Methods and systems for predicting HLA class II specific epitopes and characterizing CD4+ T cells

Abstract

The present disclosure provides a method for preparing a personalized cancer vaccine. The present disclosure also provides a method for training a machine learning HLA-peptide presentation prediction model.

Description

[0001] Cross-references

[0002] This application claims the benefit of U.S. Provisional Application No. 62 / 783,914, filed December 21, 2018, U.S. Provisional Application No. 62 / 826,827, filed March 29, 2019, U.S. Provisional Application No. 62 / 855,379, filed May 31, 2019, and U.S. Provisional Application No. 62 / 891,101, filed August 23, 2019, each of which is incorporated herein by reference in its entirety. Background Art

[0003] The major histocompatibility complex (MHC) is a gene complex that encodes human leukocyte antigen (HLA) genes. HLA genes are expressed as protein heterodimers displayed on the surface of human cells to circulating T cells. HLA genes are highly polymorphic, allowing them to fine-tune the adaptive immune system. The adaptive immune response depends in part on the ability of T cells to identify and eliminate cells that display disease-associated peptide antigens that are bound to human leukocyte antigen (HLA) heterodimers.

[0004] In humans, endogenous and exogenous proteins are processed into peptides by the proteasome, as well as by cytoplasmic and endosomal / lysosomal proteases and peptidases, and presented by two classes of cell-surface proteins encoded by MHC genes. These cell-surface proteins are known as human leukocyte antigens (HLA class I and class II), and the set of peptides that bind to them and trigger an immune response is known as HLA epitopes. HLA epitopes are key components that enable the immune system to detect danger signals such as pathogen infection and self-transformation. CD4+ T cells recognize class II MHC (HLA-DR, HLA-DQ, and HLA-DP) epitopes displayed on antigen-presenting cells (APCs) such as dendritic cells and macrophages. Endogenous processing and presentation of HLA class II ligands is a complex process involving a diverse subset of molecular chaperones and enzymes, all of which have not yet been fully characterized. HLA class II peptide presentation activates helper T cells, which subsequently promote B cell differentiation and antibody production, as well as CTL responses. Activated helper T cells also secrete cytokines and chemokines that activate and induce the differentiation of other T cells.

[0005] Understanding the peptide binding preferences of each HLA class II heterodimer is key to successfully predicting which cancer or tumor-specific antigens are likely to elicit cancer or tumor-specific T cell responses. There is a need for methods to identify and isolate specific HLA class II associated peptides (e.g., neoantigenic peptides). Such methods and isolated molecules can be used, for example, to develop therapeutic agents, including but not limited to immune-based therapeutics. Summary of the Invention

[0006] The methods and compositions described herein can be used in a wide range of applications. For example, the methods and compositions described herein can be used to identify immunogenic antigenic peptides and can be used to develop drugs, such as personalized medicine, and to isolate and characterize antigen-specific T cells.

[0007] CD4+ T cell responses can have anti-tumor activity. High CD4+ T cell response rates can be shown without using class II predictions (e.g., 60% of SLP epitopes in the NeoVax study (49% in NT-001, see Ott et al., Nature, July 13, 2017; 547(7662): 217-221), and 48% of mRNA epitopes in the Biontech study, see Sahin et al., Nature, July 13, 2017; 547(7662): 222-226). It may not be clear whether these epitopes are typically naturally presented (by tumors or phagocytic DCs). It may be desirable to convert high CD4+ T response rates into therapeutic effects by improving the identification of truly presented HLA class II binding epitopes.

[0008] The role of gene expression, enzyme cleavage, and pathway / localization preferences may not yet be robustly quantified. It may be unclear whether autophagy (HLA class II presentation by tumor cells) or phagocytosis (HLA class II presentation of tumor epitopes by APCs) is the more relevant pathway, although most existing MS data are likely to be presumed to be derived from autophagy. NetMHCIIpan may be the current predictive standard, but it may be considered inaccurate. At the three HLA class II loci (DR, DP, and DQ), data may exist only for certain common alleles of HLA-DR.

[0009] There may be different data generation methods to learn the rules of HLA class II presentation, including field standards and proposed methods. Field standards may include affinity measurements, which may be the basis of the NetMHCIIpan predictor, provide low throughput and require radioactive reagents, and it omits the role of processing. The proposed method may include mass spectrometry, where data from cell lines / tissues / tumors can help determine the processing rules of autophagy, while monoallelic MS may allow for the determination of allele-specific binding rules (assuming that multiallelic MS data are too complex for effective learning (Bassani-Sternberg. MCP. 2018)).

[0010] There may be different approaches to validate new HLA class II predictors: validation on withheld MS data, which may be the default setting; retrospective studies of vaccine studies (e.g., NT-001), where immune monitoring data may assess vaccine peptide loading on APCs rather than tumor presentation, and the data may be sparsely spread across many different alleles; biochemical affinity measurements, which can be configured to obtain measurements for discordantly predicted peptides (for only 2–3 alleles); and T cell induction, which can be configured to test the rate at which Neon-preferred and NetMHCIIpan-preferred epitopes induce ex vivo T cell responses.

[0011] For validation by T cell induction, the default approach can include evaluating predicted inconsistent neoORFs from TCGA, where the induction material can include healthy donor APCs and T cells, and induction and readout can be by SLP (approximately 15-mer peptides). Random peptides can produce high response rates, while SLP may not adequately address processing issues. Possible solutions may include induction by mRNA.

[0012] Methods disclosed herein may include generating LC-MS / MS monoallelic data for training allele-specific machine learning methods for epitope prediction. Such methods may include improving LC-MS / MS data quality using a set of quality metrics to rigorously remove false positives that increase predictive model performance; identifying allele-specific HLA class II binding cores from HLA-ligand panel LC-MS / MS datasets; utilizing machine learning algorithms to improve HLA class II ligand and epitope predictions; and / or identifying biological variables that influence HLA class II-ligand presentation and improve HLA class II epitope predictions, such as gene expression, cleavage, gene preference, cellular localization, and secondary structure.

[0013] Provided herein is a method comprising: (a) processing amino acid information of a plurality of candidate peptide sequences using a machine learning HLA peptide presentation prediction model to generate a plurality of presentation predictions, wherein each candidate peptide sequence in the plurality of candidate peptide sequences is encoded by a genome or exome of a subject, wherein the plurality of presentation predictions comprises an HLA presentation prediction for each of the plurality of candidate peptide sequences, wherein each HLA presentation prediction indicates a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject is capable of presenting a given candidate peptide sequence in the plurality of candidate peptide sequences, wherein the machine learning HLA peptide presentation prediction model is trained using training data comprising sequence information of training peptide sequences identified by mass spectrometry as to be presented by HLA proteins expressed in training cells; and (b) identifying, based at least on the plurality of presentation predictions, a peptide sequence in the plurality of peptide sequences as being presented by at least one of the one or more proteins encoded by a class II HLA allele of a cell of the subject; wherein the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.07 according to a PPV determination method.

[0014] Provided herein is a method comprising: (a) processing amino acid information for a plurality of peptide sequences encoded by a genome or exome of a subject using a machine learning HLA peptide binding prediction model to generate a plurality of binding predictions, wherein the plurality of binding predictions comprises an HLA binding prediction for each of the plurality of candidate peptide sequences, each binding prediction indicating a likelihood that one or more proteins encoded by an HLA class II allele of a cell of the subject binds to a given candidate peptide sequence in the plurality of candidate peptide sequences, wherein the machine learning HLA peptide binding prediction model is trained using training data comprising sequence information of peptide sequences identified as binding to an HLA class II protein or an HLA class II protein analog; and (b) identifying, based at least on the plurality of binding predictions, a peptide sequence in the plurality of peptide sequences having a probability of binding to at least one of the one or more proteins encoded by an HLA class II allele of a cell of the subject that is greater than a threshold binding prediction probability value; wherein the machine learning HLA peptide binding prediction model has a positive predictive value (PPV) of at least 0.1 according to a PPV determination method.

[0015] In some embodiments, the machine learning HLA peptide presentation prediction model is trained using training data comprising sequence information of training peptide sequences identified by mass spectrometry as being presented by HLA proteins expressed in training cells.

[0016] In some embodiments, the method comprises ranking at least two peptides presented by at least one of the one or more proteins identified as encoded by a class II HLA allele in cells of the subject based on the presentation prediction.

[0017] In some embodiments, the method includes selecting one or more peptides from the two or more ranked peptides.

[0018] In some embodiments, the method comprises selecting one or more peptides from the plurality of peptides that are identified as being presented by at least one of the one or more proteins encoded by a class II HLA allele in cells of the subject.

[0019] In some embodiments, the method comprises selecting one or more peptides from two or more peptides ranked based on the presentation prediction.

[0020] In some embodiments, the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.07 when processing amino acid information for a plurality of test peptide sequences to generate a plurality of test presentation predictions, each test presentation prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele in a cell of the subject is capable of presenting a given test peptide sequence in the plurality of test peptide sequences, wherein the plurality of test peptide sequences comprises at least 500 test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as to be presented by an HLA protein expressed in a cell, and (ii) at least 499 bait peptide sequences contained within a protein encoded by a genome of an organism, wherein the organism and the subject are the same species, wherein the ratio of the at least one hit peptide sequence to the at least 499 bait peptide sequences in the plurality of test peptide sequences is 1:499, and according to the machine learning HLA peptide presentation prediction model, a top percentage of the plurality of test peptide sequences are predicted to be presented by an HLA protein expressed in a cell.

[0021] In some embodiments, the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.1 when processing amino acid information for a plurality of test peptide sequences to generate a plurality of test binding predictions, each test binding prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject will bind to a given test peptide sequence in the plurality of test peptide sequences, wherein the plurality of test peptide sequences comprises at least 20 test peptide sequences comprising (i) at least one test peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in a cell; The method comprises the following steps: (i) selecting a hit peptide sequence presented by a plurality of test peptide sequences, and (ii) at least 19 bait peptide sequences contained within a protein comprising at least one peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in a cell, such as a single HLA protein expressed in a cell (e.g., a monoallelic cell), wherein the ratio of the at least one hit peptide sequence to the at least 19 bait peptide sequences in the plurality of test peptide sequences is 1:19, and a top percentage of the plurality of test peptide sequences are predicted to bind to an HLA protein expressed in a cell according to a machine learning HLA peptide presentation prediction model.

[0022] In some embodiments, there is no amino acid sequence overlap between the at least one hit peptide sequence and the bait peptide sequence.

[0023] In some embodiments, the positive predictive value (PPV) of the machine learning HLA peptide presentation prediction model is at least 0.08, 0.09, 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.2, 0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29, 0.3, 0.31, 0.32, 0.33, 0.34, 0.35, 0.36, 0.37, 0.38, 0.39, 0.4, 0.41, 0.42, 0.43, 0.44, 0.45, 0.46, 0.47, 0.48, 0.49, 0.5、0.51、0.52、0.53、0.54、0.55、0.56、0.57、0.58、0.59、0.6、0.61、0.62、0.63、0.64、0.65、0.66、0.67、0.68、0.69、0.7、0.71、0.72、0.73、0.74、 0.75, 0.76, 0.77, 0.78, 0.79, 0.8, 0.81, 0.82, 0.83, 0.84, 0.85, 0.86, 0.87, 0.88, 0.89, 0.9, 0.91, 0.92, 0.93, 0.94, 0.95, 0.96, 0.97, 0.98 or 0.99.

[0024] In some embodiments, the at least one hit peptide sequence comprises at least 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, 94, 95, 96, 97, 98, 99, or 100 hit peptide sequences.

[0025] In some embodiments, the at least 499 decoy peptide sequences comprise at least 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 500 0, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 75 00, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 8700, 8800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 10 000、11000、12000、13000、14000、15000、16000、17000、18000、19000、20000、21000、22000、23000、24000、25000、26000、27000、28000、29000、30000、 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41000, 42000, 43000, 44000, 45000, 46000, 47000, 48000, 49000, 50000, 52500 、55000、57500、60000、62500、65000、67500、70000、72500、75000、77500、80000、82500、85000、87500、90000、92500、95000、97500、100000、125000、1 50000、175000、200000、225000、250000、275000、300000、325000、350000、375000、400000、425000、450000、475000、500000、600000、700000、800000、900,000 or 1,000,000 decoy peptide sequences. One skilled in the art will recognize that changing the hit: decoy ratio changes the PPV.

[0026] In some embodiments, the at least 500 test peptide sequences comprise at least 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300 100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 8700, 8800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 1000 0, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000, 19000, 20000, 21000, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31 000、32000、33000、34000、35000、36000、37000、38000、39000、40000、41000、42000、43000、44000、45000、46000、47000、48000、49000、50000、52500、 55000、57500、60000、62500、65000、67500、70000、72500、75000、77500、80000、82500、85000、87500、90000、92500、95000、97500、100000、125000、15 0000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000,900,000 or 1,000,000 test peptide sequences.

[0027] In some embodiments, the top percentage is the top 0.20%, 0.30%, 0.40%, 0.50%, 0.60%, 0.70%, 0.80%, 0.90%, 1.00%, 1.10%, 1.20%, 1.30%, 1.40%, 1.50%, 1.60%, 1.70%, 1.80%, 1.90%, 2.00%, 2.10%, 2.20%, 2.30%, 2.40%, 2.50%, %, 2.60%, 2.70%, 2.80%, 2.90%, 3.00%, 3.10%, 3.20%, 3.30%, 3.40%, 3.50%, 3.60%, 3.70%, 3.80%, 3.90%, 4.00%, 4.10%, 4.20%, 4.30%, 4.40%, 4.50%, 4.60%, 4.70%, 4.80%, 4.90%, 5.00%, 5.10%, 5.20% %, 5.30%, 5.40%, 5.50%, 5.60%, 5.70%, 5.80%, 5.90%, 6.00%, 6.10%, 6.20%, 6.30%, 6.40%, 6.50%, 6.60%, 6.70%, 6.80%, 6.90%, 7.00%, 7.10%, 7.20%, 7.30%, 7.40%, 7.50%, 7.60%, 7.70%, 7.80%, 7.90%, 8.00%, 8.10%, 8.20%, 8.30%, 8.40%, 8.50%, 8.60%, 8.70%, 8.80%, 8.90%, 9.00%, 9.10%, 9.20%, 9.30%, 9.40%, 9.50%, 9.60%, 9.70%, 9.80%, 9.90%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19% or 20%.

[0028] In some embodiments, the at least one hit peptide sequence comprises at least 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, 94, 95, 96, 97, 98, 99, or 100 hit peptide sequences.

[0029] In some embodiments, the at least 19 decoy peptide sequences comprise at least 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400 , 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900 , 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400 , 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 8700, 8800, 8900 、9000、9100、9200、9300、9400、9500、9600、9700、9800、9900、10000、11000、12000、13000、14000、15000、16000、17000、18000、19000、20000、21000、22 000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41000, 42000, 4 3000、44000、45000、46000、47000、48000、49000、50000、52500、55000、57500、60000、62500、65000、67500、70000、72500、75000、77500、80000、82500、85,000, 87,500, 90,000, 92,500, 95,000, 97,500, 100,000, 125,000, 150,000, 175,000, 200,000, 225,000, 250,000, 275,000, 300,000, 325,000, 350,000, 375,000, 400,000, 425,000, 450,000, 475,000, 500,000, 600,000, 700,000, 800,000, 900,000, or 1,000,000 bait peptide sequences.

[0030] In some embodiments, the at least 20 test peptide sequences comprise at least 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400 400、1500、1600、1700、1800、1900、2000、2100、2200、2300、2400、2500、2600、2700、2800、2900、3000、3100、3200、3300、3400、3500、3600、3700、3800、 3900、4000、4100、4200、4300、4400、4500、4600、4700、4800、4900、5000、5100、5200、5300、5400、5500、5600、5700、5800、5900、6000、6100、6200、6300、 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 8700, 8800 , 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 10000, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000, 19000, 20000, 210 00, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41000, 4 2000、43000、44000、45000、46000、47000、48000、49000、50000、52500、55000、57500、60000、62500、65000、67500、70000、72500、75000、77500、80000、82500, 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 test peptide sequences.

