. 359145436435937881c-02 -2.256177738308906555e-02 7. 148170471191406250. 606521710753440857e-02 1. 22503511607646942] 1e-01 5. 540214106440544128e-02 -
[0756] [0756] 6. 1153713613748550420-02 -3. 7659555673599243160-02 2. 141224592924118042e-02 2. 561381843406707048e-04 2. 628492750227451324e-02 -
[0757] [0757] -5. 170010402798652649e-02 4. 7765955328941345210-02 9. 256732650101184845e-03 -2. 603030949831008911e-02 -6. 794042140245437622e-02 -
[0758] [0758] 1. 789241731166839600e-01 -8. 664064109325408936c-02 -8. 615125715732574463e-02 2. 633632421493530273e-01 -1. 550204865634441376e € -02 -
[0759] [0759] -9, 549888223409652710c-02 4. 412045329809188843ec-02 -5. S584447085857391357e-02 -1. 905645430088043213e-02 -9. 110018610954284668e-02 -
[0760] [0760] -4. 947012290358543396e-02 -7. 5545370578765869] 140-02 3. 001481853425502777e-02 2. 237997390329837799e-02 -8. 542436361312866211c-022. 268801443278789520e-03 2. 933747321367263794e-02 -1. 412031650543212891e-01 -
[0761] [0761] 1. 120156887918710709c-02 -8. 726066350936889648e-02 -A4. 278487339615821838e-02 -6. 255891919136047363ec-02 9. 387567639350891113e-03 -
[0762] [0762] 9. 359393268823623657e-02 -4. 877812787890434265e-02 -7. 489954680204391479e-02 2. 907581031322479248e-01 4. 011521115899085999e-02 1. 616817712783813477e-02 -5. 95656260848045349] 1e-02 1. 4629048 10905456543e-01 4. 229426011443138123e-02 -1. 172861754894256592e-01 1.879790723323822021e-01 4. 690033942461013794e-02 -2. 837636135518550873e-02 1. 008210182189941406e-01 8. 194539695978164673e-02 -3. 296364471316337585ec-02 2. 655319273471832275e-01 -
[0763] [0763] 5. 7150155305862426760-02 6. 4808882772922515870-02 1. 669282838702201843e-02 1. 202525794506072998e-01 6. 3957169651985168460-02 6. 400451809167861938e-02 5. 975314881652593677617 . 524275779724121094e-02 1. 314094066619873047e-01 -
[0764] [0764] 6. 225641816854476929c-02 -5. 282243341207504272 € 0-02 4. 127933457493782043e-02 -4. 712628200650215149e-02 7. 125623524188995361e-02 - 9, 528663754463195801 € -02 5. 110210552811622620e-02 -4. 803021997213363647e-02 -6. 043691188097000122e-02 4. 717775434255599976e-02 -6. 171289458870887756e- 02 5. 799505487084388733e-02 -9. 066579490900039673e-02 6. 874695420265197754e-02 -6. 4443394541 74041 748ec-02 4. 101144149899482727e-03 -
[0765] [0765] -2. 356477268040180206ec-02 -1. 091279685497283936c-01 -2. 329668262973427773e-03 9. 208843111991882324e-02 -1.239343285560607910e-01 1. 824101805686950684e-02 3. 561931103467941284e-02 1. 183044444769620895e-02 4. 296118021011352539-02 -11. 925368934869766235e-02 4. 39288690686225891 10-02 6. 38788640499] 1149902e-02 -4. 134950414299964905e-02 2. 823258005082607269e-02 7. 829239964485 1 68457e-02 -2. 60026641 1900520325e-02 4. 242255166172981262e-02 1. 201338991522789001e-01 1. 085190102458000183e-01 5. 821906402707099915e-02 -
[0766] [0766] 1. 558838933706283569e-01 5. 181598290801048279e-02 |. 124924141913652420 € -02 5. 181910097599029541e-02 -1. 441354164853692055e-03 4. 577650874853134155e-02 -9. 804491698741912842e-02 5. 727339163422584534e-02 -
[0767] [0767] 3. 567430004477500916c-02 5. 899872165173292160e-03 2. 455170080065727234e-02 1. 874354481697082520e-01 -9. 884224086999893188e-02 -
[0768] [0768] 5. T788689479231834412e-02 -4. 709724709391593933e-02 2. 889010123908519745e-02 1. 746908426284790039e-01 5. 048273876309394836e-02 4. 418041929602622986e-02 -8. 622613549232482910e-02 1. 256974488496780396e-01 9. 989419579505920410e-02 -1. 254588644951581955e-02 1. 312816441059112549e-01 9. 645503759384155273e-02 -1. 6973827034235.00061e-02 1. 609827280044555664e-01
[0769] [0769] -1. 067932881414890289c-02 1. 606239005923271179e-02 |. 514604967087507248e-02 -7. 041137665510177612e-03 4. 511967673897743225e-02 -
[0770] [0770] 1. 335728466510772705e-01 1. 8173431977629661560-02 2. 2465461 86506748199e-02 3. 2912933826446533200-01 -2. 676494652405381203e-03 -
[0771] [0771] 4. 445631057024002075e-02 -1. 023918539285659790e-01 -1. 117581315338611603e-02 1. 862979680299758911e € -01 -5. 066245794296264648e-02 8. 199230069294571877e-04 7. 692893967032432556e-03 4. 443145007826387882e-04 8. 139503747224807739e-02 -2. 634665183722972870e-02 9. 918421506881713867e-02 1. 093100681900978088e-01 3. 911758959293365479e-02 5. 851710587739944458e-02 3. 230680525302886963e-02 -3. 729220712557435036 € -03 2. 063122019171714783e-02 -
[0772] [0772] 8. 298905938863754272e-02 -7. S48166811466217041e-02 -9. 731073677539825439e-02 7. 676146179437637329e-02 -7. 419167459011077881e-02 -
[0773] [0773] 2. 9183723032474517820-02 -8. 0687731504440307620-02 4. 780821036547422409ec-03 1. 748999953269958496e-01 -4. 299267008900642395e-02 6. 467553973197937012e-02 -2. 30041 1090254783630e-02 1. 012868434190750122e-01 9. 312485903501510620e-02 -8. 379475772380828857e-02 1. 783030480146408081e-01 1. 210251897573471069e-01 -2. 531854435801506042e-02 -9. 623114019632339478e-02
[0774] [0774] -3. 535090759396553040e-02 -6. 7550726234912872310-02 -L. 075019463896751404e-01 -4. 104370623826980591e-02 4. 765389859676361084e-02 -
[0775] [0775] 3. 192454576492309570e-02 3. 307246789336204529e-02 6. 988157052546739578e-03 1. 579026132822036743e-02 6. 467562168836593628e-02 -
[0776] [0776] 4. 734509438276290894c-02 4. 632652923464775085e-02 -L1. 923647150397300720e-02 6. 391173601150512695e-02 7. 037382572889328003e-02 2. 465177886188030243e-02 -1.326739974319934845e-02 1. 130530908703804016e-01 -
[0777] [0777] 3. 043656051158905029e-02 -6. 0869596898555755620-02 9. 183261543512344360e-03 1. 070855632424354553e-01 3. 239284083247184753e-02 -
[0778] [0778] 9. 510526433587074280e-03 -1. 775580225512385368e-03 -5. 251288786530494690e-02 9. 377675596624612808e-04 -7. 234701514244079590e-02 4. 690110683441162109e-02 7. 393725961446762085e-02 -1. 300024427473545074e-02 -
[0779] [0779] 1. 435370892286300659e-01 -9, 650747478008270264e-02 1. 409323606640100479e-02 3. 022617101669311523e-01 3. 727115690708160400e-02 3. 261042758822441101e-02 5. 89585422015151515341515151515151515 5. 480205267667770386e-02 1. 147939339280128479e-01 4. 335230216383934021e-02 -6. 514886766672134399e-02 -1. 164478156715631485e-02
[0780] [0780] -2. 449510060250759125e-02 -2. 678451128304004669e-02 -3. 204980492591857910e-02 -5. 87920248508453369] 1e-02 -2. 5S48794494941830635e-03
[0781] [0781] -8. 133856207132339478e-02 -7. 968682050704956055e-02 -2. 232799679040908813e-02 -8. 912698924541473389e-02 -2. 702170796692371368e-02
[0782] [0782] 5. 622038245201110840e-02 1. 384506118483841419e0-03 -3. 178646415472030640 € -02 -3. 253740444779396057e-02 2. 842929959297180176e-02 -
[0783] [0783] 4. 876506701111793518e-02 4. 656245931982994080ec-02 -6. 872709095478057861e-02 9. 207368642091751099e-02 4. 006646573543548584e-02 7. 078028470277786255e-02 2. 194499596953392029e-02 1. 209477335214614868e-01 1. 320830732583999634e-01 430346637964248657e-02 4. 074481129646301270e-02 4. 221582785248756409e-02 4. 285238683223724365e-02 2. 575966157019138336e-02 &. 326494693756103516 € -02 9. 354557842016220093e-02 1. 698463708162307739e-01 - 9, 555853903293609619e-04 5. 1958080381 15501404e-02 5. 169894173741340637e-02
[0784] [0784] 6. 079989299178123474e-02 -5. 802433937788009644e-02 -3. 128459677100181580e-02 9. 179639071226119995e-02 8. S46645194292068481e-02 -
[0785] [0785] 9, 223939478397369385e-02 -7. 756586372852325439e0-02 2. 265060553327202797e-03 1. 173203717917203903e-02 6. 693723052740097046 € -02 -
[0786] [0786] 4. 041278362274169922e-02 -7. 590568065643310547e-02 3. 032279945909976959e-02 -2. 368949726223945618e-02 3. 753356263041496277e-023. 932601958513259888e-02 3. 488076850771903992e-02 3. 780684992671012878e-02 9. 119085967540740967e-02 -6. 106093525886535645e-02 4. 166825488209724426 € -02 1. 160556674003601074e-01 8. 207008987665176392e-02 2. 989473193883895874e-02 9. 503882378339767456e-02 -2. 137246541678905487e-03 -8. 050561696290969849e-02
[0787] [0787] 1. 623622179031372070e-01 5. 1486741751432418820-02 1. 992116495966911316 € -02 3. 487737476825714111e-01 -6. 785605102777481079e-02 -
[0788] [0788] 1. 190968900918960571e-01 6. 553560495376586914e-02 -1. 106525138020515442e-01 1.283518821001052856e € -01 -2. 475564368069171906 € -025. 395429208874702454e-02 2. 115803956985473633e-02 2. 082946151494979858e-02 6. 596980243921279907e-02 -9.272982180118560791e-029.871455281972885132e-02 1. 226999051868915558 506128072738647461e-02 8. 736342191696166992e-02 -
[0789] [0789] 3. 837553411722183228e-02 -7. 243610918521881104c-02 -2. 751245163381099701e-02 8 &. 738449960947036743e-02 6. 486799567937850952e-02 -
[0790] [0790] 7. T798896729946136475e-02 -3. 498914185911417007e-03 3. 720951825380325317e-02 -2. 186695858836174011ec-02 6. O80168485641479492e0-02 -
[0791] [0791] 1. 19845353066921234] 1e-01 -6. 251822225749492645e-03 -1. 926522701978683472e-03 3. 561934232711791992e-01 3. 679596260190010071e-02 -
[0792] [0792] 1. 328520029783248901e-01 -6. 633858382701873779c-02 -8. 514307439327239990e-02 1. 721683442592620850e-01 -2. 641975693404674530e-02 2. 220258302986621857e-02 -3. 96981947 1240043640e-02 1. 511895656585693359e-01 8. 655687421560287476e-02 3. 175893425941467285e-02 3. 417039662599563599e-02 1. 180513203144073486 € -01 -8. 425338566303253174e-02 2. 353817597031593323e-025. 307986959815025330e-02 2. 183817140758037567e-02 4. 296619445085525513e-02 -
[0793] [0793] 1. 062820181250572205e-01 -3. 021525964140892029c-02 3. 441819921135902405e-02 4. 781811684370040894e-02 -1. 090510040521621704e-01 6. 402806937694549561e-02 -9. 466243535280227661e-02 -4. 571276251226663589e-03 -
[0794] [0794] 1. 418828517198562622e-01 -6. 3668504357337951660-02 -3. 653027117252349854e-02 8. 309665322303771973e-02 -1. 022864249534904957e-03 -
[0795] [0795] 7. 168512791395187378e-02 -1. 219677738845348358e-02 -9, 493180364370346069e-02 2. 203011186793446541e-03 -9. 233405441045761 108e-02 -
[0796] [0796] -3. 323732689023017883e-02 7. 6217139139771461490-03 -2. 781111747026443481e-02 1. 904317177832126617e-02 -2. 3323595523834228520-02 -
[0797] [0797] -1. 737017557024955750e-02 -1. 127905324101448059e-01 -6. 822146475315093994e-02 -1. 461129751987755299e-03 4. 382872954010963440e-02 1. 104355230927467346 € -01 7.397020608186721802e-02 -4. 7953344881534576420-023. 069162741303443909e-02 -2. 441738266497850418e-03 1. 445270925760269165e-01 6. 761844269931316376e-03 5.733173713088035583e-02 -2. 2262923419475555420-02 1. 048047542572021484e-01 -4. 029197618365287781e-02 -1.599926035851240158e-03 -
[0798] [0798] -l. 734146289527416229e-02 6. 429485976696014404e-02 -5. 498339980840682983e-02 2. 179646305739879608e-02 7. 906819880008697510e-02 2. 928880229592323303e-02 5. 617837980389595032e-02 7. 073431462049484253e-02 -
[0799] [0799] 9. 475859999656677246c-02 -8. 615747094154357910c0-02 -5. 518051609396934509e-02 1. 193303018808364868e-01 -7. 390332967042922974e-02 7. 092832773923873901e-02 5. 151550099253654480e-02 9. 900908172130584717e-02 3. 071170300245285034e-02 -2. 205214090645313263e-02 -5. 654016975313425064e-03
[0800] [0800] 4. 8419438302516937260-02 5. 208652000874280930e-03 -4. 979505389928817749e-02 1. 631936579942703247e-01 1. 7704573692753911020-03 -
[0801] [0801] -9, 523947536945343018e-02 -3. 6273680627346038820-02 -8. 506381511688232422e0-02 -9. 913276880979537964e-02 -2. 3421965539455413820-02
[0802] [0802] -5. 690314993262290955e-02 -5. 866133142262697220e0-03 -7. 534471899271011353e-02 -9. 141033887863159180e-02 -4. 884926229715347290e-02
[0803] [0803] 2. 729833498597145081e-02 6. 4156584441661834720-02 |. 215984020382165909 € -02 1. 324468664824962616 € -02 5. 6133788079023361210-02 -
[0804] [0804] 1. 144525185227394104e0-01 1. 815601950511336327e-03 -L. 817790046334266663e-02 1. 268979012966156006 € -01 5. 706354975700378418e-02 6. 659369170665740967e-02 5. S95168564468622208e-03 9. 750044345855712891e-02 -
[0805] [0805] -3. 593383356928825378e-02 6. 626287102699279785e-02 2. 534332312643527985e-02 2. 729226648807525635e-02 -7. 638572156429290771e-02 -
[0806] [0806] 9. 698417969048023224e-03 -1. 176378056406974792e-01 6. 851708143949508667e-02 1. 194142103195190430e-01 8. 471126854419708252e-02 1. 802830770611763000e-02 3. 796647861599922180e-02 -4. 126276448369026184e-02 -
[0807] [0807] 1. 082066330127418041e-03 4. 376259632408618927e-03 6. 469377875328063965e-02 5. 911303684115409851e-02 1. 136291492730379105e-02 6. 073951721191406250e-02 -4. 855808243155479431e-02 -6. 900091469287872314e-02
[0808] [0808] 5. 947508662939071655e-02 -1. 025736182928085327e-01 -2. 945050597190856934e-02 2. 427973896265029907e-01 4. 612592980265617371e-02 -
[0809] [0809] -1. 413277816027402878e-02 -3. 187746554613113403e-02 3. 241734579205513000e-02 2. 837860397994518280e-02 4. 733189195394515991e-02 -
[0810] [0810] 1. 931124739348888397e-02 -7. 769563794136047363e-02 -8. 186532557010650635e-02 2. 006873041391372681e-01 -4.230597615242004395e-02 4. 373828694224357605e-02 -5. 577155947685241699e-02 -4. 772470146417617798e-02 -
[0811] [0811] 1. 803911291062831879ec-02 -5. 983989313244819641e-02 -9. 243328124284744263e-02 1. 108967810869216919e-01 -1.260842233896255493e-01 4. 230503737926483154e-02 -3. 601120784878730774e-02 4. 616792872548103333e-023. 417435288429260254 € -02 -1. 101735047996044159e-02 1. 1553633213043212890-01 4. 978111013770103455e-02 -3. 779812529683113098e-02 -1.214278209954500198e-02 -
[0812] [0812] 1. 183494850993156433e-01 -5. 630921944975852966 € 0-02 1. 258494239300489426 € -02 3. 422466814517974854e-01 4. 381413385272026062e-02 -
[0813] [0813] 2. 059189788997173309e-02 5. 821876227855682373e-02 A. 573731 8694591522220-02 1. 482729166746139526e-01 8. 230062574148178101ec-02 -
[0814] [0814] -3. 917984664440155029e-02 -9. 252244979143142700e-02 -5. 226548016071319580e-02 9. 673545509576797485e-02 -3. 979641571640968323e-02 1. 604867167770862579e-02 -1. 021478623151779175e-01 1. 286221742630004883e-01 -
[0815] [0815] 4. 0776897221803665160-02 2. 337147668004035950e-02 7. 450492680072784424e-02 5. 563324317336082458e-02 2. 7335556223988533020-02 8. 869221061468124390e-02 -3. 245648741722106934e-02 5. 835152789950370789e-02 1. 359995454549789429e-01 -2.725062193349003792e-03 4. 269409924745559692e-02 4. 483936354517936707e-02 -1. 069996878504753113e-01 1. 5013688802719116210-018. 724023401737213135e-02 4. 422161355614662170e-02 3. 752190619707107544e-02 -
[0816] [0816] 3. 3867150545120239260-02 6. 2944851815700531010-02 -7. 264667749404907227e-02 -1. 029711514711380005e-01 -7. 906132191419601440e-02
[0817] [0817] 7. 913240790367126465e-02 -3. 593139350414276123e-02 6. 983150541782379150e-02 1. 149878352880477905e-O01 -1. 091468334197998047e-01 -
[0818] [0818] 8. 929062634706497192c-02 -4. 8078887164592742920-02 -6. 155391037464141846 € -02 9. 159966371953487396e-03 1. 749425753951072693e-02 1. 229404658079147339e-01 -5. 475944653153419495e-02 1. 310101896524429321e-02 -
[0819] [0819] 1. 544812042266130447e-02 1. 8588557839393615720-02 -9. 394458681344985962e0-02 1. 150341257452964783e-01 6. 045040115714073181e-02 -
[0820] [0820] 1. 074690222740173340e-01 3. 676292672753334045e-02 -2. 769394218921661377e-02 1. 602123826742172241e-01 -2. 634712308645248413e-02 -
[0821] [0821] -1. 937528327107429504e-02 -7. 226168643683195114e-03 -3. 447532281279563904e-02 3. 676054254174232483e-02 -2. 856829948723316193e-02 1. 179451961070299149ec-02 -7. 8698851 16815567017e-02 3. 008047677576541901e-023. 7024248391 38984680e-02 2. 603819593787193298e-02 6. 045524775981903076e-02 -
[0822] [0822] -7. 811500132083892822e-02 -9 ,. 420144557952880859e-02 |. 133587211370468140e-02 -1.266835778951644897e-01 8. 419659733772277832e-02 6. 969964504241943359e-02 -4. 989846050739288330e-02 7. 161977142095565796e-02 -
[0823] [0823] 4. 971658065915107727e-02 5. 711320042610168457e-02 -6. 591724604368209839ec-02 9. 058685600757598877e-02 -5. 485069379210472107e-02 -