[0031] In some embodiments, the top percentile is the top 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%, or 40%.

[0032] In some embodiments, the PPV is greater than the corresponding PPV in Column 2 of Table 11 for proteins encoded by corresponding HLA alleles of Table 11. In some embodiments, the PPV is at least equal to the corresponding PPV in Column 3 of Table 11 for proteins encoded by corresponding HLA alleles of Table 11.

[0033] In some embodiments, for a protein encoded by an HLA class II allele, the PPV is equal to or greater than the corresponding PPV in column 2 of Table 12.

[0034] In some embodiments, the PPV is greater than the corresponding PPV in column 2 of Table 16 for a protein encoded by an HLA class II allele.

[0035] In some embodiments, the subject is a single subject.

[0036] In some embodiments, the subject is a mammal.

[0037] In some embodiments, the subject is a human.

[0038] In some embodiments, the training cell is a cell that expresses a single protein encoded by an HLA class II allele of the subject cell.

[0039] In some embodiments, the training cell is a monoallelic HLA cell, or a cell expressing an HLA allele with an affinity tag.

[0040] In some embodiments, the subject's cells include cancer cells.

[0041] In some embodiments, the methods are used to identify peptide sequences.

[0042] In some embodiments, the methods are used to select peptide sequences.

[0043] In some embodiments, the methods are used in preparation for cancer treatment.

[0044] In some embodiments, the methods are used to prepare a subject for a specific cancer treatment.

[0045] In some embodiments, the methods are used to prepare cancer cell-specific cancer treatments.

[0046] In some embodiments, each peptide sequence in the plurality of peptide sequences is associated with cancer.

[0047] In some embodiments, at least one peptide sequence in the plurality of peptide sequences is overexpressed by a cancer cell in the subject.

[0048] In some embodiments, each peptide sequence in the plurality of peptide sequences is overexpressed by a cancer cell in the subject.

[0049] In some embodiments, at least one peptide sequence in the plurality of peptide sequences is a cancer cell-specific peptide.

[0050] In some embodiments, each peptide sequence in the plurality of peptide sequences is a cancer cell specific peptide.

[0051] In some embodiments, each peptide sequence in the plurality of peptide sequences is expressed by a cancer cell in the subject.

[0052] In some embodiments, at least one peptide sequence in the plurality of peptide sequences is not encoded by a non-cancerous cell of the subject.

[0053] In some embodiments, each peptide sequence in the plurality of peptide sequences is not encoded by a non-cancerous cell of the subject.

[0054] In some embodiments, at least one peptide sequence in the plurality of peptide sequences is not expressed by non-cancerous cells of the subject.

[0055] In some embodiments, each peptide sequence in the plurality of peptide sequences is not expressed by non-cancerous cells of the subject.

[0056] In some embodiments, the method comprises obtaining the plurality of peptide sequences from a subject.

[0057] In some embodiments, the method comprises obtaining a plurality of polynucleotide sequences from the subject.

[0058] In some embodiments, the method comprises obtaining a plurality of polynucleotide sequences from the subject, the polynucleotide sequences encoding a plurality of peptide sequences encoded by the genome or exome of the subject or by a pathogen or virus in the subject.

[0059] In some embodiments, the method includes obtaining, by a computer processor, a plurality of polynucleotide sequences from the subject, the polynucleotide sequences encoding a plurality of peptide sequences encoded by the genome or exome of the subject.

[0060] In some embodiments, the method comprises obtaining a plurality of polynucleotide sequences from the subject by genome or exome sequencing.

[0061] In some embodiments, the method comprises obtaining a plurality of polynucleotide sequences of the subject by whole genome sequencing or whole exome sequencing.

[0062] In some embodiments, processing comprises processing with a computer processor.

[0063] In some embodiments, processing includes generating a plurality of predictor variables based at least on amino acid information of the plurality of peptide sequences.

[0064] In some embodiments, the plurality of predictor variables is processed using a machine learning HLA-peptide presentation prediction model.

[0065] In some embodiments, the one or more proteins encoded by the class II HLA alleles of cells of the subject are one or more proteins encoded by the class II HLA alleles expressed by the subject.

[0066] In some embodiments, the one or more proteins encoded by the class II HLA alleles of the subject's cells are one or more proteins encoded by class II HLA alleles expressed by cancer cells of the subject.

[0067] In some embodiments, the one or more proteins encoded by the HLA class II alleles of the subject's cells is a single protein encoded by the HLA class II alleles of the subject's cells.

[0068] In some embodiments, the one or more proteins encoded by the class II HLA alleles of the subject's cells are two, three, four, five, or six or more proteins encoded by the class II HLA alleles of the subject's cells.

[0069] In some embodiments, the one or more proteins encoded by the class II HLA alleles of the subject's cells are every protein encoded by the class II HLA alleles of the subject's cells.

[0070] In some embodiments, the method further comprises administering to the subject a composition comprising one or more peptide sequences from the selected subset of peptide sequences.

[0071] In some embodiments, identifying the plurality of peptide sequences comprises comparing a DNA, RNA, or protein sequence from a cancer cell of the subject to a DNA, RNA, or protein sequence from a normal cell of the subject, wherein each of the plurality of peptides comprises at least one mutation that is present in the cancer cell of the subject but not in the normal cell of the subject.

[0072] In some embodiments, the machine learning HLA-peptide presentation prediction model comprises a plurality of predictor variables identified based at least on training data, wherein the training data comprises training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information and a likelihood of presentation generated as an output based on the amino acid position information and the plurality of predictor variables.

[0073] In some embodiments, identifying includes identifying, based at least on the multiple presentation predictions, a peptide sequence from the multiple peptide sequences having a probability of being presented by at least one of the one or more proteins encoded by the class II HLA alleles of the subject's cells that is greater than a threshold presentation prediction probability value.

[0074] In some embodiments, one or more of 0.2% of the plurality of test peptide sequences predicted by the machine learning HLA peptide presentation prediction model to be presented have a probability of being presented by at least one of the one or more proteins encoded by a class II HLA allele of a cell of the subject that is greater than a threshold presentation prediction probability value.

[0075] In some embodiments, each of 0.2% of the plurality of test peptide sequences predicted to be presented by the machine learning HLA peptide presentation prediction model has a probability of being presented by at least one of the one or more proteins encoded by a class II HLA allele of a cell of the subject that is greater than a threshold presentation prediction probability value.

[0076] In some embodiments, the number of positives is limited to equal the number of hits.

[0077] In some embodiments, the mass spectrometry is monoallelic mass spectrometry.

[0078] In some embodiments, the peptide is presented by an HLA protein expressed in the cell via autophagy.

[0079] In some embodiments, the peptide is presented by an HLA protein expressed in a cell by phagocytosis.

[0080] In some embodiments, the plurality of predictor variables comprises predicted values ​​for expression levels of source proteins comprising peptides.

[0081] In some embodiments, the plurality of predictor variables comprises a stability prediction value for a source protein comprising the peptide.

[0082] In some embodiments, the plurality of predictor variables comprises a predicted value for the degradation rate of a source protein comprising the peptide.

[0083] In some embodiments, the plurality of predictor variables comprises a predicted value for protein cleavage of a source protein comprising a peptide.

[0084] In some embodiments, the plurality of predictor variables comprises a prediction of cellular or tissue localization of a source protein comprising the peptide.

[0085] In some embodiments, the plurality of predictor variables includes a prediction value of the intracellular processing pattern of the source protein comprising the peptide, wherein the processing pattern of the source protein includes, among other things, a prediction value of whether the source protein undergoes autophagy, phagocytosis, and intracellular trafficking.

[0086] In some embodiments, the quality of the training data is improved by using multiple quality metrics.

[0087] In some embodiments, the plurality of quality metrics includes common contaminant peptide removal, high scoring peak intensity, high score, and high mass accuracy.

[0088] In some embodiments, the scored peak intensity is at least 50%.

[0089] In some embodiments, the scored peak intensity is at least 60%.

[0090] In some embodiments, the score is at least 7.

[0091] In some embodiments, the mass accuracy is at most 5 ppm.

[0092] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single immunoprecipitated HLA protein expressed in a cell.

[0093] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single exogenous HLA protein expressed in the cell.

[0094] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single recombinant HLA protein expressed in a cell.

[0095] In some embodiments, the plurality of predictor variables includes a peptide-HLA affinity predictor variable.

[0096] In some embodiments, peptides presented by HLA proteins include peptides identified by searching a database of enzyme-specific, unmodified peptides.

[0097] In some embodiments, peptides presented by HLA proteins include peptides identified by searching a peptide database using a reverse database search strategy.

[0098] In some embodiments, the HLA protein comprises an HLA-DR, HLA-DQ, or HLA-DP protein.

[0099] In some embodiments, the HLA protein comprises an HLA class II protein selected from the group consisting of: HLA-DPB1*01:01 / HLA-DPA1*01:03, HLA-DPB1*02:01 / HLA-DPA1*01:03, HLA-DPB1*03:01 / HLA-DPA1*01:03, HLA-DPB1*04:01 / HLA-DPA1*01:03, HLA-DPB1*04:02 / HLA-DPA1*01:03, HLA-DPB1*06:01 / HLA-DPA1*01:03, HLA-DQB1*02:01 / HLA-DQA1*05:01, HLA-DQB1*02:02 / HLA-DQA1*02:01, HLA-DQB1*06:02 / HLA-DQA1*01:02, HLA-DQB1*06:04 / HLA-DQA1*01:02, HLA-DRB1*01:01, HLA-DRB1*01:02, HLA-DRB1*03:01, HLA-DRB1*03:02, HLA-DRB1*04:01, HLA-DRB1*04:02, HLA-DRB1*04:03, HLA-DRB1*04:04, HLA-DRB1*04:05, HLA-DRB1*04:07, HLA-DRB1*07:01, HLA-DRB1*08:01, HLA-DRB1*08:02, HLA-DRB1*08:03, HLA-DRB1*08:04, HLA-DRB1*09:01, HLA-DRB1*10:01, HLA-DRB1*11:01, HLA-DRB1*11:02, HLA-DRB1*11:04, HLA-DRB1*12:01, HLA-DRB1*12:02, HLA-DRB1*13:01, HLA-DRB1*13:02, HLA-DRB1*13:03, HLA-DRB1*14:01, HLA-DRB1*15:01, HLA-DRB1*15:02, HLA-DRB1*15:03, HLA-DRB1*16:01, HLA-DRB3*01:01, HLA-DRB3*02:02, HLA-DRB3*03:01, HLA-DRB4*01:01, HLA-DRB5*01:01.

[0100] In some embodiments, the HLA-DR pairs with DRA*01:01.

[0101] In some embodiments, the HLA protein is an HLA class II protein selected from the group consisting of DPA*01:03 / DPB*04:01, DRB1*01:01, DRB1*01:02, DRB1*03:01, DRB1*04:01, DRB1*04:02, DRB1*04:04, DRB1*04:05, DRB1*07:01, DRB1*08:01, DRB1*08:02, DRB1*08:03, DRB1*09:01, DRB1*11:01, DRB1*11:02, DRB 1*11:04, DRB1*12:01, DRB1*13:01, DRB1*13:02, DRB1*13:03, DRB1*14:01, DRB1*15:01, DRB1*15:02, DRB1*15 :03, DRB1*16:02, DRB3*01:01, DRB3*02:01, DRB3*02:02, DRB3*03:01, DRB4*01:01, DRB4*01:03 and DRB5*01:01.

[0102] In some embodiments, the HLA-DR protein comprises DRA*01:01 in dimeric form.

[0103] In some embodiments, the HLA protein includes an HLA-DP protein selected from the group consisting of DPB1*01:01, DPB1*02:01, DPB1*02:02, DPB1*03:01, DPB1*04:01, DPB1*04:02, DPB1*05:01, DPB1*06:01, DPB1*11:01, DPB1*13:01, and DPB1*17:01.

[0104] In some embodiments, the HLA-DP protein pair comprises DPA1*01:03.

[0105] In some embodiments, the HLA protein includes an HLA-DQ protein complex selected from the group consisting of A1*01:01+B1*05:01, A1*01:02+B1*06:02, A1*01:02+B1*06:04, A1*01:03+B1*06:03, A1*02:01+B1*02:02, A1*02:01+B1*03:03, A1*03:01+B1*03:02, A1*03:03+B1*03:01, A1*05:01+B1*02:01, and A1*05:05+B1*03:01.

[0106] In some embodiments, peptides presented by HLA proteins include peptides identified by comparing the MS / MS spectrum of the HLA-peptide to the MS / MS spectra of one or more peptides or proteins in a peptide or protein database.

[0107] In some embodiments, the mutation is selected from the group consisting of a point mutation, a splice site mutation, a frameshift mutation, a readthrough mutation, and a gene fusion mutation.

[0108] In some embodiments, the peptide presented by the HLA protein has a length of 15-40 amino acids.

[0109] In some embodiments, peptides presented by HLA proteins include peptides identified by comparing the MS / MS spectrum of the HLA-peptide with the MS / MS spectra of one or more peptides or proteins in a peptide or protein database to identify peptides presented by HLA proteins.

[0110] In some embodiments, the personalized cancer therapy further comprises an adjuvant.

[0111] In some embodiments, the personalized cancer therapy further comprises an immune checkpoint inhibitor.

[0112] In some embodiments, the training data includes structured data, time series data, unstructured data, relational data, or any combination thereof.

[0113] In some embodiments, the unstructured data includes image data.

[0114] In some embodiments, the relationship data includes data from a client system, an enterprise system, an operating system, a website, a web-accessible application programming interface (API), or any combination thereof.

[0115] In some embodiments, the training data is uploaded to a cloud-based database.

[0116] In some embodiments, the training is performed using a convolutional neural network.

[0117] In some embodiments, the convolutional neural network comprises at least two convolutional layers.

[0118] In some embodiments, the convolutional neural network includes at least one batch normalization step.

[0119] In some embodiments, the convolutional neural network includes at least one spatial dropout step.

[0120] In some embodiments, the convolutional neural network includes at least one global max pooling step.

[0121] In some embodiments, the convolutional neural network comprises at least one dense layer.

[0122] In some embodiments, identifying the peptide sequence comprises identifying a peptide sequence having a mutation that is expressed in a cancer cell of the subject.

[0123] In some embodiments, identifying a peptide sequence comprises identifying a peptide sequence that is not expressed in normal cells of the subject.

[0124] In some embodiments, identifying a peptide sequence comprises identifying a viral peptide sequence.

[0125] In some embodiments, identifying a peptide sequence comprises identifying an overexpressed peptide sequence.

[0126] Provided herein is a method for identifying an HLA class II specific peptide for immunotherapy of a subject, comprising: obtaining, by a computer processor, a candidate peptide comprising an epitope and a plurality of peptide sequences, each peptide sequence comprising the epitope; processing, by a computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences to an immune cell, each presentation prediction indicating a likelihood that one or more proteins encoded by an HLA class II allele can present a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data, the training data comprising a peptide sequence generated by a protein expressed in a cell; The method comprises the steps of: obtaining sequence information of a peptide presented by an HLA protein of the subject and identified by mass spectrometry; selecting a protein from one or more proteins encoded by an HLA class II allele of a cell of the subject, the protein being predicted to bind to the candidate peptide by a machine learning HLA-peptide presentation prediction model, wherein the probability of the protein presenting the candidate peptide to an immune cell is greater than a threshold presentation prediction probability value; contacting the candidate peptide with the selected protein such that the candidate peptide competes for a placeholder peptide associated with the selected protein; and identifying the candidate peptide as a peptide for immunotherapy specific to the selected protein based on whether the candidate peptide replaces the placeholder peptide.

[0127] In some embodiments, obtaining comprises identifying a candidate peptide, wherein identifying the candidate peptide comprises comparing a DNA, RNA, or protein sequence from a cancer cell of the subject to a DNA, RNA, or protein sequence from a normal cell of the subject.

[0128] In some embodiments, processing includes determining a plurality of predictor variables based at least on amino acid information of the plurality of peptide sequences, and processing the plurality of predictor variables using a machine learning HLA-peptide presentation prediction model.

[0129] In some embodiments, the machine learning HLA-peptide presentation prediction model comprises a plurality of predictor variables identified based at least on training data, wherein the training data comprises training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information and a likelihood of presentation generated as an output based on the amino acid position information and the plurality of predictor variables.

[0130] In some embodiments, the number of positives is limited to equal the number of hits.

[0131] In some embodiments, the mass spectrometry is monoallelic mass spectrometry.

[0132] In some embodiments, the plurality of predictor variables include any one or more of the following: a predicted value for the expression level of a source protein comprising the peptide, a predicted value for stability, a predicted value for degradation rate, a predicted value for cleavage, a predicted value for cell or tissue localization, and a predicted value for intracellular processing modes including autophagy, phagocytosis, and intracellular transport.

[0133] In some embodiments, the quality of the training data is improved by using multiple quality metrics.

[0134] In some embodiments, the plurality of quality metrics includes common contaminant peptide removal, high scoring peak intensity, high score, and high mass accuracy.

[0135] In some embodiments, the scored peak intensity is at least 50%.

[0136] In some embodiments, the scored peak intensity is at least 60%.

[0137] In some embodiments, the placeholder peptide is a CLIP peptide.

[0138] In some embodiments, the placeholder peptide is a CMV peptide.

[0139] In some embodiments, the method further comprises measuring the IC50 of replacement of the placeholder peptide by the peptide of interest.

[0140] In some embodiments, the IC50 of the placeholder peptide being replaced by the peptide of interest is less than 500 nM.

[0141] In some embodiments, at least one protein from one or more proteins encoded by an HLA class II allele in a cell of the subject is an HLA class II tetramer or multimer.

[0142] In some embodiments, the target peptide is further identified by mass spectrometry.

[0143] In some embodiments, the at least one protein encoded by an HLA class II allele of a cell of the subject is a recombinant protein.

[0144] In some embodiments, the at least one protein encoded by an HLA class II allele in cells of the subject is expressed in a eukaryotic cell.

[0145] In some embodiments, the peptide is presented by an HLA protein expressed in the cell via autophagy.

[0146] In some embodiments, the peptide is presented by an HLA protein expressed in a cell by phagocytosis.

[0147] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single immunoprecipitated HLA protein expressed in a cell.

[0148] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single exogenous HLA protein expressed in the cell.

[0149] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single recombinant HLA protein expressed in a cell.

[0150] In some embodiments, the plurality of predictor variables includes a peptide-HLA affinity predictor variable.

[0151] In some embodiments, peptides presented by HLA proteins include peptides identified by searching a database of enzyme-specific, unmodified peptides.

[0152] In some embodiments, peptides presented by HLA proteins include peptides identified by searching a peptide database using a reverse database search strategy.

[0153] In some embodiments, the HLA protein comprises an HLA-DR, HLA-DQ, or HLA-DP protein.

[0154] In some embodiments, the immunotherapy is cancer immunotherapy.

[0155] In some embodiments, the epitope is a cancer-specific epitope.

[0156] In some embodiments, the at least one protein encoded by the HLA class II allele comprises at least the α1 subunit and the β1 subunit of the HLA protein that exist in the form of a dimer.

[0157] In some embodiments, the identity of the peptide is known.

[0158] In some embodiments, the identity of the peptide is unknown.

[0159] In some embodiments, the identity of the peptide is determined by mass spectrometry.

[0160] In some embodiments, the peptide exchange assay comprises detecting a peptide fluorescent probe or tag.

[0161] In some embodiments, the placeholder peptide is a CLIP peptide. In some embodiments, the placeholder peptide has the amino acid sequence PVSKMRMATPLLMQA.

[0162] In some embodiments, the polynucleic acid construct comprises an expression vector, which further comprises one or more of the following: a promoter, a secretion signal, a dimerization factor, a ribosomal skipping sequence, and one or more tags for purification and / or detection.

[0163] In some embodiments, the placeholder peptide sequence is encoded by a nucleic acid sequence within a vector.

[0164] In some embodiments, the sequence encoding the cleavage domain is placed between the sequence encoding the placeholder peptide and the sequence encoding the HLA β1 peptide.

[0165] Provided herein is a method for determining the immunogenicity of an MHC class II binding peptide, comprising: selecting a protein encoded by an HLA class II allele that is predicted to bind to an MHC class II binding peptide by a machine learning HLA-peptide presentation prediction model, wherein the machine learning HLA-peptide presentation prediction model is configured to generate a presentation prediction for a given peptide sequence, the presentation prediction indicating a likelihood that one or more proteins encoded by the HLA class II allele can present the given peptide sequence, and wherein the probability that the protein presents the MHC class II binding peptide is greater than a threshold presentation prediction probability value; contacting the peptide with the selected protein such that the peptide competes for and displaces a placeholder peptide associated with the selected protein, thereby forming a complex comprising the HLA class II protein and the MHC class II binding peptide; contacting the complex with a CD4+ T cell, and measuring one or more activation parameters of the CD4+ T cell, the parameters selected from the group consisting of: induction of cytokines, induction of chemokines, and expression of cell surface markers.

[0166] In some embodiments, the HLA class II allele is a tetramer or a multimer.

[0167] In some embodiments, the cytokine is IL-2.

[0168] Provided herein is a method for inducing CD4+ T cell activation in a subject for cancer immunotherapy, the method comprising: identifying a peptide sequence associated with cancer and comprising a cancer mutation, wherein identifying the peptide sequence comprises comparing a DNA, RNA, or protein sequence from a cancer cell of the subject with a DNA, RNA, or protein sequence from a normal cell of the subject; selecting a protein encoded by an HLA class II allele, the protein being normally expressed by cells of the subject and predicted to bind to the peptide by a machine learning HLA-peptide presentation prediction model; wherein the prediction model has a positive predictive value of at least 0.1 at a recall rate of at least 0.1%, 0.1%-50%, or at most 50%, and wherein the probability that the protein presents the identified peptide sequence is greater than a threshold presentation prediction probability value; contacting the identified peptide with the selected protein encoded by the HLA class II allele to verify whether the identified peptide competes with the selected protein encoded by the HLA class II allele. The invention relates to a method for identifying a placeholder peptide associated with a selected protein encoded by a class II allele, thereby replacing the placeholder peptide with an IC50 value of less than 500 nM; optionally, purifying the identified peptide; and administering to the subject an effective amount of a polypeptide comprising the sequence of the identified peptide or a polynucleotide encoding the polypeptide.

[0169] Provided herein is a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, comprising: obtaining, by a computer processor, a plurality of peptide sequences of the polypeptide sequence; processing, by a computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by a class I or class II MHC allele of a cell of the subject is capable of presenting an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information associated with HLA proteins expressed in cells; determining or predicting, based on the plurality of presentation predictions, that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject; and administering to the subject a composition comprising the drug.

[0170] Provided herein is a method for preparing HLA class II tetramers or multimers by conjugating four separate HLA protein α1 and β1 heterodimers, the method comprising: expressing in eukaryotic cells a vector comprising nucleic acid sequences encoding the α and β chains of the HLA proteins, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, such that each HLA protein α1 and β1 heterodimer is secreted in a dimerized state, wherein the heterodimer is associated with a placeholder peptide, purifying the secreted heterodimers from the cell culture medium, verifying peptide binding activity using a peptide exchange assay, adding streptavidin to conjugate the heterodimers into tetramers, and purifying the tetramers with a yield greater than 1 mg / L. Multimers, such as pentamers, hexamers, or octamers, can also be produced and are also contemplated herein.

[0171] In some embodiments, the vector comprises a CMV promoter.

[0172] In some embodiments, the vector comprises a sequence encoding a placeholder peptide linked to the β1 chain via a cleavable site.

[0173] In some embodiments, peptide exchange assays involve pre-cleavage of a placeholder peptide from a beta chain.

[0174] In some embodiments, the cleavable site is a thrombin cleavage site.

[0175] In some embodiments, the peptide exchange assay is a FRET assay.

[0176] In some embodiments, the purification is performed by any of the following: column chromatography, ion exchange chromatography, size exclusion chromatography, affinity chromatography, or LC-MS.

[0177] Provided herein is an HLA class II tetramer or multimer comprising HLA-DR or HLA-DP or HLA-DQ heterodimers, each heterodimer comprising an α and a β chain, wherein the heterodimers are purified and present at a concentration greater than 1 mg / L.

[0178] In some embodiments, the HLA class II tetramer is selected from Tables 8A-8C.

[0179] In some embodiments, the HLA class II tetramer comprises a heterodimeric pair selected from the group consisting of HLA-DR, HLA-DP, and HLA-DQ proteins.

[0180] In some embodiments, the HLA protein is an HLA class II protein selected from the group consisting of HLA-DPB1*01:01 / HLA-DPA1*01:03, HLA-DPB1*02:01 / HLA-DPA1*01:03, HLA-DPB1*03:01 / HLA-DPA1*01:03, HLA-DPB1*04:01 / HLA-DPA1*01:03, HLA-DPB1*04:02 / HLA-DPA1*01:03, HLA-DPB1*06:01 / HLA-DPA1*01:03, HLA-DQB1*02:01 / HLA-DQA 1*05:01, HLA-DQB1*02:02 / HLA-DQA1*02:01, HLA-DQB1*06:02 / HLA-DQA1*01:02, HLA-DQB1*06:04 / HLA-DQA1*01:02, HLA-DR B1*01:01, HLA-DRB1*01:02, HLA-DRB1*03:01, HLA-DRB1*03:02, HLA-DRB1*04:01, HLA-DRB1*04:02, HLA-DRB1*04:03, HLA-D RB1*04:04, HLA-DRB1*04:05, HLA-DRB1*04:07, HLA-DRB1*07:01, HLA-DRB1*08:01, HLA-DRB1*08:02, HLA-DRB1*08:03, HLA- DRB1*08:04, HLA-DRB1*09:01, HLA-DRB1*10:01, HLA-DRB1*11:01, HLA-DRB1*11:02, HLA-DRB1*11:04, HLA-DRB1*12:01, HLA -DRB1*12:02, HLA-DRB1*13:01, HLA-DRB1*13:02, HLA-DRB1*13:03, HLA-DRB1*14:01, HLA-DRB1*15:01, HLA-DRB1*15:02, HL A-DRB1*15:03, HLA-DRB1*16:01, HLA-DRB3*01:01, HLA-DRB3*02:02, HLA-DRB3*03:01, HLA-DRB4*01:01 and HLA-DRB5*01:01).

[0181] In some embodiments, the heterodimeric pair is expressed in a eukaryotic cell.

[0182] In some embodiments, the heterodimer pair is encoded by a vector.

[0183] Provided herein is a vector, wherein the vector comprises a nucleic acid sequence encoding the α chain and β chain of the HLA protein described herein, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, such that each HLA protein α1 and β1 heterodimer is secreted in a dimerized state, wherein the secreted heterodimer is optionally associated with a placeholder peptide.

[0184] Provided herein is a cell comprising a vector described herein.

[0185] In some embodiments, the HLA class II heterodimer is secreted from eukaryotic cells into the cell culture medium, which is further purified by any of the following methods: column chromatography, ion exchange chromatography, size exclusion chromatography, affinity chromatography, or LC-MS.

[0186] Provided herein is a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, comprising: obtaining, by a computer processor, a plurality of peptide sequences of the polypeptide sequence; processing, by a computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by class I or class II MHC alleles of a cell of the subject can present an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; and determining or predicting, based on the plurality of presentation predictions, that at least one of the plurality of peptide sequences of the polypeptide sequence is immunogenic to the subject.

[0187] Provided herein is a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that an epitope sequence of a given peptide sequence is presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject; wherein the machine learning HLA-peptide presentation prediction model comprises: a plurality of predictor variables determined based at least on training data; wherein the training data comprises: sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables; determining or predicting, based on the set of presentation predictions, that each of the peptide sequences of the polypeptide sequence is not immunogenic to the subject; and administering a composition comprising the drug to the subject.

[0188] Provided herein is a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that an epitope sequence of a given peptide sequence is presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject; wherein the machine learning HLA-peptide presentation prediction model comprises: a plurality of predictor variables determined based at least on training data; wherein the training data comprises: sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables; and determining or predicting, based on the set of presentation predictions, that at least one of the peptide sequences of the polypeptide sequence is immunogenic to the subject.

[0189] Provided herein is a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, comprising: obtaining, by a computer processor, a plurality of peptide sequences of the polypeptide sequence; processing, by a computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by a class I or class II MHC allele of a cell of the subject is capable of presenting an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information associated with HLA proteins expressed in cells; determining or predicting, based on the plurality of presentation predictions, that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject; and administering to the subject a composition comprising the drug.

[0190] In some embodiments, the method further comprises deciding not to administer the drug to the subject.

[0191] In some embodiments, the drug comprises an antibody or a binding fragment thereof.

[0192] In some embodiments, the peptide sequence of the polypeptide sequence has a length of 8, 9, 10, 11, or 12 amino acids, and wherein the protein encoded by a class I or class II MHC allele of a cell of the subject is a protein encoded by a class I MHC allele of a cell of the subject.

[0193] In some embodiments, the peptide sequence of the polypeptide sequence has a length of 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25 amino acids, and wherein the protein encoded by a class I or class II MHC allele of a cell of the subject is a protein encoded by a class II MHC allele of a cell of the subject.

[0194] Provided herein is a method of treating a subject having an autoimmune disease or condition, comprising: (a) identifying or predicting an epitope of an expressed protein presented by class I or class II MHC on cells of the subject, wherein a complex comprising the identified or predicted epitope and class I or class II MHC is targeted by CD8 or CD4 T cells of the subject; (b) identifying a T cell receptor (TCR) that binds to the complex; (c) expressing the TCR in regulatory T cells or allogeneic regulatory T cells from the subject; and (d) administering regulatory T cells expressing the TCR to the subject.

[0195] In some embodiments, the autoimmune disease or condition is diabetes.

[0196] In some embodiments, the cells are pancreatic islet cells.

[0197] Provided herein is a method of treating a subject having an autoimmune disease or condition, comprising administering to the subject a regulatory T cell expressing a T cell receptor (TCR) that binds to a complex comprising: (i) an epitope of an expressed protein identified or predicted to be presented by class I or class II MHC on cells of the subject, and (ii) class I or class II MHC, wherein the complex is targeted by CD8 or CD4 T cells of the subject.

[0198] Provided herein is a computer system for identifying peptide sequences for personalized cancer treatment of a subject, comprising: a database configured to store a plurality of peptide sequences of a subject; and one or more computer processors operably coupled to the database, wherein the one or more computer processors are individually and collectively programmed to: process amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by class II MHC alleles of a cell of the subject is capable of presenting a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information of peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; and select a subset of the plurality of peptide sequences for personalized cancer treatment of the subject based at least on the plurality of presentation predictions.