[0824] [0824] 9. 388184547424316406c-02 -6. 535119563341140747e0-02 -9. 216629900038242340e-03 3. 881492838263511658e-02 -6. 139004603028297424e-02 -
[0825] [0825] 1. 1013125069439411160-02 -7. 820514589548110962e0-02 1. 644919998943805695e-02 1. 333054751157760620e-01 4. 800546169281005859e-02 -
[0826] [0826] 7. 626184076070785522e-02 -4 ,. 128406196832656860e-02 6. 194528192281723022e-03 -3. 991593420505523682e-02 1. 068871654570102692e-02 -
[0827] [0827] -3. 049043193459510803c-02 2. 230124548077583313e-02 -7. 959044724702835083e-02 -4. 829500336199998856e-03 -7. 715594768524169922e-02 -
[0828] [0828] 8. 512626402080059052e0-03 -2. 800592780113220215e-02 -1. 905749924480915070e-02 -7. 994933053851127625e-03 -4. 526712745428085327e-02
[0829] [0829] 6. 359946727752685547e-02 -6. 367459893226623535e-02 1. 439768704585731030e-03 4. 144346714019775391e-01 8. 793116547167301178e-03 9. 585755877196788788e-03 -2. 005S44304847717285e-02 1. 964464038610458374e-01 -
[0830] [0830] -l. 405541040003299713e-02 1. 007734984159469604e-01 3. 295860812067985535e-02 7. 6443068683 14743042e0-02 -7. 301811873912811279e-02 4. 453234001994132996e-02 -8. 700695633888244629e-02 5. 5420439690351486210-02 1. 077778488397598267e-01 -7. 343017309904098511ec-02 7. 176268845796585083e-02 1. 833534426987171173e-02 4. 853336885571479797e-02 2. 010704251006245613e-03 1. 019486486911773682e-01 -7. 219821214675903320e-02 3. 426502272486686707e-02 -
[0831] [0831] 9. 495856612920761108e-02 2. 365492843091487885e-02 -7. 261023670434951782e-02 8. 086591213941574097e-02 -7. 802540063858032227e-02 2. 694368362426757812e0-02 -1. 653958112001419067e-02 -4. 153492301702499390e-02
[0832] [0832] -1. 626765541732311249e-02 5. 274569243192672729e-02 -6. 251925975084304810e-02 1. 551971305161714554e-02 6. 571535766124725342e0-02 -
[0833] [0833] 1. 872961670160293579e-01 -8. 568450063467025757e-02 -1. 074861064553260803e-01 3. 769880831241607666 € -01 1. 131844427436590195e-02 -
[0834] [0834] 1. 023939773440361023e-01 5. 629530921578407288e-02 |. 547605451196432114 € -02 1.831026282161474228e-03 -1.055494323372840881e-014. 693903774023056030e-02 -9. 868312627077102661e-02 -5. 087112635374069214e-02 -
[0835] [0835] 6. 508504599332809448e-02 -3. 988321498036384583e-02 -1. 030667871236801147e-01 1. 517282724380493164e-01 3. 712902963161468506 € -02 -
[0836] [0836] 1. 695588417351245880c-02 -4. 190247133374214172e0-02 -L1. 108506619930267334e-01 1. 388823390007019043e-01 1. 841226220130920410e-02 -
[0837] [0837] 1. 549934875220060349e-02 -8. 315423876047134399e-02 -L1. 577772386372089386 € -02 5. 878822878003120422e-02 1. 486529712565243244 € -03 6. 505486369132995605e-02 1. 799016445875167847e-02 6. 8749763071537017820-02 5. 4414637386798858756. 603618618100881577e-03 5. 875011160969734192e-02 7. 437109202146530151e-02 -1. 723186075687408447e-01 -
[0838] [0838] 7. 467529177665710449e-02 -2. 547362260520458221e-02 1. 847620308399200439e-02 5. 649776011705398560e-02 6. 935506314039230347e-02 1. 696843095123767853e-02 -3. 029418550431728363e-02 8. 984336256980895996e-02 7. 461861521005630493e-02 5. 565826431 848108768e-04 1. 584191620349884033e-01 3. 326614946126937866e-02 -8. 445455878973007202e-02 1. 7808383703231811520-015. 096849054098129272e-02 1. 615845598280429840e-02 6. 1006240546703338620-02 -
[0839] [0839] 2. 703832462430000305e-02 7. 569682318717241287e-03 -5. 901194363832473755e-02 3. 002952039241790771e-02 -5.707353726029396057e-02 1. 899783872067928314e-02 -7.222885638475418091e-02 1. 5458059497 17760086 € -02 4.
[0840] [0840] 8. 303497731685638428e-02 2. 994180470705032349e-02 -7. 316526025533676147e-02 1. 880865842103958130e-01 9. 231487661600112915ec-02 -
[0841] [0841] 4. 192541353404521942e-03 -3. 229407221078872681e-02 -5. 791775882244110107e-02 2. 046165987849235535e-02 -3. 079301118850708008e-02 -
[0842] [0842] -3. 186006098985671997e-02 4. 188629984855651855e-02 |. 966482028365135193e-02 1. 423168182373046875e-01 4. 931379109621047974e-02 -
[0843] [0843] 3. 510799258947372437e-02 -8. 943509310483932495e-02 -2. 943567559123039246 € -02 7. S65789856016635895e-03 -6. 579076498746871948e-02 -
[0844] [0844] -2. 509402669966220856c-02 -7. 2445683181285858150-02 -1. 156093180179595947e-01 8. 540309220552444458e-02 -4. 748164489865303040e-02 4. 687562957406044006 € -02 -9. 249224513769149780e-02 -2. 113502658903598785e-02
[0845] [0845] -4. 066236689686775208e-02 9, 155753254890441895e-02 -1. 601463370025157928e-02 8. 901791274547576904e-02 -5. 799761041998863220e-02 1. 089053526520729065e-01 -1. 068192645907402039e-01 6.357172876596450806e € -02 1. 201343759894371033e-01 2. 644992433488368988e-02 1. 062887683510780334e-01 2. 260592207312583923e-02 -1. 058975905179977417e-01 6. 5697036683559417720-023. 352943342179059982e-03 -7. 212850451469421387e-02 -3. 818150982260704041e-02 -
[0846] [0846] 1. 258508861064910889c-01 -4. 5662205666303634640-02 7. 407991681247949600e-03 1. 224736794829368591e-01 -5. 166685953736305237e-02 -
[0847] [0847] 1. 104597896337509155e-01 7. 2823137044906616210-02 -9. 893217682838439941e-02 -7. 497874554246664047e-03 5. 683084204792976379e-02 -
[0848] [0848] 3. 160682320594787598e-02 -3. 285820782184600830c-02 -1. 179315336048603058e-02 5. 097493529319763184e-02 5. 039264913648366928e-03 -
[0849] [0849] 2. 084301412105560303e-02 -3. O68604692816734314e-02 |. 182319130748510361e-02 5. 421790853142738342e-02 -2. 733012288808822632e-02 -
[0850] [0850] 1. 101736724376678467e-01 4. 4577069580554962160-02 1. 908039674162864685e-03 3. 994913697242736816e-01 -4. 799120500683784485e-02 -
[0851] [0851] -9, 806135669350624084e-03 6. 153792142868041992e-02 |. 361553557217121124 € -02 6. 820578128099441528e-02 -1. 847075484693050385e-02 2. 226619049906730652e-02 -1. 082147806882858276e-01 5. 864118691533803940e-03 6. 402990967035293579e-02 -6. 206588074564933777e-02 -3. 104354487732052803e-03 -
[0852] [0852] 8. 531540632247924805e-02 6. 703392416238784790ec-02 -A4. 450594633817672729e-02 2. 179663777351379395e-01 -3. 072215244174003601e-02 -
[0853] [0853] -4. 826126620173454285e-02 6. 405718624591827393e-02 -6. 425921618938446045e-02 -7. 123930752277374268e-02 -3. 816322609782218933e-02
[0854] [0854] 1. 357984989881515503e-01 -1. 098650395870208740e-01 3. 4257438965141773220-03 3. 13045591 1159515381e-01 -6. 297829002141952515e-02 -
[0855] [0855] -6. 363824754953384399ec-02 -5. O84878951311111450c-02 -6. 851577758789062500e-02 -2. 644426003098487854e-02 -7. 553866505622863770e-02 -
[0856] [0856] -8. 6232172325253486630-03 2. 1474413573741912840-02 5. 256289988756179810e-02 1. 306586265563964844e-01 -4. 361163824796676636e-02 -
[0857] [0857] 6. 731378287076950073e-02 4. 540525749325752258e-02 1. 128863021731376648e-01 9. 305796772241592407e-02 -9. 432055056095123291e-02 -
[0858] [0858] 1. 727583259344100952e-02 -5. 854962766170501709e-02 5. 003688111901283264e-02 1. 575747281312942505e-O01 4. 496030136942863464e-02 -
[0859] [0859] 2. 835433371365070343c-02 -5. 872105062007904053e-02 4. 938599467277526855e-02 5. 584099143743515015e-02 7. 954009622335433960e-02 3. 595757856965065002e-02 -5.22990152239799499e-02 9. 0627387166023254 031617149710655212e-01 -8. 254364132881164551e-02 1. 115744374692440033e-02 -
[0860] [0860] 1. 397663820534944534e-02 4. 662597179412841797e0-02 4. 663140326738357544e-02 -2. 134256996214389801e-02 3. 420338034629821777e-02 -
[0861] [0861] 8. 034001290798187256c-02 2. 335308305919170380ec-02 -8. 568230271339416504e-02 1. 544220596551895142e0-01 2. 945798914879560471e-03 1. 766280154697597027e-03 -1. 337741268798708916e-03 -3. 684893622994422913e-02
[0862] [0862] 9. 750191122293472290c-02 6. 8249344825744628910-02 8. 178575336933135986 € -02 1. 192890182137489319e-01 -2. 802307426463812590e-04 -