[0199] Provided herein is a computer system for identifying HLA class II-specific peptides for immunotherapy of a subject, comprising: a database configured to store candidate peptides comprising an epitope and a plurality of peptide sequences, each peptide sequence comprising the epitope; and one or more computer processors operably coupled to the database, wherein the one or more computer processors are individually and collectively programmed to: process amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences to immune cells, each presentation prediction indicating a likelihood that one or more proteins encoded by an HLA class II allele are capable of presenting a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; Selecting a protein from one or more proteins encoded by a class II allele, wherein the protein is predicted to bind to the candidate peptide by a machine learning HLA-peptide presentation prediction model, wherein the probability of the protein presenting the candidate peptide to immune cells is greater than a threshold presentation prediction probability value; and identifying the candidate peptide as a peptide specific for the selected protein for immunotherapy based on whether the candidate peptide replaces the placeholder peptide when the candidate peptide contacts the selected protein such that the candidate peptide competes for the placeholder peptide associated with the selected protein.

[0200] Provided herein is a computer system for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, comprising: a database configured to store a plurality of peptide sequences of the polypeptide sequences; and one or more computer processors operably coupled to the database, wherein the one or more computer processors are individually and collectively programmed to: process amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by class I or class II MHC alleles of a cell of the subject is capable of presenting an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information associated with HLA proteins expressed in cells; and determine or predict, based on the plurality of presentation predictions, that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject, wherein a composition comprising the drug is administered to the subject.

[0201] Provided herein is a computer system for screening drugs comprising polypeptide sequences for immunogenicity in a subject, comprising: a database configured to store a plurality of peptide sequences of the polypeptide sequences; and one or more computer processors operably coupled to the database, wherein the one or more computer processors are individually and collectively programmed to: process amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by class I or class II MHC alleles of a cell of the subject can present an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information of peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; and determine or predict, based on the plurality of presentation predictions, that at least one of the plurality of peptide sequences of the polypeptide sequence is immunogenic to the subject.

[0202] Provided herein is a non-transitory computer-readable medium comprising machine-executable code that, when executed by one or more computer processors, implements a method for identifying peptide sequences for personalized cancer treatment of a subject, the method comprising: obtaining a plurality of peptide sequences for the subject; processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by class II MHC alleles of a cell of the subject is capable of presenting a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information of peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; and selecting a subset of the plurality of peptide sequences for personalized cancer treatment of the subject based at least on the plurality of presentation predictions.

[0203] Provided herein is a non-transitory computer-readable medium comprising machine-executable code that, when executed by one or more computer processors, implements a method for identifying HLA class II-specific peptides for immunotherapy of a subject, the method comprising: obtaining a candidate peptide comprising an epitope and a plurality of peptide sequences, each peptide sequence comprising the epitope; processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences to immune cells, each presentation prediction indicating a likelihood that one or more proteins encoded by an HLA class II allele are capable of presenting a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; and obtaining a plurality of peptide sequences from the HLA cells of the subject. Selecting a protein from one or more proteins encoded by a class II allele, wherein the protein is predicted to bind to the candidate peptide by a machine learning HLA-peptide presentation prediction model, wherein the probability of the protein presenting the candidate peptide to immune cells is greater than a threshold presentation prediction probability value; and identifying the candidate peptide as a peptide specific for immunotherapy for the selected protein based on whether the candidate peptide replaces the placeholder peptide when the candidate peptide contacts the selected protein such that the candidate peptide competes with the placeholder peptide.

[0204] Provided herein is a non-transitory computer-readable medium comprising machine-executable code that, when executed by one or more computer processors, implements a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: obtaining a plurality of peptide sequences of the polypeptide sequence; processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by a class I or class II MHC allele of a cell of the subject is capable of presenting an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information associated with HLA proteins expressed in cells; and determining or predicting, based on the plurality of presentation predictions, that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject, wherein a composition comprising the drug is administered to the subject.

[0205] Provided herein is a non-transitory computer-readable medium comprising machine-executable code that, when executed by one or more computer processors, implements a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: obtaining a plurality of peptide sequences of the polypeptide sequence; processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by class I or class II MHC alleles of a cell of the subject can present an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data comprising sequence information of peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; and determining or predicting, based on the plurality of presentation predictions, that at least one of the plurality of peptide sequences of the polypeptide sequence is immunogenic to the subject.

[0206] Provided herein is a method comprising: processing amino acid information of a plurality of candidate peptide sequences using a machine learning HLA peptide presentation prediction model to generate a plurality of presentation predictions, wherein each candidate peptide sequence in the plurality of candidate peptide sequences is encoded by a genome or exome of a subject, wherein the plurality of presentation predictions comprises an HLA presentation prediction for each of the plurality of candidate peptide sequences, wherein each presentation prediction indicates a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject is capable of presenting a given candidate peptide sequence in the plurality of candidate peptide sequences, wherein the machine learning HLA peptide presentation prediction model is trained using training data comprising sequence information of training peptide sequences identified by mass spectrometry as to be presented by HLA proteins expressed in training cells; and identifying, based at least on the plurality of presentation predictions, a peptide sequence in the plurality of peptide sequences having a probability of being presented by at least one of the one or more proteins encoded by the class II HLA allele of a cell of the subject that is greater than a threshold presentation prediction probability. value; wherein the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.07 when processing amino acid information of a plurality of test peptide sequences to generate a plurality of test presentation predictions, each test presentation prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject is capable of presenting a given test peptide sequence in the plurality of test peptide sequences, wherein the plurality of test peptide sequences comprises at least 500 test peptide sequences, the sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as to be presented by an HLA protein expressed in a cell, and (ii) at least 499 decoy peptide sequences contained within a protein encoded by a genome of an organism, wherein the organism and the subject are the same species, wherein the ratio of the at least one hit peptide sequence to the at least 499 decoy peptide sequences in the plurality of test peptide sequences is 1:499, and according to the machine learning HLA peptide presentation prediction model, 0.2% of the plurality of test peptide sequences are predicted to be presented by an HLA protein expressed in the cell.

[0207] Provided herein is a method comprising: processing amino acid information of a plurality of peptide sequences encoded by a genome or exome of a subject using a machine learning HLA peptide binding prediction model to generate a plurality of binding predictions, wherein the plurality of binding predictions comprises an HLA binding prediction for each of the plurality of candidate peptide sequences, each binding prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject will bind to a given candidate peptide sequence in the plurality of candidate peptide sequences, wherein the machine learning HLA peptide binding prediction model is trained using training data comprising proteins identified to bind to HLA class II proteins or HLA class II proteins. sequence information for a peptide sequence that is bound by a class II protein analog; and identifying, based at least on the plurality of binding predictions, a peptide sequence from the plurality of peptide sequences that has a probability of binding to at least one of the one or more proteins encoded by the class II HLA alleles of the subject's cells that is greater than a threshold binding prediction probability value; wherein the machine learning HLA peptide binding prediction model has a positive predictive value (PPV) of at least 0.1 when processing the amino acid information for the plurality of test peptide sequences to generate a plurality of test binding predictions, each test binding prediction indicating a likelihood that one or more proteins encoded by the class II HLA alleles of the subject's cells will bind to a given test peptide sequence from the plurality of test peptide sequences , wherein the plurality of test peptide sequences comprises at least 50 test peptide sequences, the sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in a cell, and (ii) at least 19 bait peptide sequences contained within a protein, the protein comprising a peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in a cell, wherein the organism and the subject are the same species, wherein the ratio of the at least one hit peptide sequence to the at least 19 bait peptide sequences in the plurality of test peptide sequences is 1:19, and 5% of the plurality of test peptide sequences are predicted to bind to an HLA protein expressed in a cell according to a machine learning HLA peptide presentation prediction model.

[0208] In some embodiments, the machine learning HLA peptide presentation prediction model is trained using training data comprising sequence information of training peptide sequences identified by mass spectrometry as being presented by HLA proteins expressed in training cells.

[0209] In some embodiments, one or more of 0.2% of the plurality of test peptide sequences predicted to be presented by the machine learning HLA peptide presentation prediction model have a probability of being presented by at least one of the one or more proteins encoded by a class II HLA allele of a cell of the subject that is greater than a threshold presentation prediction probability value.

[0210] In some embodiments, each of 0.2% of the plurality of test peptide sequences predicted to be presented by a machine learning HLA peptide presentation prediction model has a probability of being presented by at least one of the one or more proteins encoded by a class II HLA allele of a cell of the subject that is greater than a threshold presentation prediction probability value.

[0211] In some embodiments, the PPV is greater than the corresponding PPV in column 2 of Table 11 for proteins encoded by corresponding HLA alleles of Table 13. In some embodiments, the PPV is at least equal to the corresponding PPV in column 3 of Table 11 for proteins encoded by corresponding HLA alleles of Table 11.

[0212] In some embodiments, the PPV is greater than the corresponding PPV in column 2 of Table 12 for a protein encoded by an HLA class II allele.

[0213] In some embodiments, for the protein encoded by the corresponding HLA allele of Table 16, the PPV is at least equal to the corresponding PPV of column 2 of Table 16.

[0214] Provided herein is a method for preparing a personalized cancer therapy, the method comprising: identifying a peptide sequence, wherein the peptide sequence is associated with cancer, wherein the identifying comprises comparing a DNA, RNA, or protein sequence from a cancer cell of a subject with a DNA, RNA, or protein sequence from a normal cell of the subject; and inputting amino acid position information of the identified peptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the identified peptide sequence, each presentation prediction representing a peptide sequence presented by an HLA of a cell of the subject. The method of claim 1 wherein the machine learning HLA-peptide presentation prediction model comprises: a plurality of predictor variables determined at least based on training data; wherein the training data comprises: sequence information of peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between amino acid position information received as input and a likelihood of presentation generated as output based on the amino acid position information and the predictor variables; and selecting a subset of the identified peptide sequences for use in preparing personalized cancer treatment based on the set of presentation predictions; wherein the prediction model has a positive predictive value of at least 0.1 at a call rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0215] Provided herein is a method comprising training a machine learning HLA-peptide presentation prediction model, wherein the training comprises inputting, into the HLA-peptide presentation prediction model, using a computer processor, amino acid position information sequences of HLA-peptides isolated from one or more HLA-peptide complexes from cells expressing an HLA class II allele; the machine learning HLA-peptide presentation prediction model comprising: a plurality of predictor variables determined based at least on training data, the training data comprising: sequence information of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information of training peptides, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information received as input and a presentation likelihood generated as output based on the amino acid position information and the predictor variables.

[0216] In some embodiments, the presentation model has a positive predictive value of at least 0.25 at a call rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0217] In some embodiments, the presentation model has a positive predictive value of at least 0.4 at a call rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0218] In some embodiments, the presentation model has a positive predictive value of at least 0.6 at a call rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0219] In some embodiments, the mass spectrometry is monoallelic mass spectrometry.

[0220] In some embodiments, the peptide is presented by an HLA protein expressed in the cell via autophagy.

[0221] In some embodiments, the peptide is presented by an HLA protein expressed in a cell by phagocytosis.

[0222] In some embodiments, the quality of the training data is improved by using multiple quality metrics.

[0223] In some embodiments, the plurality of quality metrics includes common contaminant peptide removal, high scoring peak intensity, high score, and high mass accuracy.

[0224] In some embodiments, the scored peak intensity is at least 50%.

[0225] In some embodiments, the scored peak intensity is at least 60%.

[0226] In some embodiments, the score is at least 7.

[0227] In some embodiments, the mass accuracy is at most 5 ppm.

[0228] In some embodiments, the mass accuracy is at most 2 ppm.

[0229] In some embodiments, the backbone cleavage score is at least 5.

[0230] In some embodiments, the backbone cleavage score is at least 8.

[0231] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single immunoprecipitated HLA protein expressed in a cell.

[0232] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single exogenous HLA protein expressed in the cell.

[0233] In some embodiments, the peptide presented by an HLA protein expressed in a cell is a peptide presented by a single recombinant HLA protein expressed in a cell.

[0234] In some embodiments, the plurality of predictor variables includes a peptide-HLA affinity predictor variable.

[0235] In some embodiments, the plurality of predictor variables includes a source protein expression level predictor variable.

[0236] In some embodiments, the plurality of predictor variables includes a peptide cleavage predictor variable.

[0237] In some embodiments, the training peptide sequence information includes sequences of peptides presented by HLA proteins, including peptides identified by searching a database of enzyme-specific, unmodified peptides. In some embodiments, the peptides presented by HLA proteins include peptides identified by searching a de novo peptide sequencing tool.

[0238] In some embodiments, peptides presented by HLA proteins include peptides identified by searching a peptide database using a reverse database search strategy.

[0239] In some embodiments, the HLA proteins include HLA-DR and HLA-DP or HLA-DQ proteins. In some embodiments, the HLA proteins include HLA-DR proteins selected from HLA-DR and HLA-DP or HLA-DQ proteins. In some embodiments, the HLA proteins include HLA-DR proteins selected from the group consisting of HLA-DPB1*01:01 / HLA-DPA1*01:03, HLA-DPB1*02:01 / HLA-DPA1*01:03, HLA-DPB1*03:01 / HLA-DPA1*01:03, HLA-DPB1*04:01 / HLA-DPA1*01:03, HLA-DPB1*04:02 / HLA-DPA1*01:03, HLA-DPB1*06:01 / HLA-DPA1*01:03, HLA-DPB1*07:01 / HLA-DPA1*07:03, HLA-DPB1*08:01 / HLA-DPA1*08:03 A-DQB1*02:01 / HLA-DQA1*05:01, HLA-DQB1*02:02 / HLA-DQA1*02:01, HLA-DQB1*06:02 / HLA-DQA1*01:02, HLA-DQB1*06:04 / HLA- DQA1*01:02, HLA-DRB1*01:01, HLA-DRB1*01:02, HLA-DRB1*03:01, HLA-DRB1*03:02, HLA-DRB1*04:01, HLA-DRB1*04:02, HLA-DRB 1*04:03, HLA-DRB1*04:04, HLA-DRB1*04:05, HLA-DRB1*04:07, HLA-DRB1*07:01, HLA-DRB1*08:01, HLA-DRB1*08:02, HLA-DRB1* 08:03, HLA-DRB1*08:04, HLA-DRB1*09:01, HLA-DRB1*10:01, HLA-DRB1*11:01, HLA-DRB1*11:02, HLA-DRB1*11:04, HLA-DRB1*12 :01, HLA-DRB1*12:02, HLA-DRB1*13:01, HLA-DRB1*13:02, HLA-DRB1*13:03, HLA-DRB1*14:01, HLA-DRB1*15:01, HLA-DRB1*15:0 2. HLA-DRB1*15:03, HLA-DRB1*16:01, HLA-DRB3*01:01, HLA-DRB3*02:02, HLA-DRB3*03:01, HLA-DRB4*01:01 and HLA-DRB5*01:01).

[0240] In some embodiments, peptides presented by HLA proteins include peptides identified by comparing the MS / MS spectrum of the HLA-peptide to the MS / MS spectra of one or more HLA-peptides in a peptide database.

[0241] In some embodiments, the mutation is selected from the group consisting of a point mutation, a splice site mutation, a frameshift mutation, a readthrough mutation, and a gene fusion mutation.

[0242] In some embodiments, the peptide presented by the HLA protein is 15-40 amino acids in length.

[0243] In some embodiments, peptides presented by HLA proteins include peptides identified by: (a) isolating one or more HLA complexes from a cell line expressing a single HLA class II allele; (b) isolating one or more HLA-peptides from the one or more isolated HLA complexes; (c) obtaining MS / MS spectra of the one or more isolated HLA-peptides; and (d) obtaining peptide sequences corresponding to the MS / MS spectra of the one or more isolated HLA-peptides from a peptide database; wherein the sequences of the one or more isolated HLA-peptides are identified from the one or more sequences obtained in step (d).

[0244] In some embodiments, the personalized cancer therapy further comprises an adjuvant.

[0245] In some embodiments, the personalized cancer therapy further comprises an immune checkpoint inhibitor.

[0246] In some embodiments, the training data includes structured data, time series data, unstructured data, relational data, or any combination thereof.

[0247] In some embodiments, the unstructured data includes image data.

[0248] In some embodiments, the relationship data includes data from a client system, an enterprise system, an operating system, a website, a web-accessible application programming interface (API), or any combination thereof.

[0249] In some embodiments, the training data is uploaded to a cloud-based database.

[0250] In some embodiments, the training is performed using a convolutional neural network.

[0251] In some embodiments, the convolutional neural network comprises at least two convolutional layers.

[0252] In some embodiments, the convolutional neural network (CNN) includes at least one batch normalization step.

[0253] In some embodiments, the convolutional neural network includes at least one spatial dropout step.

[0254] In some embodiments, the convolutional neural network includes at least one global max pooling step.

[0255] In some embodiments, the convolutional neural network comprises at least one dense layer.

[0256] In some embodiments, identifying the peptide sequence comprises identifying a peptide sequence having a mutation that is expressed in a cancer cell of the subject.

[0257] In some embodiments, identifying a peptide sequence comprises identifying a peptide sequence that is not expressed in normal cells of the subject.

[0258] In some embodiments, identifying a peptide sequence comprises identifying an overexpressed peptide sequence.

[0259] In some embodiments, identifying a peptide sequence comprises identifying a viral peptide sequence. In one aspect, the present invention provides a method for identifying an HLA class II specific peptide for immunotherapy that is specific to a subject, the method comprising: identifying a candidate peptide comprising an epitope; inputting amino acid information of a plurality of peptide sequences (each comprising an epitope) into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of HLA presentation predictions for the peptide sequence to immune cells, each presentation prediction representing a probability that one or more proteins encoded by an HLA class II allele of a cell of the subject presents a given peptide sequence comprising the epitope; wherein the prediction model has a positive predictive value of at least 0.1 at a determination rate of at least 0.1%, 0.1%-50%, or at most 50%, selecting a protein from the one or more proteins encoded by an HLA class II allele of a cell of the subject, the protein being predicted by the prediction model to bind to the candidate peptide, wherein the probability that the protein presents the candidate peptide to an immune cell is greater than a threshold presentation prediction probability value; contacting the candidate peptide with a protein encoded by an HLA class II allele such that the candidate peptide competes with the HLA class II allele for the present peptide sequence to form a peptide sequence that binds to the candidate peptide; a placeholder peptide associated with a protein encoded by an HLA class II allele; and identifying the candidate peptide as a peptide for immunotherapy that is specific to the protein encoded by the HLA class II allele based on whether the candidate peptide replaces the placeholder peptide.

[0260] In some embodiments, the immunotherapy is cancer immunotherapy.

[0261] In some embodiments, identifying comprises comparing a DNA, RNA, or protein sequence from a cancer cell of the subject to a DNA, RNA, or protein sequence from a normal cell of the subject.In some embodiments, the epitope is a cancer-specific epitope.

[0262] In some embodiments, the at least one protein encoded by the HLA class II allele comprises at least the α1 subunit and the β1 subunit of an HLA protein or fragment thereof present in dimer form. In some embodiments, the placeholder peptide is a CLIP peptide. In some embodiments, the placeholder peptide is a CMV peptide. In some embodiments, the method further comprises measuring the IC50 of the placeholder peptide being replaced by the target peptide. In some embodiments, the IC50 of the placeholder peptide being replaced by the target peptide is less than 500 nM. In some embodiments, at least one protein from one or more proteins encoded by the HLA class II alleles of the subject's cells is an HLA class II tetramer or multimer. In some embodiments, the target peptide is further identified by mass spectrometry. In some embodiments, the at least one protein encoded by the HLA class II alleles of the subject's cells is a recombinant protein. In some embodiments, the at least one protein encoded by the HLA class II alleles of the subject's cells is expressed in a eukaryotic cell.

[0263] In one aspect, provided herein is an assay method for validating the specificity of a candidate peptide binding to an HLA class II protein, the method comprising: expressing in a eukaryotic cell a polynucleic acid construct comprising a nucleic acid sequence encoding an HLA class II protein comprising an α chain and a β chain, or a portion thereof, capable of binding to a peptide comprising an MHC-II binding epitope, and wherein the expressed HLA class II protein or portion thereof remains associated with a placeholder peptide; isolating the HLA class II protein or portion thereof expressed in the eukaryotic cell; and performing a peptide exchange assay by: (a) adding increasing amounts of a candidate peptide to determine whether the candidate peptide displaces the placeholder peptide associated with the HLA class II protein or portion thereof; and (b) calculating the IC50 of the displacement reaction to determine the affinity of the candidate peptide for the HLA class II protein or portion thereof relative to the placeholder peptide, thereby validating the specificity of the candidate peptide binding to the HLA class II protein.

[0264] In some embodiments, the identity of the peptide is known. In some embodiments, the identity of the peptide is unknown. In some embodiments, the identity of the peptide is determined by mass spectrometry.

[0265] In some embodiments, the peptide exchange assay comprises detecting a peptide fluorescent probe or tag. In some embodiments, the placeholder peptide is a CLIP peptide.

[0266] In some embodiments, the polynucleic acid construct comprises an expression vector, which further comprises one or more of the following: a promoter, a linker, one or more protease cleavage sites, a secretion signal, a dimerization factor, a ribosomal skipping sequence, one or more tags for purification and / or detection.

[0267] In one aspect, provided herein is a method for determining the immunogenicity of an MHC class II-binding peptide, the method comprising: selecting a protein encoded by an HLA class II allele that is predicted to bind to a peptide by a machine learning HLA-peptide presentation prediction model; wherein the prediction model has a positive predictive value of at least 0.1 at a call rate of at least 0.1%, 0.1%-50%, or at most 50%, and wherein the probability that the protein presents the identified peptide sequence is greater than a threshold presentation prediction probability value; contacting the peptide with the selected protein encoded by the HLA class II allele such that the peptide competes for and displaces a placeholder peptide associated with the selected protein encoded by the HLA class II allele, thereby forming a complex comprising the HLA class II protein and the identified peptide; contacting the complex of the HLA class II protein and the identified peptide with CD4+ T cells, and determining one or more activation parameters of the CD4+ T cells selected from the group consisting of cytokine induction, chemokine induction, and expression of cell surface markers.

[0268] In some embodiments, the HLA class II allele is a tetramer or multimer. In some embodiments, the cytokine is IL-2. In some embodiments, the cytokine is IFN-γ.

[0269] In one aspect, the present invention provides a method for inducing CD4+ T cell activation in a subject for cancer immunotherapy, the method comprising: identifying a peptide sequence associated with cancer and comprising a cancer mutation, wherein the identification comprises comparing a DNA, RNA or protein sequence from a cancer cell of the subject with a DNA, RNA or protein sequence from a normal cell of the subject; selecting a protein encoded by an HLA class II allele, the protein being normally expressed by the cells of the subject and predicted to bind to the peptide by a machine learning HLA-peptide presentation prediction model; wherein the prediction model has a positive predictive value of at least 0.1 at a call rate of at least 0.1%, 0.1%-50% or at most 50%, and wherein the probability of the protein presenting the identified peptide sequence is greater than a threshold presentation prediction probability value; contacting the identified peptide with the selected protein encoded by the HLA class II allele to verify whether the identified peptide competes with the selected protein encoded by the HLA class II allele. The method comprises the steps of: identifying a placeholder peptide associated with a selected protein encoded by a class II allele, thereby replacing the placeholder peptide with an IC50 value of less than 500 nM; purifying the identified peptide; and administering an effective amount of the identified peptide to the subject.

[0270] In one aspect, the present invention provides a method for preparing HLA class II tetramers or multimers, the method comprising: expressing in a eukaryotic cell a vector comprising a nucleic acid sequence encoding the α chain and β chain of an HLA protein, a linker, one or more protease cleavage sites, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, such that each HLA protein α1 and β1 heterodimer is secreted in a dimerized state, wherein the heterodimer is associated with a placeholder peptide, purifying the secreted heterodimers from the cell culture medium, verifying the peptide binding activity using a peptide exchange assay, adding streptavidin to conjugate the heterodimers into tetramers, purifying the tetramers, and having a yield greater than 1 mg / L.

[0271] In some embodiments, the vector comprises a CMV promoter. In some embodiments, the vector comprises a sequence encoding a placeholder peptide connected to the β1 chain via a cleavable site. In some embodiments, the peptide exchange assay involves pre-cleaving the placeholder peptide from the β chain. In some embodiments, the cleavable site is a thrombin cleavage site. In some embodiments, the peptide exchange assay is a FRET assay. In some embodiments, the purification is performed by any of the following: column chromatography, batch chromatography, ion exchange chromatography, size exclusion chromatography, affinity chromatography, or LC-MS.

[0272] In one aspect, provided herein is a composition comprising an HLA class II tetramer comprising an HLA-DR, HLA-DP, or HLA-DQ heterodimer, each heterodimer comprising an α and a β chain, purified and present at a concentration greater than 0.25 mg / L. In some embodiments, the HLA class II tetramer comprises a heterodimer pair selected from the group consisting of: a protein selected from HLA-DR and an HLA-DP or HLA-DQ protein.In some embodiments, the HLA protein is selected from: HLA-DPB1*01:01 / HLA-DPA1*01:03, HLA-DPB1*02:01 / HLA-DPA1*01:03, HLA-DPB1*03:01 / HLA-DPA1*01:03, HLA-DPB1*04:01 / HLA-DPA1*01:03, HLA-DPB1*04:02 / HLA-DPA1*01:03, HLA-DPB1*06:01 / HLA-DPA1*01:03, HLA-DQB1*02:01 / HLA-DQA1*05:01, HLA-DQB1*02:02 / HLA-DQA1*02:01, HLA-DQB1*06:02 / HLA-DQA1*01:02, HLA-DQB1*06:04 / HLA-DQA1*01:02, HLA-DRB1*01:01, HLA-DRB1*01:02, HLA-DRB1*03:01, HLA-DRB1*03:02, HLA-DRB1*04:01, HLA-DRB1*04:02, HLA-DRB1*04:03, HLA-DRB1*04:04, HLA-DRB1*04:05, HLA-DRB1*04:07, HLA-DRB1*07:01, HLA-DRB1*08:01, HLA-DRB1*08:02, HLA-DRB1*08:03, HLA-DRB1*08:04, HLA-DRB1*09:01, HLA-DRB1*10:01, HLA-DRB1*11:01, HLA-DRB1*11:02, HLA-DRB1*11:04, HLA-DRB1*12:01, HLA-DRB1*12:02, HLA-DRB1*13:01, HLA-DRB1*13:02, HLA-DRB1*13:03, HLA-DRB1*14:01, HLA-DRB1*15:01, HLA-DRB1*15:02, HLA-DRB1*15:03, HLA-DRB1*16:01, HLA-DRB3*01:01, HLA-DRB3*02:02, HLA-DRB3*03:01, HLA-DRB4*01:01, HLA-DRB5*01:01).

[0273] In some embodiments, the heterodimer pair is expressed in a eukaryotic cell. In some embodiments, the heterodimer pair is encoded by a vector. In some embodiments, the vector comprises: a nucleic acid sequence encoding the α chain and β chain of the HLA protein, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, such that each HLA protein α1 and β1 heterodimer is secreted in a dimerized state, wherein the secreted heterodimer is associated with a placeholder peptide. In some embodiments, the vector comprises: a nucleic acid sequence encoding the α chain and β chain of the HLA protein, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, such that each HLA protein α1 and β1 heterodimer is secreted in a dimerized state, wherein the secreted heterodimer is associated with a placeholder peptide.

[0274] In some embodiments, HLA class II heterodimers are secreted from eukaryotic cells into the cell culture medium and purified by any of the following methods: column or batch chromatography, ion exchange chromatography, size exclusion chromatography, affinity chromatography, or LC-MS.

[0275] In one aspect, provided herein is a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that an epitope sequence of a given peptide sequence is presented by one or more proteins encoded by an HLA class I or class II allele of a cell of the subject; wherein the machine learning HLA-peptide presentation prediction model comprises: a plurality of predictor variables determined based at least on training data; wherein the training data comprises: sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables; (b) determining or predicting, based on the set of presentation predictions, that each of the peptide sequences of the polypeptide sequence is not immunogenic to the subject; and (c) administering a composition comprising the drug to the subject.

[0276] In one aspect, provided herein is a method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: (a) inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that an epitope sequence of a given peptide sequence is presented by one or more proteins encoded by an HLA class I or class II allele of a cell of the subject; wherein the machine learning HLA-peptide presentation prediction model comprises: a plurality of predictor variables determined based at least on training data; wherein the training data comprises: sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables; and (b) determining or predicting, based on the set of presentation predictions, that at least one of the peptide sequences of the polypeptide sequence is immunogenic to the subject.

[0277] In one embodiment, the method further comprises deciding not to administer the drug to the subject.

[0278] In one embodiment, the medicament comprises an antibody or a binding fragment thereof.

[0279] In one embodiment, the peptide sequence of the polypeptide sequence comprises each contiguous peptide sequence of the polypeptide sequence, which has a length of 8, 9, 10, 11 or 12 amino acids, and wherein the protein encoded by an HLA class I or class II allele of a cell of the subject is a protein encoded by an HLA class I allele of a cell of the subject.

[0280] In one embodiment, the peptide sequence of the polypeptide sequence comprises each contiguous peptide sequence of the polypeptide sequence, which has a length of 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or 25 amino acids, and wherein the protein encoded by an HLA class I or class II allele of a cell of the subject is a protein encoded by an MHC class II allele of a cell of the subject.

[0281] In one aspect, provided herein is a method of treating a subject having an autoimmune disease or condition, comprising: (a) identifying or predicting an epitope of an expressed protein presented by HLA class I or class II of cells of the subject, wherein a complex comprising the identified or predicted epitope and HLA class I or class II is targeted by CD8 or CD4 T cells of the subject; (b) identifying a T cell receptor (TCR) that binds to the complex; (c) expressing the TCR in regulatory T cells or allogeneic regulatory T cells from the subject; and (d) administering regulatory T cells expressing the TCR to the subject.

[0282] In one embodiment, the autoimmune disease or condition is diabetes.

[0283] In one embodiment, the cells are pancreatic islet cells.

[0284] In one aspect, provided herein is a method of treating a subject having an autoimmune disease or condition, comprising administering to the subject a regulatory T cell expressing a T cell receptor (TCR) that binds to a complex comprising: (i) an epitope of an expressed protein identified or predicted to be presented by HLA class I or class II of a cell of the subject, and (ii) HLA class I or class II, wherein the complex is targeted by a CD8 or CD4 T cell of the subject.

[0285] Other aspects and advantages of the present disclosure will become apparent to those skilled in the art based on the following detailed description, which merely shows and describes illustrative embodiments of the present disclosure. It will be appreciated that the present disclosure is capable of other different embodiments and that its several details are capable of modifications in various obvious respects, all without departing from the present disclosure. Accordingly, the drawings and description are to be regarded as illustrative in nature and not restrictive.

[0286] MAPTAC TM Can be used for high-throughput peptide binding assays where MAPTAC is used at different time points and under different conditions (e.g., heating at 37°C) TM After construct isolation, peptides binding to HLA class II were measured to obtain the sequences of peptide populations with different stabilities using LC-MS / MS.

[0287] In one aspect, provided herein is a method for treating cancer in a subject, the method comprising: identifying a peptide sequence, wherein the peptide sequence is associated with the cancer, wherein the identifying comprises comparing a DNA, RNA, or protein sequence from a cancer cell of the subject with a DNA, RNA, or protein sequence from a normal cell of the subject; inputting amino acid information of the identified peptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the identified peptide sequence, each presentation prediction representing a peptide sequence presented by an HLA of a cell of the subject; The method of claim 1 , wherein the machine learning HLA-peptide presentation prediction model comprises: a plurality of predictor variables determined based at least on training data, wherein the training data comprises: sequence information of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between amino acid position information received as input and a likelihood of presentation generated as output based on the amino acid position information and the predictor variables; and selecting a subset of the identified peptide sequences for preparation of personalized cancer treatment based on the set of presentation predictions; and administering to the subject a composition comprising one or more of the peptides, wherein the predictive model has a positive predictive value of at least 0.1 at a call rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0288] In some embodiments, the machine learning HLA-peptide presentation prediction model comprises sequence information of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry following reverse phase offline fractionation.