[0863] [0863] 1. 516768336296081543e-01 3. 809823840856552124e-02 -9, 779969230294227600e-03 7. 889612764120101929e-02 7. 974613457918167114e-02 1. 775179244577884674e-02 3.6617406-7 278109058737754822e-02 6. 465785205364227295e-02 3. 841586410999298096 € -03 5. 543022602796554565e-02 2. 339226454496383667 € -01 1. 869145222008228302e-02 1.765297353267669678e-02. 453186735510826111c-022. 895061392337083817e-03 -3. 982309252023696899e-02 5. 759987607598304749e-02 1. 769271120429039001e-02 1. 546288728713989258e-01 -1. 841125637292861938e-02 -
[0864] [0864] 1. 795746013522148132e-02 -9. 615703672170639038e-02 7. 352245040237903595e-03 6. 004977598786354065e-02 -6. 1020683497 19047546e € -02 -
[0865] [0865] 6. 242174282670021057e-02 -1. 397683378309011459c-02 3. 690619021654129028e-02 -7. 0653401315212249760-02 4. 6602442860603332520-023. 903285786509513855e-02 3. 2474536448717117310-02 -3. 085780330002307892e-02 -
[0866] [0866] 6. 910125166177749634e-02 7. 630631327629089355e-02 -2. 406744100153446198e-02 -4. 677357897162437439e-02 -3. 498730808496475220e-02 -
[0867] [0867] 1. 166847273707389832e-01 -3. 715934231877326965e-02 4. 789013043045997620e-02 2. 130470424890518188e-01 1. 923215948045253754e-02 5. S68984895944595337e-02 -3. 808736801147460938e-02 1. 005978286266326904e-01 9. 103122353553771973e-02 -9. 865847416222095490e-03 7. 994836568832397461e-02 4. 422549903392791748e-02 -1. 021462604403495789e-01 4. 461938515305519104e-02 6. 906905770301818848e-02 4. 237182065844535828e-02 1. 667861640453338623e-01 -
[0868] [0868] -3. 308371081948280334c-02 -1. 0367181152105331420-02 -1. 535094901919364929e-02 5. 003887787461280823e-02 3. 411285579204559326 € -02 -
[0869] [0869] 9. 240733087062835693e-02 -3. 487472981214523315e-02 -6. 709354370832443237e-02 1.023806780576705933e-01 -4. 461290687322616577e-029. 127608686685562134e-02 -5. 094216763973236084e-02 -1. 3902475126087665560-02
[0870] [0870] -1. 778232306241989136e-02 -2. 454098127782344818e-02 2. 657401142641901970e-03 2. 952568605542182922e0-02 8. 3549454808235168460-02 -
[0871] [0871] 5. 941914394497871399c-02 -1. 0663821361958980560-02 6. 414017081260681152e-02 4. 182731211185455322e-01 8. 235490322113037109e-02 -
[0872] [0872] -8. 413249999284744263e-02 4. 008056595921516418e-02 5. 144314467906951904e-02 6. 8276174366474151610-02 -8. 644512295722961426 € -02 1. 469829399138689041e-02 -6. 635034922510385513e-03 1. 307150628417730331e-023. 910980001091957092e-02 -6. 188061460852622986e-02 3. 867447841912508011e-03 4. T43856843560934067e-03 -8. 206780999898910522e-02 1. 260694265365600586e € -01 -
[0873] [0873] -6. 030004937201738358e-03 -9. 2403918504714965820-02 -1. 857880316674709320e-02 1. 41392961 1444473267e-01 8. 326789736747741699e-02 -
[0874] [0874] 9. 284706413745880127e-02 -5. 670753866434097290c0-02 -8. 413679897785186768e-02 -7. 827807962894439697e-02 5. 781856924295425415e-02 6. 598381698131561279e-02 -2. 714345976710319519e-02 -1. 456629857420921326 € -02 -
[0875] [0875] 5. 987243354320526123e-02 -8. 7489500641822814940-02 -4. 739358648657798767e-02 1. 842073649168014526e-01 -1. 190731022506952286e-02 -
[0876] [0876] 2. 986291237175464630e-02 -2. 687921375036239624e-02 6. 299826502799987793e-02 1. 438053208403289318e-03 -5. 5.00067025423049927e-02
[0877] [0877] 2. 4937152862548828120-02 -5. 5572427809238433840-02 - &. 795661479234695435e-02 -1. 409878255799412727e-03 -5. 3486358374357223510-02
[0878] [0878] 4. 86910715699] 958618e-02 1. 9722018390893936160-02 5. 451372265815734863e-02 1. 037200614809989929e-01 -4. 668588563799858093e-02 -
[0879] [0879] -4. 930094350129365921e-03 -6. 391929835081100464e-02 -6. 575210392475128174e-02 5.727827921509742737e-02 -3. 9893217384815216060-02 2. 443857677280902863e-02 -1. 071977764368057251e-01 3. 364928066730499268e-02 -
[0880] [0880] 1. 094543039798736572e-01 4. 671913385391235352e0-02 -7. 780002057552337646e-02 1. 556625664234161377e-01 2. 435254119336605072e-02 1. 696345955133438110c-02 2. 826086431741714478e-02 2. 708872780203819275e-02 -
[0881] [0881] 1. 686608046293258667e-01 2. S584923803806304932e0-02 5. 612814426422119] 141e-02 7. 497839629650115967e-02 -4. 946025088429450989e-02 -
[0882] [0882] 1. 017296686768531799e-01 -4. 583080857992172241e-02 6. 557719409465789795e-02 4. 816550388932228088e-02 3. 837331756949424744e-02 -
[0883] [0883] -2. 660217508673667908e-02 -1. 086283326148986816e-01 -1. 679304987192153931e-02 6. 821155548095703125e-02 3. 872316330671310425e-02 -
[0884] [0884] -5. 855180695652961731e-02 -7. 080493867397308350e-02 -7. 643019407987594604e-02 5. 622763186693191528e-02 3. 6194462329149246220-02 -
[0885] [0885] -9. 212182462215423584c0-02 -1. 070652604103088379e-01 4. 055893793702125549ec-02 -6. 756686419248580933e-02 2. 866046875715255737e-02 -
[0886] [0886] 1. 199186593294143677e-01 -4. 516707267612218857e-03 -2. 243466302752494812e-02 1. 319016069173812866 € -01 -4. 294145852327346802e-02 -
[0887] [0887] -1. 059765219688415527e-01 4. 222847521305084229c-02 -3. 808278590440750122e-03 -1. 830457337200641632e-02 -6. 451085954904556274e-02
[0888] [0888] 1. 057598069310188293e-01 7. 204839587211608887e-02 7. 993333041667938232e-02 7. 262804359197616577e-02 9. 874258190393447876e-02 -
[0889] [0889] -4, 427675157785415649e-02 -5. 723756179213523865e-02 6. 398452445864677429e-03 1. 135059073567390442e-01 -8. 347614109516143799e-02 2. 417769655585289001e-02 4. 380713775753974915e-02 1. 274052262306213379e-02 5. 734059587121009827e-02 -1. 024644076824188232e-01 -7.239122688770294189ec-02 -
[0890] [0890] -3. 665628284215927124e-02 -6. 260617077350616455e-02 -3. 417005762457847595e-02 -2. 461953647434711456 € -02 -9. 000421315431594849e-02
[0891] [0891] 3. 692010790109634399e-02 4. 770493134856224060e-02 -1. 778219453990459442e-02 4. 632871598005294800e-02 2. 143774740397930145e-02 -
[0892] [0892] 3. 279715403914451599ec-02 -1. 204383000731468201e-02 1. 721214735880494118e-03 8. 710324019193649292e-02 9. 941030293703079224e-02 -
[0893] [0893] 1. 042182147502899170e-01 3. 519155085086822510e-02 -7. 921601831912994385e-02 -3. 339725360274314880e-02 4. 629015550017356873e-02 -
[0894] [0894] W ": -1. 415030807256698608e-01 -9. 5422111451625823970-02
[0895] [0895] FIG. 14 illustrates an example computer 1400 for implementing the entities shown in FIGS. 1 and 3. The computer 1400 includes at least one processor 1402 coupled to a 1404 chipset. The 1404 chipset includes a 1420 memory controller hub and an 1422 input / output (I / O) controller hub. A 1406 memory and an adapter graphics 1412 are attached to the memory controller hub 1420 and a monitor 1418 is attached to the graphics adapter 1412. A storage device 1408, an input device 1414 and network adapter 1416 are attached to the hub of the I / O controller 1422. Others Computer 1400 modes have different architectures.
[0896] [0896] Storage device 1408 is a non-transitory, computer-readable storage medium, such as a hard disk, read-only compact disc (CD-ROM), DVD, or a solid-state memory device. Memory 1406 contains instructions and data used by processor 1402. Input interface 1414 is a touch screen interface, a mouse, track ball or other type of pointing device, a keyboard or some combination thereof, and is used to insert data on computer 1400. In some embodiments, computer 1400 can be configured to receive input (for example, commands) from the input interface 1414 through user gestures. The graphics adapter 1412 displays images and other information on the display 1418. The network adapter 1416 couples the computer 1400 to one or more computer networks.
[0897] [0897] The 1400 computer is adapted to run computer program modules to provide the functionality described here. As used in this document, the term "module" refers to the logic of the computer program used to provide the specified functionality. Thus, a module can be implemented in hardware, firmware and / or software. In one embodiment, the program modules are stored in storage device 1408, loaded into memory 1406 and executed by processor 1402.