[0289] In some embodiments, the prediction model exhibits a 1.1x to 100x improvement compared to NetMHC II pan. In some embodiments, the prediction model exhibits a 1.1, 2, 3, 4, 5, 6, 7, 7.4, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 18, 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 41, 42, 43, 44, 45, 50, 55, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 8, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100-fold or more improvement.

[0290] Incorporation by reference

[0291] All publications, patents, and patent applications mentioned in this specification are incorporated herein by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference. In the event of a conflict between the disclosure contained in this specification and the publications and patents or patent applications incorporated by reference, the present specification is intended to supersede and / or take precedence over any such conflicting material. BRIEF DESCRIPTION OF THE DRAWINGS

[0292] The novel features of the present invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description of illustrative embodiments utilizing the principles of the present invention and the accompanying drawings (also referred to herein as "Figures"), in which:

[0293] Figure 1A is a diagram showing a peptide docked on an MHC class I protein.

[0294] Figure 1B Depicted are exemplary diagrams representing peptides docked on MHC class II proteins.

[0295] Figure 2 An exemplary experimental method for generating monoallelic HLA class II binding peptide data is described. HLA class II peptides are introduced into any cell, including cells that do not express HLA class II, to express a specific HLA class II allele in the cell. A population of genetically engineered HLA-expressing cells is harvested, lysed, their HLA-peptide complexes labeled (e.g., biotinylated), and immunopurified (e.g., using a biotin-streptavidin interaction). HLA-associated peptides specific for a single HLA can be eluted from their labeled (e.g., biotinylated) complexes and evaluated (e.g., by sequencing using high-resolution LC-MS / MS).

[0296] Figure 3 Depicted are exemplary sequence logo representations of HLA class II-DRB1*11:01 associated peptides in: NeonBAP, Expi293 cell line; Neon BAP, A375 cell line; IEDB, affinity <50 nM; and full HLA class II antibody, homozygous LCL. Figure 3 Examples of MS-derived motifs are shown to match known patterns and show consistency across transfected cell lines.

[0297] Figure 4 is an exemplary depiction of HLA class II binding predictor performance. Figure 4is a bar graph showing the performance of the binding predictor (neonmhc2) and NetMHC IIpan applied to a validation dataset consisting of observed mass spectrometry peptides and decoy peptides generated by randomly shuffling hit peptides at a ratio of 1:19 (hit:bait). For the NEON binding predictor, neonmhc2, a separate model was constructed for each MHC II allele shown. The height of the bar shows the positive predictive value (PPV), which is defined as the fraction of predicted binders in the validation set that are actually hit peptides. Alleles are ranked by the model's performance in predicting that allele.

[0298] Figure 5 Depicted are exemplary effects of score peak intensity (SPI) thresholds on binding predictor validation. Figure 5 The performance of the HLA class II binding predictor when trained / validated on sets of peptides with different scored peak intensity (SPI) cutoffs is shown. For each allele-specific model trained, the performance of the model is shown in three settings: training and evaluating the dataset using observed MS hit peptides with 70 SPI or greater, training on peptides with 50 SPI or greater and validating on peptides with 70 SPI or greater, and training and validating on peptides with 50 SPI or greater.

[0299] Figure 6 Depicted is an exemplary bar graph showing representative data from the number of observed peptides using a score peak intensity (SPI) cutoff of greater than or equal to 70 for allele profiling by LC-MS / MS. Each bar represents the total number of peptides observed for the allele. Data for 35 HLA-DR alleles were collected. The collected data for the 35 HLA-DR alleles had >95% HLA-DR population coverage (US allele frequency).

[0300] Figure 7A The PPV of the model when applied to the test partition of the data for the indicated HLA class II alleles is shown. The decoy peptides used were scrambled sequences of the positive (hit) peptide sequences, with a hit to decoy ratio of 1:19. The PPV was determined by identifying the top 5% scoring peptides in the test partition and determining the fraction of them that were positive for binding to the protein encoded by the corresponding HLA class II allele.

[0301] Figure 7B Depicted are exemplary prediction performance as a function of training set size (curves obtained by artificially downsampling the training set). Figure 7B showed that, in general, for the collected 35 HLA-DR alleles, the value of PPV increased when the training set size increased.

[0302] Figure 8 An exemplary diagram is depicted demonstrating that treatment-related variables can further improve predictions. Distinguishing peptides observed by MS can distinguish random sequences selected from the protein-coding exon set. On the training data partition, logistic regression can be fit to predict HLA class II presentation using binding strength (NetMHCIIpan or Neon predictors) and treatment features (RNA-Seq expression and derived gene-level bias terms). In a separate evaluation partition, exon positions that overlap with MHC II peptides observed by MS ("hits") can be scored together with random exon positions not observed in MS (1:499 ratio). The top 0.2% (1 / 500) can be called positive, and the positive predictive value can assess this threshold.

[0303] Figure 9 An exemplary neural network architecture is depicted. Input peptides are represented as 20-mers, with shorter peptides padded with "missing" characters. Each peptide has a 31-dimensional embedding, so the input to the neural network is a 20x31 matrix. Before being processed by the neural network, the 20x31 matrix is ​​feature normalized according to the mean and standard deviation of the eigenvalues ​​in the training set. The first convolutional layer has a kernel of 9 amino acids and 50 filters (also called channels), with a rectified linear unit (ReLU) activation function. This is followed by batch normalization, then spatial dropout with a dropout rate of 20%. This is followed by another convolutional layer with a kernel of 3 amino acids and 20 filters, with a ReLU activation function, followed by batch normalization and spatial dropout with a dropout rate of 20%. Global max pooling is then applied to take the neuron with the maximum activation in each of the 20 filters; these 20 values ​​are then fed into a fully connected (dense) layer with a single neuron using a sigmoid activation function. The output of this layer is considered the bind / non-bind prediction. L2 regularization is applied to the weights of the first convolutional layer, the second convolutional layer, and the dense layer with weights of 0.05, 0.1, and 0.01, respectively. The other models used vary the number of convolutional layers and the kernel size of each layer.

[0304] Figure 10 Depicted are exemplary computer control systems programmed or otherwise configured to implement the methods provided herein.

[0305] Figure 11A Depicts MAPTAC TM Illustrative overview of the experimental workflow.

[0306] Figure 11B Exemplary per-allele peptide counts pooled across replicates are depicted.

[0307] Figure 11C Depicted by MAPTACTM Exemplary peptide length distributions for the analyzed HLA class I and HLA class II alleles.

[0308] Figure 11D Describes the MAPTAC TM and exemplary per-residue cysteine ​​frequencies observed with IEDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01, and DRB1*11:01), the human proteome, and multi-allelic MS data from a previous publication.

[0309] Figure 12A Depicted are the Caucasian frequencies of HLA-DR, -DP, and -DQ alleles present in >1% of individuals, as well as counts of peptides from the indicated sources measured as strong binders (<50 nM).

[0310] Figure 12B Depicted are exemplary length distributions of IEDB peptides with associated HLA class II affinity measurements.

[0311] Figure 12C Depicted are exemplary Western blots of (1) Expi293, (2) HeLa, and (3) A375 cell lines transfected with two HLA class I and two HLA class II alleles, HLA-A*02:01, HLA-B*45:01, HLA-DRB1*01:01, and HLA-DRB1*11:01, respectively. The membrane was blotted with an anti-biotin ligase epitope tag to display a biotin-accepting peptide (BAP) and anti-β-tubulin as a loading control. Lanes correspond to the expression of the 5′-HLA-DAP-1 protein in the presence of 1% β-tubulin. TM The following fractions were collected during the protocol: lane 1, input; lane 2, biotinylated input; lane 3, input after pull-down.

[0312] Figure 12D Describes the MAPTAC TM and exemplary amino acid frequencies per residue observed with IEDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01, and DRB1*11:01), the human proteome, and multi-allelic MS data from a previous publication.

[0313] Figure 12E Depicted are the Caucasian frequencies of HLA-DR, -DP, and -DQ alleles present in >1% of individuals, and counts of peptides from the indicated sources that were measured as strong binders (<50 nM). Figure 12AAdditional data is available from tools.iedb.org / main / datasets / .

[0314] Figure 12F Describes the MAPTAC TM (reduced and alkylated), MAPTAC TM Exemplary amino acid frequencies per residue observed for (no treatment) and IEDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01, and DRB1*11:01), the human proteome, and multi-allelic MS data from a previous publication.

[0315] Figure 13 Depicts each MAPTAC TM Figure 1. Example representation of core binding sequence logos for MHC II alleles from the IEDB. The sequence logo is a graphical representation where the height of each amino acid is proportional to its frequency of occurrence in peptides bound to the MHC protein encoded by the allele. Positions with the lowest entropy are indicated by color, where the color corresponds to the amino acid property. Peptides are from the dataset shown and aligned according to a CNN-based predictor (Methods). The logo represents all peptides, including those that do not closely match the overall motif (e.g., no peptides are isolated in a "junk" cluster).

[0316] Figure 14A Depicted are exemplary sequence logos of HLA-A*02:01 binding peptides (ligands) analyzed in two different cell lines (A375 and expi293) using different HLA-ligand profiling techniques, including binding assays, stability assays, soluble HLA (sHLA) mass spectrometry, monoallelic mass spectrometry, and MAPTAC TM .

[0317] Figure 14B Depicts MAPTACs showing 0, 1, 2, 3, and 4 heuristically defined anchors TM Exemplary fractions of peptides.

[0318] Figure 14C Describes the MAPTAC TM Exemplary distribution of observed peptides (20 peptides per allele, each nested group with SPI>70 and size>=2) and NetMHCIIpan predicted binding affinities for length-matched baits sampled from the proteome.

[0319] Figure 15ADepicts an exemplary architecture of a convolutional neural network (CNN) trained to discriminate monoallelic MHC peptides from scrambled length-matched decoys. The schematic shows the use of amino acid feature embeddings, two convolutional layers with different filter sizes, and global max pooling as input to the final logistic output node.

[0320] Figure 15B Figure 2 is an example result showing the Kendall Tau statistic for the correlation of measured IEDB affinities with binding predictions from neoNMHC2 or NetMHCIIpan. Peptides evaluated included only those published to IEDB one year after NetMHCIIpan release.

[0321] Figure 16 is an exemplary depiction of the performance of neonmhc2 as a function of the training dataset size.

[0322] Figure 17A Delineation of MAPTACs incorporated into pan-DR and pan-class II MHC MS datasets TM Example cluster assignments of peptides (20 per allele). The dataset was deconvoluted using GibbsCluster. Each box represents a MAPTAC TM Peptides. The color of the box indicates which cluster it was assigned to, and the gray bar indicates which allele the peptide actually came from. The total number of clusters in the Gibbs clustering scheme (right) was selected using the mutual information (MI) metric. The MI score also determines how the samples are ranked; samples with a high MI score appear at the top.

[0323] Figure 17B Depicted are exemplary core binding sequence identities for multi-allelic MS data deconvoluted by GibbsCluster. Each set of peptides corresponds to a MAPTAC TM The cluster with the best match of the spike.

[0324] Figure 17C Describes the use of MAPTAC TM Data or deconvoluted multi-allelic data predicted retention MAPTAC TM For each allele, the larger of the two data sources (usually MAPTAC TM ) are downsampled so that the predictors are based on the same number of training examples. NetMHCIIpan performance is shown as an additional comparison.

[0325] Figure 17D Depicted are exemplary core binding sequence signatures derived from multi-allelic MS data from the indicated sources.

[0326] Figure 18A Depicted are exemplary graphs of the fraction of peptide to source gene expression (transcripts per million (TPM)) for MS observed peptides and random proteomic baits (data replotted from Schuster et al., 2017).

[0327] Figure 18B Depicted are exemplary observed and expected numbers of class II peptides per gene as determined by joint analysis of colorectal cancer, melanoma, and ovarian cancer datasets ( et al., 2018 and Schuster et al., 2017). Expected counts are calculated by multiplying gene length by expression level. Expected and observed counts for the relevant samples were summed. Genes known to be present in plasma are labeled according to their concentration (inset).

[0328] Figure 18C Depicted are exemplary distributions of enrichment scores (ratios of observed to expected observations, as Figure 18B ).

[0329] Figure 18D Depicted is an exemplary distribution of enrichment scores according to the location of each source gene. Source gene locations were determined using Uniprot (uniprot_sprot.dat).

[0330] Figure 18E Depicted are exemplary data representing the expected versus observed frequency of the fraction of the total number of peptides with MHC-II binding affinity separated based on their cellular localization properties.

[0331] Figure 18F Depicts exemplary representative data of the relative consistency of the observed peptides for two different gene expression profiles. For each sample, the peptide counts at the gene level were modeled as a linear combination of bulk tumor gene expression and professional APC (macrophage) gene expression profiles. The ratio of the coefficients determines the relative consistency of each expression profile with the peptide repertoire. The error bars correspond to the 95% confidence intervals calculated by bootstrap resampling.

[0332] Figure 19A Depicted are exemplary representative data of HLA-DRB1 expression levels from five example studies. Each spot represents expression averaged across cells for an individual cell type in an individual patient.

[0333] Figure 19BDepicts exemplary representative data of HLA-DRB1 expression derived from tumors and stroma of RNA-Seq input from TCGA patients. Horizontal bars correspond to individual patients and are grouped by tumor type. Patients were included if they had mutations in HLA class II pathway genes (CIITA, CD74, or CTSSS) as determined by DNA-based mutation determination. For each patient, the fraction of HLA-DRB1 expression attributed to the tumor was estimated as min(1,2f), where f is the fraction of RNA-Seq reads in CIITA, CD74, or CTSS that showed mutations.

[0334] Figure 19C Depicted are exemplary representative data from additional single-cell RNA-Seq studies including biopsies before and after checkpoint blockade immunotherapy.

[0335] Figure 20 Depicted are exemplary representative experimental data evaluating the overall performance of predictions on native donor tissue.

[0336] Figure 21A Depicted are exemplary representative data showing that the integrated presentation model predicts cellular HLA class II ligandomes. It represents the PPV of the pan-DR dataset at a hit:bait ratio of 1:499 (also in Figure 30B and Figure 32E The predictor uses a combination of predictions (NetMHCIIpan or neonmhc2) and optionally gene expression, gene preference (based on Figure 32A ) and overlapped with previously observed HLA-DQ peptides. For each candidate peptide, the binding score was calculated as the maximum value among the HLA-DR alleles in the sample genotype.

[0337] Figure 21B Depicted are exemplary representative data showing the predictive performance of tumor-derived peptides identified using SILAC that were presented by dendritic cells (analyzed from cell lysates) using the same Figure 21A Identical hit:decoy ratios and performance metrics with and without treatment features.

[0338] Figure 21C Depicted are exemplary expression and gene preference scores of heavily labeled peptides (red spots, plotted against K562 expression) observed in UV treatment experiments compared to lightly labeled peptides (grey spots, plotted against DC expression).

[0339] Figure 21D Depicted is an example diagram representing the overlap of heavily labeled peptide-derived genes from lysate and UV treatment experiments. Gene names are colored by functional category.

[0340] Figure 22A Depicted is an exemplary flow diagram representing the assay protocol disclosed herein to validate HLA class II driven CD4+ T cells and T cell responses.

[0341] Figure 22B Depicted are an exemplary HLA protein dimer construct design for a peptide exchange assay (top panel) and a schematic of an exemplary assay workflow (bottom panel).

[0342] Figure 23 Depicted is an exemplary diagram of an exemplary vector design for MHC-II expression used for screening of novel binding peptides, and representation of the expressed protein product.