[0898] [0898] The types of computers 1400 used by the entities of FIG. 1 may vary depending on the modality and the processing power required by the entity. For example, presentation ID system 160 can run on a single computer 1400 or on multiple computers 1400 communicating with each other over a network, such as a server farm. 1400 computers may not have some of the components described above, such as 1412 graphics adapters and displays
1418.
References
1. Desrichard, A., Snyder, A. & Chan, T. A. Cancer Neoantigens and Applications for Immunotherapy. Clin. Cancer Res. Off. JJ. Am. Assoc. Cancer Res. (2015). doi: 10. 1158 / 1078-0432. CCR-14-3175
2. Schumacher, T. N. & Schreiber, R. D. Neoantigens in cancer immunotherapy. Science 348, 69-74 (2015).
3. Gubin, M. M., Artyomov, M. N., Mardis, E. R. & Schreiber, R. D. Tumor neoantigens: building a framework for personalized cancer immunotherapy. J. Clin. Invest. 125, 3413-3421 (2015).
4. Rizvi, N. A. etal. Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 348, 124—128 (2015).
5. Snyder, A. et al. Genetic basis for clinical response to CTLA-4 blockade in melanoma. N. Engl. J. Med. 371, 2189-2199 (2014).
6. Carreno, B. M. etal. Cancer immunotherapy. A dendritic cell vaccine increases the breadth and diversity of neoantigen-specific melanoma T cells. Science 348, 803—-808 (2015).
7. Tran, E. etal. Cancer immunotherapy based on mutation-specific CD4 + T cells in a patient with epithelial cancer. Science 344, 641645 (2014).
8. Hacohen, N. & Wu, C. J.-Y. United States Patent Application: 0110293637 -
COMPOSITIONS AND METHODS OF IDENTIFYING TUMOR SPECIFIC NEOANTIGENS. (Al). at & lthttp: // appftli uspto - gov / netacgi / nph- Parser Sect1 = PTO1 & Sect2 = HITOFF & d = PG01 & p = 1 & u = / netahtml / PTO / srchnum. html & r = 1 & fFG & = 50 & s1 = 20110293637. PGNR. &gt;
9. Lundegaard, C., Hoof, L., Lund, O. & Nielsen, M. State of the art and challenges in sequence based T-cell epitope prediction. Immunome Res. 6 Suppl 2, S3 (2010).
10. Yadav, M. etal. Predicting immunogenic tumor mutations by combining mass spectrometry and exome sequencing. Nature 515, 572—576 (2014).
11. Bassani-Sternberg, M., Pletscher-Frankild, S., Jensen, L. J. & Mann, M. Mass spectrometry of human leukocyte antigen class I peptidomes reveals strong effects of protein abundance and turnover on antigen presentation. Mol. Cell. Proteomics MCP 14, 658673 (2015).
12. Van Allen, E. M. etal. Genomic correlates of response to CTLA-4 blockade in metastatic melanoma. Science 350, 207-211 (2015).
13. Yoshida, K. & Ogawa, S. Splicing factor mutations and cancer. Wiley Interdiscip. Rev. RNA 5, 445459 (2014).
14. Cancer Genome Atlas Research Network. Comprehensive molecular profiling of lung adenocarcinoma. Nature 511, 543-550 (2014).
15. Rajasagi, M. etal Systematic identification of personal tumor-specific neoantigens in chronic lymphocytic leukemia. Blood 124, 453—462 (2014).
16. Downing, S. R. etal United States Patent Application: 0120208706 - OPTIMIZATION OF MULTIGENE ANALYSIS OF TUMOR SAMPLES. (Al). at & http: // appftl. uspto. gov / netacgi / nph- Parser Sect1l = PTO1 & Sect2 = HITOFF & d = PG01 & p = 1 & u = / netahtml / PTO / srchnum. html & r = 1 & FG & F = 50 & s1 = 20120208706. PGNR. &gt;
17. Target Capture for NextGen Sequencing - IDT. at &lt; http: // iwww. idtdna. com / pages / products / nextgen / target-capture &gt;
18. Shukla, S. A. etal. Comprehensive analysis of cancer-associated somatic mutations in class I HLA genes. Nat. Biotechnol. 33, 1152-1158 (2015).
19. Cieslik, M. etal. The use of exome capture RNA-seq for highly degraded RNA with application to clinical cancer sequencing. Genome Res. 25, 1372—1381 (2015).
20. Bodini, M. etal. The hidden genomic landscape of acute myeloid leukemia: subclonal structure revealed by undetected mutations. Blood 125, 600-605 (2015).
21. Saunders, C. T. etal. Strelka: accurate somatic small-variant calling from sequenced tumor-normal sample pairs. Bioinforma. Oxf. Engl. 28, 1811-1817 (2012).
22. Cibulskis, K. etal. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat. Biotechnol. 31, 213—219 (2013).
23. Wilkerson, M. D. etal. Integrated RNA and DNA sequencing improves mutation detection in low purity tumors. Nucleic Acids Res. 42, e107 (2014).
24. Mose, L. E., Wilkerson, M. D., Hayes, D. N., Perou, C. M. & Parker, J. S. ABRA: improved coding indel detection via assembly-based realignment. Bioinforma. Oxf. Engl. 30, 2813-2815 (2014).
25. Ye, K., Schulz, M. H., Long, Q., Apweiler, R. & Ning, Z. Pindel: a pattern growth approach to detect break points of large deletions and medium sized insertions from paired-end short reads. Bioinforma. Oxf. Engl. 25, 2865-2871 (2009).
26. Lam, H. Y. K. etal. Nucleotide-resolution analysis of structural variants using BreakSegq and a breakpoint library. Nat. Biotechnol. 28, 47-55 (2010).
27. Frampton, G. M. etal. Development and validation of a clinical cancer genomic profiling test based on massively parallel DNA sequencing. Nat. Biotechnol. 31, 1023-1031 (2013).
28. Boegel, S. etal. HLA typing from RNA-Seq sequence reads. Genome Med. 4, 102 (2012).
29. Liu, C. etal. ATHLATES: accurate typing of human leukocyte antigen through exome sequencing. Nucleic Acids Res. 41, e142 (2013).
30. Mayor, N.P. etal. HLA Typing for the Next Generation. PloS One 10, and € 0127153 (2015).
31. Roy, C. K., Olson, S., Graveley, B. R., Zamore, P. D. & Moore, M. J. Assessing long-distance RNA sequence connectivity via RNA-templated DNA-DNA ligation. eLife 4, (2015).
32. Song, L. & Florea, L. CLASS: constrained transcript assembly of RNA-seq reads. BMC Bioinformatics 14 Suppl 5, S14 (2013).
33. Maretty, L., Sibbesen, J. A. & Krogh, A. Bayesian transcriptome assembly. Genome Biol. 15, 501 (2014).
34. Pertea, M. etal. StringTie enables improved reconstruction of a transcriptome from RNA-segq reads. Nat. Biotechnol. 33, 290-295 (2015).
35. Roberts, A., Pimentel, H., Trapnell, C. & Pachter, L. Identification of novel transcripts in annotated genomes using RNA-Seq. Bioinforma. Oxf. Engl. (2011). doi: 10. 1093 / bioinformatics / btr355
36. Vitting-Seerup, K .., Porse, B. T., Sandelin, A. & Waage, J. spliceR: an R package for classification of alternative splicing and prediction of coding potential from RNA-seq data. BMC Bioinformatics 15, 81 (2014).
37. Rivas, M. A. etal. Human genomics. Effect of predicted protein-truncating genetic variants on the human transcriptome. Science 348, 666669 (2015).
38. Skelly, D. A., Johansson, M., Madeoy, J., Wakefield, J. & Akey, J. M. A powerful and flexible statistical framework for testing hypotheses of allele-specific gene expression from RNA-seq data. Genome Res. 21, 1728-1737 (2011).
39. Anders, S., Pyl, P. T. & Huber, W. HTSeg - a Python framework to work with high-throughput sequencing data. Bioinforma. Oxf. Engl. 31, 166-169 (2015).
40. Furney, S. J. etal. SF3B1 mutations are associated with alternative splicing in uveal melanoma. Cancer Discov. (2013). doi: 10. 1158 / 2159-8290. CD-13-0330
41. Zhou, Q. etal. A chemical genetics approach for the functional assessment of novel cancer genes. Cancer Res. (2015). doi: 10. 1158 / 0008-5472. CAN-14-2930
42. Maguire, S. L. etal. SF3B1 mutations constitute a novel therapeutic target in breast cancer. J. Pathol. 235, 571-580 (2015).
43. Carithers, L. J. etal. A Novel Approach to High-Quality Postmortem Tissue Procurement: The GTEx Project. Biopreservation Biobanking 13, 311—-319 (2015).
44. Xu, G. etal. RNA CoMPASS: a dual approach for pathogen and host transcriptome analysis of RNA-seq datasets. PloS One 9, e89445 (2014).
45. Andreatta, M. & Nielsen, M. Gapped sequence alignment using artificial neural networks: application to the MHC class I system. Bioinforma. Oxf. Engl. (2015). doi: 10. 1093 / bioinformatics / btv639
46. Jorgensen, K. W., Rasmussen, M., Buus, S. & Nielsen, M. NetMHCstab - predicting stability of peptide-MHC-I complexes; impacts for cytotoxic T lymphocyte epitope discovery. Immunology 141, 18-26 (2014).