[0343] Figure 24 An exemplary flow chart for transfection, purification, and cleavage of the placeholder peptide from the beta chain is depicted.

[0344] Figure 25A Depicted are exemplary diagrams showing vectors encoding CLIP peptides that are associated with increased secretion of expressed MHC-II peptides.

[0345] Figure 25B Depicted are exemplary diagrams having shorter and longer forms of nucleic acids encoding CLIPO and CLIP1, respectively.

[0346] Figure 25C Depicted are exemplary representative results of Coomassie gel analysis of α and β chains with and without a longer clip.

[0347] Figure 26A An exemplary diagram of a TR-FRET assay is depicted.

[0348] Figure 26B Depicted are exemplary representative polarization data for an HLA class II peptide binding assay using a fluorescence resonance energy transfer (FRET) assay using specific peptides.

[0349] Figure 26C Depicted are exemplary representative polarization data for an HLA class II peptide binding assay using a fluorescence resonance energy transfer (FRET) assay using specific peptides.

[0350] Figure 26D Depicted are exemplary percentage substitutions of MHC-construct binding peptides calculated from fluorescence increases.

[0351] Figure 26E Depicted are exemplary percentage substitutions of MHC-construct binding peptides calculated from fluorescence increases.

[0352] Figure 26FDepicted is an exemplary peptide exchange obtained using an assay employing differential scanning fluorimetry (DSF). The diagram illustrates an exemplary mechanism for detecting the dissociation of a peptide from MHC class II using heat, which also dissociates MHC class II heterodimers, resulting in binding of the fluorophore and high fluorescence. Also depicted is an exemplary schematic diagram of a placeholder peptide displaced by an epitope peptide. Also depicted is an exemplary melting curve plotted as a function of temperature.

[0353] Figure 26G Depicted are exemplary soluble HLA-DM constructs and their use for MHC class II peptide exchange. The depicted construct comprises a CMV promoter, the coding sequence for the HLA-DM β chain, and the coding sequence for the HLA-DM α chain downstream of the secretory sequence (leader), along with a BAP sequence 3' to the β chain coding sequence and a His tag 3' to the α chain coding sequence. The two chains are separated by an intermediate ribosomal skipping sequence. The construct is expressed in Expi-CHO cells, and the protein secreted into the culture medium is purified.

[0354] Figure 26H Exemplary size exclusion chromatography data for peptide exchange using HLA-sDM are shown.

[0355] Figure 27A An exemplary diagram of an exemplary DRB tetramer repertoire construction is depicted.

[0356] Figure 27B An exemplary diagram of an exemplary class II tetramer repertoire construction is depicted.

[0357] Figure 27C An exemplary diagram depicting a summary of DRB tetramer repertoire coverage of DRB1 alleles for peptide exchange is shown.

[0358] Figure 27D Depicted are exemplary coverage generated by human MHC class II alleles.

[0359] Figure 27E Shown are exemplary results of tetramer staining from samples induced with either influenza epitopes (anamnestic response) or HIV epitopes (primary response).

[0360] Figure 28A Depicted is an exemplary diagram of a method for assessing HLA class II restriction of peptides by a fluorescence polarization assay that enables a screening method to rapidly identify the allelic restriction of epitope peptides. Figure 28A The assay principle depicted in allows for affinity measurements and unambiguous measurement of peptide exchange.

[0361] Figure 28BAn exemplary summary of multiple assay conditions explored in a fluorescence polarization assay using DRB1*01:01 is depicted (top panel). Also depicted are soluble MHC class II alleles and full-length MHC class II alleles with transmembrane domains in detergent micelles (bottom panel), both constructed with placeholder peptides with cleavable linkers for use in the assay.

[0362] Figure 28C Describes the previous study Figure 28B The following figure shows an exemplary illustration of the determination of full-length and soluble alleles. In short, both full-length and soluble alleles are expressed in cells. The membrane-bound full-length allele form is collected by permeabilizing the membrane, while the secreted form is collected from the cell supernatant. The harvested class II HLA allele proteins are purified by passing through nickel (Ni 2+ ) column for purification.

[0363] Figure 28D Depicted are exemplary data showing that the purification method does not affect the potency of the peptide. Shown on the left are full-length HLA-DR1 and Ni from purification using L243. 2+ Average IC50 values ​​from experiments with purified full-length HLA-DR1.

[0364] Figure 28E Depicted are exemplary data showing that the choice of soluble form (sDR1) or full-length form (fDR1) does not affect peptide potency. Shown on the left are mean IC50 values ​​from experiments using sDR1 form or fDR1. FP, fluorescence polarization.

[0365] Figure 28F Depicted are exemplary graphical views of exemplary evaluation of neoNMC2- and NetMHCIIpan-predicted peptides in binding assays and identification of discordant peptides.

[0366] Figure 28G Depicted are exemplary fluorescence polarization binding screening data for evaluating neoNMHC2-predicted peptides; displayed as a heat map with the percent inhibition of probe binding indicated for each concentration of the peptide used. Green depicts good binding, proportional to the color intensity. Yellow depicts moderate binding, and red depicts poor binding, also indicated by the corresponding percent inhibition value.

[0367] Figure 28H Depicted is a summary of the evaluation of NeoMHC2-predicted peptides in exemplary binding assays.

[0368] Figure 29 Depicted are the average MAPTAC for each HLA allele. TM Exemplary average peptide counts from experimental replicates (50 million cells).

[0369] Figures 30A-30C Delineation of HLA class II MAPTAC TM Exemplary binding core analysis of allele + / - HLA-DM and multi-allelic deconvolution fidelity. Figure 30A Depicts the MAPTAC TM Example sequence logos for a representative HLA-DR, -DQ, and -DP allele, with and without HLA-DM co-transfection (expi293 cell line) and IEDB (where the height of each amino acid is proportional to its frequency). Amino acids with a frequency greater than 10% are colored according to their chemical nature; all others are shown in gray. Peptides were aligned using the GibbsCluster tool (Supplementary Methods), and the logos represent all peptides, including those that do not closely match the overall motif (e.g., no peptides are isolated in a "junk" cluster). Figure 30B Depicting MAPTAC incorporation into the pan-DR MS dataset TM Example depiction of cluster assignments for peptides (20 per allele). The dataset was deconvoluted using GibbsCluster. Each colored box represents a MAPTAC TM The color of the box indicates which cluster it is assigned to, and the gray bar indicates which allele the peptide is derived from. Figure 30C Depicted is an exemplary diagram showing Figure 30B The fraction of peptides showing 0, 1, 2, 3, or 4 expected residues at anchor positions for the alleles shown in . Anchor positions were defined as the four positions with the lowest entropy, and "expected" residues were defined as those with a frequency ≥ 10% at these positions.

[0370] Figures 31A-31F Depicted is an exemplary architecture and benchmarking of the neoNMHC2 binding prediction algorithm. Figure 31A Depicts an exemplary architecture of a convolutional neural network (CNN) trained to discriminate monoallelic HLA class II peptides from scrambled, length-matched decoys. The schematic demonstrates the use of an amino acid feature embedding layer, two convolutional layers of width 6, the presence of skip-to-end connections, and a combination of average and max pooling operations as input to the final logic output node. Figure 31B Depicts MAPTAC before training or hyperparameter optimization. TMExample positive predictive values ​​(PPVs) of NetMHCIIpan and neonMHC2 evaluated on data partitions. For each allele, n MS-observed peptides were scored together with 19n length-matched decoys sampled from the same set of source genes, and the n top-ranked peptides (e.g., top 5%) for each predictor were called positive. According to this evaluation scheme, PPV is the same as recall because the number of false positives and false negatives is necessarily equal. Figure 31C Depicted are exemplary PPVs for NetMHCIIpan and neoNeonMHC2 on the TGEM dataset. For each allele, the n top-ranked peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. Figure 31D Depicted are exemplary ex vivo T cell induction results for neoantigenic peptides. Peptides were selected based on a high neoMHC2 score and a weak NetMHCIIpan score for HLA-DRB1*11:01. Figure 31E Depicted is a comparison of a model trained on monoallelic MAPTAC data and evaluated on deconvolved multiallelic data held out monoallelic data. Values ​​are as for neonMHC2, where the training dataset was downsampled to match the size of the deconvolved training set. Figure 31F PPVs on the TGEM dataset for NetMHCIIpan-v3.1, a deconvolution-trained predictor, and neonMHC2 (with and without downsampling) are shown. For each allele, the n top-ranked peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set.

[0371] Figures 32A-32E Depicted are exemplary gene representation and protein processing in the HLA class II tumor peptidomes. Figure 32A Depicted are exemplary results comparing the observed and expected number of HLA class II peptides for each gene, as determined by a joint analysis of colorectal cancer, melanoma, and ovarian cancer datasets. Expected counts are calculated by multiplying the gene length by the expression level. Expected and observed counts for the relevant samples were summed. Genes known to be present in plasma are labeled according to their concentration. Figure 32B Example results of the expected versus observed frequency of localized peptides per cell are depicted. Figure 32C The enrichment scores of genes regulated by the proteasome are depicted (ratio of observed to expected observations, partly as Figure 32B ) distribution. The gene set includes genes with known ubiquitination sites and genes whose abundance increases after application of proteasome inhibitors. Figure 32DDepicted is a diagram presenting three exemplary working models for how HLA class II peptides are processed, according to which: i) cathepsins and other enzymes cleave proteins into peptide fragments that are then bound by HLA, ii) proteins or unfolded polypeptides bind to HLA and are subsequently cleaved into peptide lengths, and iii) proteins are partially digested before binding and further trimmed after binding. Each model corresponds to a different prediction method. Figure 32E Depicted are the absolute increases in PPV observed for logistic regression models that included treatment-related variables and the neoMHC2 binding prediction compared to models using binding prediction alone. Eleven samples analyzed by HLA-DR antibody assay ( Figure 30B Same samples analyzed in ); each point corresponds to one sample. Asterisks indicate significant improvement according to a two-tailed paired t-test (*: p < 0.01, **: p < 0.001, ***: p < 0.0001). The same analysis is shown in Figure 40B Instead, NetMHCIIpan was used as the base predictor. The methods used for bait selection and PPV calculation were the same as those used in Figure 31B The same method is used in .

[0372] Figures 33A-33G Depicted are exemplary results for identifying and predicting tumor antigens presented by dendritic cells. Figure 33A Depicts an exemplary schematic of the experimental workflow for identifying HLA-II ligands presented by DCs derived from cancer cells (K562). Cancer cells were grown in SILAC medium until fully incorporated, lysed or irradiated, and then seeded with monocyte-derived dendritic cells. Presented peptides were isolated using a pan-DR antibody and sequenced by LC-MS / MS. Figure 33B Depicts the use of Figure 21A Example data showing the predictive performance of the same hit:decoy ratio and performance metrics for tumor-derived peptides presented by dendritic cells. The performance of NetMHCIIpan and neoNeonMHC2-based models with and without processing features is shown. Figure 33C Depicted are exemplary gene expression profiles of the source gene of the heavily labeled peptide (red curve, plotted against K562 expression) observed in UV treatment experiments compared to the source gene of the lightly labeled peptide (grey curve, plotted against DC expression). Figure 33DAn exemplary illustration of the PPV for predicting presented tumor antigens at a hit:decoy ratio of 1:499 using NetMHCIIpan and neomhc2-based models with and without treatment features is shown. Data points from left to right represent the following samples: Donor 1 HOCl-treated cells: NetMHCIIpan continuous expression; NetMHCIIpan continuous expression + gene bias; NetMHCIIpan continuous expression + gene bias + DQ overlap, fully treated mode; Donor 1, UV-treated: neomhc2; neomhc2 + threshold expression; neomhc2 + continuous expression; neomhc2 + continuous expression + gene bias; neomhc2 + continuous expression + gene bias + DQ overlap. Figure 33E Depicted are the significance of various gene locations and functional categories in predicting heavy (K562-derived) and light (DC-derived) peptides, respectively. P values ​​were calculated from logistic regression controlling for NeoMHC2 binding score and source gene expression. Bar color indicates the sign associated with the regression coefficient. Figure 33F Depicted is an exemplary graph of the results showing the overlap of tumor cell-derived peptide-derived genes (colored by functional category) in UV and HOCl-treated experiments. Figure 33G Depicted are exemplary data showing the PPV for predicting tumor antigens presented in a second donor using a logistic model fit on heavily labeled peptides observed in a first donor. Models were fitted using neoNHC2 binding alone; binding and expression; or binding, expression, and a binary variable indicating whether the peptide was derived from a mitochondrial gene.

[0373] Figures 34A-34B Depicts the Figure 29 Related MAPTAC TM Exemplary representation of data. Figure 34A Depicts the MAPTAC expression of HLA-A*02:01-BAP by encoding affinity tags TM Exemplary HLA cell surface analysis by FACS of Expi293 cell line transfected with constructs. Figure 34B Depicts the MAPTAC expression of HLA-DRB1*11:01-BAP by encoding affinity-tagged TMExemplary HLA cell surface analysis by FACS of Expi293 cell lines transfected with constructs (bottom). HLA cell surface expression of transfected Expi293 cells (orange) was compared to stained untransfected Expi293 (blue), unstained untransfected Expi293 (red), stained PBMC (dark green), and unstained PBMC (light green). All HLA class I stains used W6 / 32 (pan-HLA class I), while HLA class II stains used REA332 (pan-HLA class II).

[0374] Figure 35 Depicts the Figure 30A Related MAPTAC TM Figure 2. Example comparison of IEDB identifications. Measured affinity observed by MS and predicted affinity by NetMHCIIpan for peptides that did not exhibit a good NetMHCIIpan score but were well supported by MS (scoring peak intensity >70 and nested group size ≥1).

[0375] Figures 36A-36C Depicts the Figures 30A-30C HLA-DR1-associated MAPTAC TM Exemplary analysis of data fidelity. Figure 36A Depicted are the HLA-DR1MAPTAC profiles for the common allele compared to 50,000 length-matched bait peptides (blue) randomly sampled from the proteome. TM Exemplary NetMHCIIpan3.1 scores for peptides (green) (length 12-23). Figure 36B Depicted are exemplary measured affinities and NetMHCIIpan-predicted affinities for exemplary MS-observed peptides that did not exhibit a good NetMHCIIpan score but were well supported by MS (scored peak intensity >70 and nested group size ≥1). Figure 36C Depicted as by MAPTAC TM Exemplary HLA class II sequence signatures of HLA-DRB1 alleles identified in different cell types.

[0376] Figures 37A-37C Depicts the Figures 30A-30C Related MAPTAC TM Additional exemplary analysis of motifs. Figure 37A Depicts MAPTAC used in experiments with and without HLA-DM co-transfection (expi293 cell line) TM Derived sequence identifier. Figure 37B Depicts the MAPTAC TMSequence logos of several HLA class I alleles from the IEDB. Note that A*32:01 does not show a high frequency of Q at P2, and C*03:03 does not show a high frequency of Y at P9, which is different from previous studies using multi-allelic deconvolution; the logo of B*52:01 has not been previously published. Figure 37C Depicting MAPAC TM Exemplary alignment of observed peptides with the gene sequence of CD74.