47. Larsen, M. V. etal. An integrative approach to CTL epitope prediction: a combined algorithm integrating MHC class I binding, TAP transport efficiency, and proteasomal cleavage predictions. Eur. J. Immunol. 35, 2295-2303 (2005).
48. Nielsen, M., Lundegaard, C., Lund, O. & Kesmir, C. The role of the proteasome in generating cytotoxic T-cell epitopes: insights obtained from improved predictions of proteasomal cleavage. Immunogenetics 57, 3341 (2005).
49, Boisvert, F.-M. etal. A Quantitative Spatial Proteomics Analysis of Proteome Turnover in Human Cells. Mol. Cell. Proteomics 11, M111. 011429-M111. 011429 (2012).
50. Duan, F. et al. Genomic and bioinformatic profiling of mutational neoepitopes reveals new rules to predict anticancer immunogenicity. J. Exp. Med. 211, 2231-2248 (2014).
51. Janeway's Immunobiology: 9780815345312: Medicine & Health Science Books (2 Amazon.com at &lt; http: // www. Amazon.com / Janeways-Immunobiology-Kenneth- Murphy / dp / 0815345313 &gt;
52. Calis, J. J. A. etal. Properties of MHC Class I Presented Peptides That Enhance Immunogenicity. PLoS Comput. Biol. 9, e1003266 (2013).
53. Zhang, J. et al. Intratumor heterogeneity in localized lung adenocarcinomas delineated by multiregion sequencing. Science 346, 256259 (2014)
54. Walter, M. J. etal. Clonal architecture of secondary acute myeloid leukemia. N. Engl. J. Med. 366, 1090-1098 (2012).
55. Hunt DF, Henderson RA, Shabanowitz J, Sakaguchi K, Michel H, Sevilir N, Cox AL, Appella E, Engelhard VH. Characterization of peptides bound to the class T MHC molecule HLA-A2. 1 by mass spectrometry. Science 1992. 255: 1261-1263.
56. Zarling AL, Polefrone JM, Evans AM, Mikesh LM, Shabanowitz J, Lewis ST, Engelhard VH, Hunt DF. Identification of class I MHC-associated phosphopeptides as targets for cancer immunotherapy. Proc Natl Acad Sci US A. 2006 Oct 3; 103 (40): 14889-
94.
57. Bassani-Sternberg M, Pletscher-Frankild S, Jensen LJ, Mann M. Mass spectrometry of human leukocyte antigen class I peptidomes reveals strong effects of protein abundance and turnover on antigen presentation. Mol Cell Proteomics. 2015 Mar; 14 (3): 658-
73. doi: 10. 1074 / mcp. M114. 042812.
58. Abelin JG, Trantham PD, Penny SA, Patterson AM, Ward ST, Hildebrand WH, Cobbold M, Bai DL, Shabanowitz J, Hunt DF. Complementary IMAC enrichment methods for HLA-associated phosphopeptide identification by mass spectrometry. Nat Protoc. 2015 Sep; 10 (9): 1308-18. doi: 10. 1038 / nprot. 2015. 086. Epub 2015 Aug 6
59. Barnstable CJ, Bodmer WF, Brown G, Galfre G, Milstein C, Williams AF, Ziegler A. Production of monoclonal antibodies to group A erythrocytes, HLA and other human cell surface antigens-new tools for genetic analysis. Cell. 1978 May; 14 (1): 9-20.
60. Goldman JM, Hibbin J, Kearney L, Orchard K, Th'ng KH. HLA-DR monoclonal antibodies inhibit the proliferation of normal and chronic granulocytic leukaemia myeloid progenitor cells. Br J Haematol. 1982 Nov; 52 (3): 411-20.
61. Eng JK, Jahan TA, Hoopmann MR. Comet: an open-source MS / MS sequence database search tool. Proteomics. 2013 Jan; 13 (1): 22-4. doi: 10. 1002 / pmic. 201200439. Epub 2012 Dec 4.
62. Eng JK, Hoopmann MR, Jahan TA, Egertson JD, Noble WS, MacCoss MJ. À deeper look into Comet - implementation and features. J Am Soc Mass Spectrom. 2015 Nov; 26 (11): 1865-74. doi: 10. 1007 / s13361-015-1179-x. Epub 2015 Jun 27.
63. Lukas Kill, Jesse Canterbury, Jason Weston, William Stafford Noble and Michael J. MacCoss. Semi-supervised learning for peptide identification from shotgun proteomics datasets. Nature Methods 4: 923 - 925, November 2007
64. Lukas Kãll, John D. Storey, Michael J. MacCoss and William Stafford Noble. Assigning confidence measures to peptides identified by tandem mass spectrometry. Journal of Proteome Research, 7 (1): 29-34, January 2008
65. Lukas Kãll, John D. Storey and William Stafford Noble. Nonparametric estimation of posterior error probabilities associated with peptides identified by tandem mass spectrometry. Bioinformatics, 24 (16): 142-148, August 2008
66. Bo Li and C. olin N. Dewey. RSEM: accurate transcript quantification from RNA-Segq data with or without a referenfe genome. BMC Bioinformatics, 12: 323, August 2011
67. Hillary Pearson, Tariq Daouda, Diana Paola Granados, Chantal Durette, Eric Bonneil, Mathieu Courcelles, Anja Rodenbrock, Jean-Philippe Laverdure, Caroline Côté, Sylvie Mader, Sébastien Lemieux, Pierre Thibault, and Claude Perreault. MHC class I- associated peptides derive from selective regions of the human genome. The Journal of Clinical Investigation, 2016,
68. Juliane Liepe, Fabio Marino, John Sidney, Anita Jeko, Daniel E. Bunting, Alessandro Sette, Peter M. Kloetzel, Michael P. H. Stumpf, Albert J. R. Heck, Michele Mishto. A large fraction of HLA class I ligands are proteasome-generated spliced peptides. Science, 21, October 2016.
69. Mommen GP., Marino, F., Meiring HD., Poelen, MC., Van Gaans-van den Brink, JA., Mohammed S., Heck AJ., And van Els CA. Sampling From the Proteome to the Human Leukocyte Antigen-DR (HLA-DR) Ligandome Proceeds Via High Specificity. Mol Cell Proteomics 15 (4): 1412-1423, April 2016.
70. Sebastian Kreiter, Mathias Vormehr, Niels van de Roemer, Mustafa Diken, Martin Lôwer, Jan Diekmann, Sebastian Boegel, Barbara Schrôrs, Fulvia Vascotto, John C. Castle, Arbel D. Tadmor, Stephen P. Schoenberger, Christoph Huber, Ôzlem Túreci, and Ugur Sahin. Mutant MHC class II epitopes drive therapeutic immune responses to caner. Nature 520, 692-696, April 2015. 71. Tran E., Turcotte S., Gros A., Robbins P. F, Lu YC ,, Dudley ME, Wunderlich JR, Somerville RP, Hogan K .., Hinrichs CS, Parkhurst MR, Yang JC, Rosenberg SA Cancer immunotherapy based on mutation-specific CD4 + T cells in a patient with epithelial cancer. Science 344 (6184) 641-645, May 2014. 72. Andreatta M., Karosiene E., Rasmussen M., Stryhn A., Buus S., Nielsen M. Accurate pan-specific prediction of peptide-MHC class II binding affinity with improved binding core identification. Immunogenetics 67 (11-12) 641-650, November 2015.
73. Nielsen, M., Lund, O. NN-align. An artificial neural network-based alignment algorithm for MHC class II peptide binding predictionn BMC Bioinformatics 10: 296, September 2009.
74. Nielsen, M., Lundegaard, C., Lund, O. Prediction of MHC class II binding affinity using SMM-align, a novel stabilization matrix alignment method. BMC Bioinformatics 8: 238, July 2007.
75. Zhang, J., etal. PEAKS DB: again sequencing assisted database search for sensitive and accurate peptide identification. Molecular & Cellular Proteomics. 11 (4): 1-8. 1/2/2012.
76. Jensen, Kamilla Kjaergaard, et al. “Improved Methods for Prediting Peptide Binding Affinity to MHC Class II Molecules. “Immunology, 2018, doi: 10. 1111 / imm.
12889.
77. Carter, SL, Cibulskis, K., Helman, E., McKenna, A., Shen, H., Zack, T., Laird, PW, Onofrio, RC, Winckler, W., Weir, BA, et al . (2012). Absolute quantification of somatic DNA alterations in human cancer. Nat. Biotechnol. 30, 413421
78. McGranahan, N., Rosenthal, R., Hiley, CT, Rowan, AJ, Watkins, TBK, Wilson, GA, Birkbak, NJ, Veeriah, S., Van Loo, P., Herrero, J., et al . (2017). Allele- Specific HLA Loss and Immune Escape in Lung Cancer Evolution. Cell 171, 1259-1271. ell.
79. Shukla, S. A., Rooney, M. S., Rajasagi, M., Tiao, G., Dixon, P. M., Lawrence, M.S ,, Stevens, J., Lane, W. J., Dellagatta, J. L., Steelman, S., et al. (2015). Comprehensive analysis of cancer-associated somatic mutations in class I HLA genes. Nat. Biotechnol. 33, 1152-1158.
80. Van Loo, P., Nordgard, SH, Lingjerde, OC, Russnes, HG, Rye, IH, Sun, W., Weigman, VJ, Marynen, P., Zetterberg, A., Naume, B., et al . (2010). Allele-specific copy number analysis of tumors. Proc. Natl. Acad. Sci. U.S.A. 107, 1691016915.
81. Van Loo, P., Nordgard, SH, Lingjerde, OC, Russnes, HG, Rye, IH, Sun, W., Weigman, VJ, Marynen, P., Zetterberg, A., Naume, B., et al . (2010). Allele-specific copy number analysis of tumors. Proc. Natl. Acad. Sci. U.S.A. 107, 16910-16915.

权利要求:
Claims (33)
[1]
1. Method for generating an output for the construction of a personalized cancer vaccine, identifying one or more neoantigens from one or more tumor cells of a subject that are likely to be presented on a tumor cell surface, characterized by the fact that it comprises the steps of: obtaining at least one of the exome, transcriptome, or entire genome nucleotide sequencing data from the subject's tumor cells and normal cells, where the nucleotide sequencing data is used to obtain data representing peptide sequences from each one of a set of neoantigens identified by comparing the nucleotide sequencing data of tumor cells and the nucleotide sequencing data of normal cells and in which the peptide sequence of each neoantigen comprises at least one alteration that distinguishes it from the corresponding wild type peptide sequence identified from the subject's normal cells O; encode the peptide sequences of each of the neoantigens in a corresponding numeric vector, each numeric vector including information about a plurality of amino acids that make up the peptide sequence and a set of positions of the amino acids in the peptide sequence; introduce the numerical vectors, using a computer processor, into a profound learning presentation model to generate a set of presentation probabilities for the set of neoantigens, each presentation probability in the set representing the probability that a corresponding neoantigen will be presented by a or more class II MHC alleles on the surface of the subject's tumor cells, the deep learning presentation model comprising: a plurality of parameters identified at least based on a training data set comprising: markers obtained by mass spectrometry that measure the presence of peptides linked to at least one MHC class II allele identified as present in at least one of a plurality of samples; train peptide sequences encoded as numeric vectors, including information about a plurality of amino acids that make up the peptide sequence and a set of amino acid positions in the peptide sequence; and at least one HLA allele associated with the training peptide sequences; and a function that represents a relationship between the numeric vector received as an input and the presentation probability generated as an output based on the numeric vector and the parameters, select a subset of the set of neoantigens based on the set of presentation probabilities to generate a set selected neoantigens; and generate the way for the construction of a personalized cancer vaccine based on the set of selected neoantigens.
[2]
2. Method according to claim 1, characterized by the fact that encoding the peptide sequence comprises encoding the peptide sequence using a one-hot coding scheme.
[3]
3. Method according to any one of claims 1 to 2, characterized by the fact that the entry of the numerical vector in the deep learning presentation model comprises: applying the deep learning presentation model to the peptide sequence of the neoantigen to generate a dependency score for each of the MHC class II alleles, indicating whether the MHC class II allele will present the neoantigen based on specific amino acids at the particular positions of the peptide sequence.
[4]
4. Method, according to claim 3, characterized by the fact that the entry of the numerical vector in the deep learning presentation model also comprises:
transform the dependency scores to generate a probability by corresponding allele for each class II MHC allele, indicating a probability that the corresponding class II MHC allele has the corresponding neoantigen; and combine the probabilities by allele to generate the probability of presenting the neoantigen.
[5]
5. Method, according to claim 4, characterized by the fact that the dependence transformation score models the presentation of the neoantigen as mutually exclusive between one or more MHC class II alleles.
[6]
6. Method, according to claim 3, characterized by the fact that the entry of the numerical vector in the profound learning presentation model further comprises: transforming a combination of the dependency scores to generate the presentation probability, in which to transform the combination dependency scores models the presentation of the neoantigen as interfering between one or more MHC class II alleles.
[7]
7. Method, according to claim 3, characterized by the fact that the set of presentation probabilities is still identified by at least one or more non-interactive allele characteristics and further comprising: applying the presentation model to the non-interactive characteristics of allele to generate a dependency score for non-interactive allele characteristics, indicating whether the peptide sequence of the corresponding neoantigen will be presented based on the non-interactive allele characteristics.
[8]
8. Method according to claim 7, characterized by the fact that it further comprises: combining the dependency score for each class II MHC allele into one or more class II MHC alleles with the dependency score for the non-characteristic allele interactive; and transforming the combined dependency scores for each class II MHC allele to generate a probability per allele for each class II MHC allele, indicating a probability that the corresponding class II MHC allele will present the corresponding neoantigen; and combine the odds by allele to generate the probability of presentation.
[9]
9. Method according to claim 8, characterized by the fact that it further comprises: transforming a combination of the dependency scores for each of the MHC class II alleles and the dependency score for the non-interactive allele characteristics to generate the probability of presentation.
[10]
10. Method according to any one of the claims | to 9, characterized by the fact that the one or more MHC class II alleles include two or more MHC class II alleles.
[11]
11. Method according to any of the claims | to 10, characterized by the fact that at least one class II MHC allele includes two or more different types of class II MHC alleles.
[12]
12. Method according to any one of the claims | to 11, characterized by the fact that the plurality of samples comprises at least one of: (a) one or more cell lines engineered to express a single class II MHC allele; (b) one or more cell lines engineered to express a plurality of MHC class II alleles; (c) one or more human cell lines obtained or derived from a plurality of patients; (d) fresh or frozen tumor samples obtained from a plurality of patients; and (e) fresh or frozen tissue samples obtained from a plurality of patients.
[13]
13. Method according to any one of claims 1 to 12, characterized by the fact that the training data set further comprises at least one of: (a) data associated with MHC peptide-binding affinity measurements for at least least one of the isolated peptides; and (b) data associated with measurements of MHC-peptide binding stability for at least one of the isolated peptides.
[14]
14. Method according to any one of the claims | to 13, characterized by the fact that the set of presentation probabilities is still identified by at least levels of expression of one or more MHC class II alleles in the subject, as measured by RNA-seq or mass spectrometry.
[15]
15. Method according to any one of claims 1 to 14, characterized by the fact that the set of presentation probabilities is further identified by at least interactive allele characteristics, comprising at least one of: (a) predicted affinity between a neoantigen in the set of neoantigens and one or more MHC alleles; and (b) predicted stability of the neoantigen-encoded peptide-MHC complex.
[16]
16. Method according to any one of claims 1 to 15, characterized by the fact that the set of numerical probabilities is further identified by at least non-interactive characteristics of the MHC allele comprising at least one of: (a) The C sequences -terminals that flank the peptide encoded by neoantigen within its original protein sequence; and (b) The N-terminal sequences that flank the peptide encoded by neoantigen within its original protein sequence.
[17]
17. Method according to any one of claims 1 to 16, characterized by the fact that the selection of the set of selected neoantigens comprises selecting neoantigens that have an increased probability of being presented on the surface of the tumor cell in relation to the neoantigens not selected with based on the presentation template.
[18]
18. Method according to any of the claims | to 17, characterized by the fact that the selection of the set of selected neoantigens comprises selecting neoantigens that have an increased probability of being able to induce a tumor specific immune response in the subject in relation to neoantigens not selected based on the presentation model.
[19]
19. Method according to any one of claims 1 to 18, characterized by the fact that the selection of the set of selected neoantigens comprises selecting neoantigens that have an increased probability of being able to be presented to naive T cells by antigen presenting cells professionals (APCs) in relation to unselected neoantigens based on the presentation model, optionally, in which the APC is a dendritic cell (DC).
[20]
20. Method according to any one of the claims | to 19, characterized by the fact that the selection of the set of selected neoantigens comprises selecting neoantigens that have a reduced probability of being subject to inhibition via central or peripheral tolerance in relation to non-selected neoantigens based on the presentation model.
[21]
21. Method according to any one of claims 1 to 20, characterized by the fact that the selection of the set of selected neoantigens comprises selecting neoantigens that have a reduced probability of being able to induce an autoimmune response to normal tissue in the subject in relation to to neoantigens not selected based on the presentation model.
[22]
22. Method according to any of the claims | to 21, characterized by the fact that one or more tumor cells are selected from the group consisting of: lung cancer, melanoma, breast cancer, ovarian cancer, prostate cancer, kidney cancer, gastric cancer, colon cancer, testicular cancer, head and neck cancer, pancreatic cancer, brain cancer, B-cell lymphoma, acute myeloid leukemia, chronic myeloid leukemia, chronic lymphocytic leukemia and T-cell lymphocytic leukemia, non-small cell lung cancer and cancer small cell lung cancer.
[23]
23. Method for treating a subject with a tumor, characterized by the fact that it comprises performing the steps of any one of the claims | to 22, and further comprising obtaining a tumor vaccine comprising the set of selected neoantigens and administering the tumor vaccine to the subject.
[24]
24. A method of making a tumor vaccine, comprising carrying out the steps of any one of claims 1 to 22, and further comprising producing or having produced a tumor vaccine comprising the set of selected neoantigens.
[25]
25. Method according to any of the claims | to 24, characterized by the fact that it also comprises identifying one or more T cells that are specific to the antigen for at least one of the neoantigens in the subset.
[26]
26. Method according to claim 25, characterized in that the identification comprises co-culturing one or more T cells with one or more neoantigens in the subset under conditions that expand the one or more antigen-specific T cells.
[27]
27. Method according to claim 25, characterized by the fact that the identification comprises contacting one or more T cells with a tetramer comprising one or more of the neoantigens in the subset under conditions that allow the connection between the T cell and the tetramer .
[28]
28. Method according to any one of claims 25 to 27, characterized in that it further comprises identifying one or more T cell receptors (TCR) from one or more identified T cells.
[29]
29. The method of claim 28, characterized in that the identification of one or more T cell receptors comprises sequencing the T cell receptor sequences of one or more identified T cells.
[30]
30. Isolated T cell that is antigen specific, characterized by the fact that it is for at least one neoantigen selected in the subset as defined in any of claims 1-28.
[31]
31. Method according to any one of claims 28 to 29, characterized by the fact that it further comprises:
genetic engineering of a plurality of T cells to express at least one of the one or more T cell receptors identified; culturing the plurality of T cells under conditions that expand the plurality of T cells; and infuse the expanded T cells into the subject.
[32]
32. The method of claim 31, characterized in that the genetic engineering of the plurality of T cells to express at least one or more of the identified T cell receptors comprises: cloning the T cell receptor sequences from a or more T cells identified in an expression vector; and transfecting each of the plurality of T cells with the expression vector.
[33]
33. Method according to any one of claims 25 to 29 and 31 to 32, characterized by the fact that it further comprises: culturing the one or more identified T cells under conditions that expand the one or more identified T cells; and infuse the expanded T cells into the subject.

类似技术:

---

公开号 | 公开日 | 专利标题

---

US11183286B2|2021-11-23|Neoantigen identification, manufacture, and use

---

BR112019021782A2|2020-08-18|identification, manufacture and use of neoantigens

---

US11264117B2|2022-03-01|Neoantigen identification using hotspots

---

US20210011026A1|2021-01-14|Reducing junction epitope presentation for neoantigens

---

US20200363414A1|2020-11-19|Neoantigen Identification for T-Cell Therapy

---

US20200105377A1|2020-04-02|Neoantigen identification, manufacture, and use

同族专利:

---

公开号 | 公开日

---

IL269855D0|2019-11-28|

---

WO2018195357A1|2018-10-25|

---

KR20190140935A|2019-12-20|

---

CA3060569A1|2018-10-25|

---

US20210113673A1|2021-04-22|

---

RU2019136762A|2021-05-19|

---

CO2019012345A2|2020-01-17|

---

CN110636852A|2019-12-31|

---

EP3612965A1|2020-02-26|

---

JP2020519246A|2020-07-02|

---

SG11201909652WA|2019-11-28|

---

AU2018254526A1|2019-11-14|

---

MX2019012433A|2019-12-11|

---

EP3612965A4|2021-01-13|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

KR20210156320A|2013-04-07|2021-12-24|더 브로드 인스티튜트, 인코퍼레이티드|Compositions and methods for personalized neoplasia vaccines|

---

US20170199961A1|2015-12-16|2017-07-13|Gritstone Oncology, Inc.|Neoantigen Identification, Manufacture, and Use|WO2014180490A1|2013-05-10|2014-11-13|Biontech Ag|Predicting immunogenicity of t cell epitopes|

---

WO2016128060A1|2015-02-12|2016-08-18|Biontech Ag|Predicting t cell epitopes useful for vaccination|

---

US20170199961A1|2015-12-16|2017-07-13|Gritstone Oncology, Inc.|Neoantigen Identification, Manufacture, and Use|

---

WO2019075112A1|2017-10-10|2019-04-18|Gritstone Oncology, Inc.|Neoantigen identification using hotspots|

---

US20210181188A1|2018-08-24|2021-06-17|The Regents Of The University Of California|Mhc-ii genotype restricts the oncogenic mutational landscape|

---

EP3906045A1|2019-01-03|2021-11-10|Evaxion Biotech ApS|Vaccines targeting neoepitopes|

---

CN111621564A|2019-02-28|2020-09-04|武汉大学|Method for identifying effective tumor neoantigen|

---

AU2020232844A1|2019-03-06|2021-10-28|Gritstone Bio, Inc.|Identification of neoantigens with MHC class II model|

---

CN113905756A|2019-03-11|2022-01-07|伊沃逊生物科技股份公司|Nucleic acid vaccination using constructs encoding neoepitopes|

---

WO2020185010A1|2019-03-12|2020-09-17|신테카바이오|System and method for providing neoantigen immunotherapy information by using artificial-intelligence-model-based molecular dynamics big data|

---

WO2021048400A1|2019-09-13|2021-03-18|Evaxion Biotech Aps|Method for identifying t-cell epitopes|

---

WO2021123232A1|2019-12-18|2021-06-24|Evaxion Biotech Aps|Nucleic acid vaccination using neo-epitope encoding constructs|

---

KR102278586B1|2020-01-07|2021-07-16|한국과학기술원|Method and System for Screening Neoantigens, and Use thereof|

---

WO2021204911A1|2020-04-07|2021-10-14|Evaxion Biotech A/S|Neoepitope immunotherapy with apc targeting unit|

---

WO2021257879A1|2020-06-18|2021-12-23|Personalis Inc.|Machine-learning techniques for predicting surface-presenting peptides|

---

WO2022013277A1|2020-07-14|2022-01-20|Evaxion Biotech A/S|Apc targeting units for immunotherapy|

法律状态:
2021-08-03| B11A| Dismissal acc. art.33 of ipl - examination not requested within 36 months of filing|

---

2021-10-19| B11Y| Definitive dismissal - extension of time limit for request of examination expired [chapter 11.1.1 patent gazette]|

---

2021-11-03| B350| Update of information on the portal [chapter 15.35 patent gazette]|

优先权:

---

申请号 | 申请日 | 专利标题

---

US201762487469P| true| 2017-04-19|2017-04-19|

---

US62/487,469|2017-04-19|

---

PCT/US2018/028438|WO2018195357A1|2017-04-19|2018-04-19|Neoantigen identification, manufacture, and use|

---