[0377] Figure 38Ai-38D Depicts the Figures 31A-31D Exemplary NeoMHC2 performance statistics and associated T cell flow staining. Figure 38Ai Plots the exemplary performance of neonmhc2 as a function of the training dataset size. Figure 31B PPV was evaluated in the same manner and using the same evaluation peptides; however, the training data were randomly downsampled to simulate a smaller training dataset. Figure 38Aii Depicted are exemplary sequence logos for peptide clusters derived from the multi-allelic HLA-DR ligand set using GibbsCluster (default settings; "allow junk clusters"). Figure 38B Depicted are exemplary representative flow cytometry plots of IFN-γ expression in CD4+ cells from induced samples determined using neoantigen peptides predicted by neomhc2. Delta values ​​are calculated by subtracting the percentage of CD4+ cells expressing IFN-γ when determined using neoantigens (+ peptides) from the percentage of CD4+ expressing IFN-γ when determined in the absence of neoantigens (no peptides). The two flow cytometry plots on the left represent neoantigens (PEASLYGALSKGSGG) that induce CD4+ T cell T cell responses and neoantigens (PATYILILKEFCLVG) that do not induce T cell responses. Figure 38C Depicted are exemplary delta values ​​from wells called with a single neonmch2 neoantigen peptide. If a peptide had a positive response (delta response above 3%, highlighted), it was considered an induced hit. Figure 38D Shown are exemplary sequence logos for peptide clusters derived for a multi-allelic HLA-DR ligand panel using GibbsCluster (default settings; allowing "junk" clusters).

[0378] Figures 39A-39C Depicts the Figures 32A-32E Additional exemplary cell-of-origin analysis of relevant HLA class II. Figure 39A Describes the presence of the peptide-derived gene in human plasma, in which the pan-DR antibody (from Figure 30BExample percentile neonMHC2 scores for HLA class II peptides observed in four PBMC samples analyzed using the RG1248, RG1104, RG1095, and HDSC from the donor. For each peptide, the best (lowest) percentile ranking was used among the alleles present in the donor. Scores for randomly length-matched proteomic baits are shown for comparison. The boxplots are labeled with the 5th, 25th, 50th, 75th, and 95th percentiles. Figure 39B Describes the use and Figure 32A Example counts of observed versus expected peptides for each HLA class I gene using the same method as in Figure 2. The data correspond to the same tumor types (colorectal cancer, ovarian cancer, and melanoma). Genes present in human plasma are highlighted in blue and sized according to their concentration. Figure 39C Depicted are exemplary relative concordances of peptide observations for two different gene expression profiles. For each sample, gene-level peptide counts were modeled as a linear combination of the bulk tumor gene expression and the professional APC gene expression profile. The ratio of the coefficients determines the relative concordance of each expression profile with the peptide repertoire. Error bars correspond to 95% confidence intervals calculated by bootstrap resampling.

[0379] Figures 40A-40B Depicts the Figures 32A-32E Additional exemplary analysis of relevant processing motifs. Figure 40A Depicted are the results of studies on donor PBMC, monocyte-derived dendritic cells, colorectal cancer, melanoma, ovarian cancer, and expi293 cell lines (used in most MAPTACs). TM Exemplary amino acid frequencies near the N-terminal and C-terminal peptide cleavage sites relative to the average proteome frequency (for upstream positions U3-U1 and downstream positions D1-D3) or relative to the average peptide frequency (for internal positions N1-C1) as observed in the data generation (data generation). Figure 40B Depicts the use of NetMHCIIpan as a basic predictor with Figure 32E The same analysis was performed for the 8 samples analyzed by HLA-DR antibody ( Figure 31B Figure 3. Absolute increase in PPV observed for the logistic regression model that included treatment-related variables in addition to NetMHCIIpan predictions (compared to the model using NetMHCIIpan alone) (same samples analyzed in ). Asterisks indicate significant improvements according to a two-tailed paired t-test (*: p < 0.01, **: p < 0.001, ***: p < 0.0001).

[0380] Figure 41 An exemplary nomenclature system for referring to positions upstream of a peptide, within a peptide, and downstream of a peptide is depicted.

[0381] Figure 42ADepicted is a diagram representing an exemplary workflow for analysis of endogenously processed and HLA-1 and HLA class II presented peptides by nLC-MS / MS.

[0382] Figure 42B Depicted are diagrams showing exemplary experimental results from nLC-MS / MS analysis of tryptic peptides with or without FAIMS. Also depicted are representative overlaps of HLA-1 and HLA class II peptides detected by nLC-MS / MS analysis with or without FAIMS at the indicated analytical scale.

[0383] Figure 43A Depicted are exemplary HLA class I acidic and basic reverse phase fractionation peptide detection with and without FAIMS.

[0384] Figure 43B Depicted are exemplary experimental results showing the detection of unique peptides for HLA class I binding plotted against retention time.

[0385] Figure 44A Depicted are exemplary HLA class II acidic and basic reverse phase fractionation peptide detection with and without FAIMS.

[0386] Figure 44B Depicted are exemplary experimental results showing the detection of unique peptides for HLA class II binding plotted against retention time.

[0387] Figure 45 Depicted are an exemplary graphical representation of the cross-size of HLA class I binding peptides detected using the indicated methods (left) and a Venn diagram of exemplary standard and optimized workflows for LC-MS / MS detection of HLA class I binding peptides (right).

[0388] Figure 46 Depicted are an exemplary graphical representation of the cross-size of HLA class II binding peptides detected using the indicated methods (left) and a Venn diagram of exemplary standard and optimized workflows for LC-MS / MS detection of HLA class II binding peptides (right). DETAILED DESCRIPTION

[0389] All terms should be understood as they would be understood by those skilled in the art. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

[0390] The section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.

[0391] Although various features of the present disclosure may be described in the context of a single embodiment, these features may also be provided separately or in any suitable combination. Conversely, although the present disclosure may be described herein in the context of separate embodiments for clarity, the present disclosure may also be implemented in a single embodiment.

[0392] The present disclosure is based on the important discovery that presentation of antigens, particularly cancer antigens, by specific HLA class II α and β chain pairs can be predicted with high confidence using a new computer-based machine learning HLA-peptide presentation prediction model that allows for improved immunotherapy using HLA class II-specific peptides.

[0393] In one aspect, the present disclosure provides methods for predicting peptides that can accurately pair or bind to specific HLA class II α and β chain heterodimers, such that high-fidelity binding of the peptide to HLA class II proteins (including α and β chain heterodimers) ensures that the specific peptide is presented to T lymphocytes, thereby eliciting a specific immune response and avoiding any cross-reactivity or immune promiscuity. Several recent studies have shown that CD4+ T cells can also recognize ligands presented by HLA class II and help control tumors. Ideally, cancer vaccines and other immunotherapies would utilize guided CD4+ T cell responses, but current efforts have completely abandoned HLA class II antigen prediction due to the insufficient accuracy of current prediction tools.

[0394] In one aspect, the present disclosure provides methods for predicting peptides that accurately bind to specific HLA class II proteins, such that when the peptides are therapeutically administered to subjects expressing specific cognate HLA class II proteins, the peptides can be used to activate a more sustained and potent immune response, leveraging the ability of HLA class II proteins to activate CD4+ T cells and stimulate immune memory. In some embodiments, the methods provided herein demonstrate improvements in prediction of specific HLA class II proteins relative to currently available predictors. In some embodiments, the methods provided herein demonstrate at least about a 1.1-fold improvement in prediction of specific HLA class II proteins relative to currently available predictors. In some embodiments, the methods provided herein demonstrate at least about a 2-fold improvement in prediction of specific HLA class II proteins relative to currently available predictors. In some embodiments, the methods provided herein demonstrate at least about a 3-fold improvement in prediction of specific HLA class II proteins relative to currently available predictors. In some embodiments, the methods provided herein demonstrate at least about a 4-fold improvement in prediction of specific HLA class II proteins relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 5-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 6-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 7-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about an 8-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 9-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 10-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 15-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 20-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein demonstrate at least about a 30-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein demonstrate at least about a 40-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors.In some embodiments, the methods provided herein exhibit at least about a 50-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors. In some embodiments, the methods provided herein exhibit at least about a 60-fold improvement in the prediction of a particular HLA class II protein relative to currently available predictors.

[0395] In one aspect, the present invention provides a customized or personalized immunotherapy approach for a specific subject. Each subject or patient expresses a specific set of HLA class I and HLA class II proteins. HLA typing is a well-known technique that allows determination of the specific repertoire of HLA proteins expressed by a subject. Once the HLA heterodimers expressed by a particular subject are known, improved, sophisticated, and reliable methods as described herein for predicting peptides that bind to specific HLA class II α and β chain heterodimers with high fidelity can ensure that a specific immune response tailored specifically for that subject can be generated.

[0396] In this application, unless otherwise specifically stated, the use of the singular includes the plural. It must be noted that, as used in this specification, the singular forms "a", "an" and "the" include plural referents unless the context clearly indicates otherwise. In this application, the use of "or" means "and / or" unless otherwise stated. In addition, the use of the term "includes" and other forms such as "comprises", "contains" and "has" is not restrictive. The term "one or more" or "at least one", for example one or more or at least one member of a group of members, is clear in itself and, by way of further illustration, the term particularly includes reference to any one of the members, or any two or more of the members, for example, any 3, 4, 5, 6 or 7 of the members, etc., up to all of the members.

[0397] Reference in this specification to "some embodiments," "an embodiment," "one embodiment," or "other embodiments" means that a feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments of the present disclosure, but not necessarily all embodiments.

[0398] As used in this specification and claims, the words "comprising" (and any form of comprising), "having" (and any form of having), "including" (and any form of including), or "containing" (and any form of containing) are inclusive or open-ended and do not exclude other unrecited elements or method steps. It is contemplated that any embodiment discussed in this specification can be implemented using any method or composition of the present disclosure, and vice versa. In addition, the compositions of the present disclosure can be used to implement the methods of the present disclosure.

[0399] As used herein, the terms "about" or "approximately" when referring to a measurable value such as a parameter, amount, duration, etc., are intended to encompass variations of + / - 20% or less, + / - 10% or less, + / - 5% or less, or + / - 1% or less of the specified value, so long as such variations are suitable for use in the present disclosure. It is understood that the value to which the modifier "about" or "approximately" refers is itself specifically disclosed.

[0400] The term "immune response" includes T cell-mediated and / or B cell-mediated immune responses that are affected by T cell costimulation regulation. Exemplary immune responses include T cell responses, such as the generation of cytokines and the cytotoxicity of cells. In addition, the term immune response includes immune responses that are indirectly affected by T cell activation, such as antibody production (humoral response) and cytokine responsive cells such as the activation of macrophages.

[0401] "Receptor" should be understood to refer to a biological molecule or group of molecules capable of binding a ligand. Receptors can be used to transmit information within a cell, cell formation, or organism. A receptor comprises at least one receptor unit and may contain two or more receptor units, each of which may be composed of a protein molecule, such as a glycoprotein molecule. A receptor has a structure complementary to that of a ligand and can complex with the ligand as a binding partner. Signal information can be transmitted through conformational changes following binding of the receptor to a ligand on the cell surface. According to the present disclosure, a receptor may refer to MHC class I and class II proteins that are capable of forming a receptor / ligand complex with a ligand, such as a peptide or peptide fragment of appropriate length. The class I and class II MHC peptides encoded by HLA class I and class II alleles are generally referred to herein as HLA class I and HLA class II peptides, or HLA class I and HLA class II peptides, or HLA class I and class II proteins, or HLA class I and class II proteins, or HLA class I and class II molecules, or common variants thereof, as known to those of ordinary skill in the art in the context of the discussion.

[0402] A "ligand" is a molecule that is capable of forming a complex with a receptor. According to the present disclosure, a ligand is understood to mean a peptide or peptide fragment, for example, of a suitable length and a suitable binding motif in its amino acid sequence, such that the peptide or peptide fragment is capable of binding to and forming a complex with an MHC class I or MHC class II protein (i.e., HLA class I and HLA class II proteins).

[0403] "Antigen" is a molecule that can stimulate an immune response and can be produced by cancer cells, infectious agents, or autoimmune diseases. Antigens recognized by T cells (whether helper T lymphocytes (T helper (TH) cells) or cytotoxic T lymphocytes (CTL)) are not recognized as complete proteins, but as small peptides associated with HLA class I or class II proteins on the cell surface. In the process of a naturally occurring immune response, antigens recognized by association with HLA class II molecules on antigen presenting cells (APCs) are obtained from the extracellular space, internalized, and processed into small peptides associated with HLA class II molecules. APCs can also cross-present peptide antigens by processing exogenous antigens and presenting the processed antigens to HLA class I molecules. Antigens that produce peptides recognized by association with HLA class I MHC molecules are generally peptides produced intracellularly, and these antigens are processed and associated with class I MHC molecules. It is now understood that peptides that associate with a given HLA class I or class II molecule are characterized as having a common binding motif, and binding motifs have been identified for a large number of different HLA class I and class II molecules. Synthetic peptides can also be synthesized that correspond to the amino acid sequence of a given antigen and contain a binding motif for a given HLA class I or class II molecule. These peptides can then be added to appropriate APCs, and the APCs can be used to stimulate T helper cell or CTL responses in vitro or in vivo. Binding motifs, methods for synthesizing peptides, and methods for stimulating T helper cell or CTL responses are all known to those of ordinary skill in the art and are readily available.

[0404] In this specification, the term "peptide" is used interchangeably with "mutant peptide" and "neoantigenic peptide". Similarly, in this specification, the term "polypeptide" is used interchangeably with "mutant polypeptide" and "neoantigenic polypeptide". "Neoantigen" or "neo-epitope" refers to a class of tumor antigens or tumor epitopes produced by tumor-specific mutations in expressed proteins. The present disclosure further includes peptides comprising tumor-specific mutations, peptides comprising known tumor-specific mutations, and mutant polypeptides or fragments thereof identified by the methods of the present disclosure. These peptides and polypeptides are referred to herein as "neoantigenic peptides" or "neoantigenic polypeptides". These polypeptides or peptides may have a variety of lengths, may be in their neutral (uncharged) form or in salt form, and may be free of modifications such as glycosylation, side chain oxidation, phosphorylation or any post-translational modification, or may contain these modifications, provided that the modification does not destroy the biological activity of the polypeptides described herein. In some embodiments, the neoantigenic peptides of the present disclosure may comprise: for HLA class I, a length of 22 or fewer residues, for example, about 8 to about 22 residues, about 8 to about 15 residues, or 9 or 10 residues; for HLA class II, a length of 40 or fewer residues, for example, about 8 to about 40 residues, about 8 to about 24 residues, about 12 to about 19 residues, or about 14 to about 18 residues. In some embodiments, the neoantigenic peptide or neoantigenic polypeptide comprises a neoepitope.

[0405] The term "epitope" includes any protein determinant that can specifically bind to an antibody, antibody peptide and / or antibody-like molecule (including but not limited to a T cell receptor) as defined herein. Epitope determinants are typically composed of chemically active surface groups of molecules such as amino acids or sugar side chains, and typically have specific three-dimensional structural characteristics and specific charge characteristics.

[0406] A "T cell epitope" is a peptide sequence that can be bound by an MHC class I or II molecule in the form of a peptide-presenting MHC molecule or MHC complex and is then recognized and bound in this form by cytotoxic T lymphocytes or T helper cells, respectively.

[0407] As used herein, the term "antibody" includes IgG (including IgG1, IgG2, IgG3 and IgG4), IgA (including IgA1 and IgA2), IgD, IgE, IgM and IgY, and is intended to include complete antibodies, including single-chain complete antibodies, and antigen-binding (Fab) fragments thereof. Antigen-binding antibody fragments include, but are not limited to, Fab, Fab' and F(ab')2, Fd (consisting of VH and CH1), single-chain variable fragments (scFv), single-chain antibodies, disulfide-linked variable fragments (dsFv) and fragments comprising VL or VH domains. Antibodies can be from any animal source. Antigen-binding antibody fragments, including single-chain antibodies, can comprise a variable region alone or in combination with all or part of the following: hinge region, CH1, CH2 and CH3 domains. Any combination of variable region and hinge region, CH1, CH2 and CH3 domains is also included. The antibodies can be, for example, monoclonal antibodies, polyclonal antibodies, chimeric antibodies, humanized antibodies, and human monoclonal and polyclonal antibodies that specifically bind to an HLA-associated polypeptide or an HLA-HLA binding peptide (HLA-peptide) complex. Those skilled in the art will recognize that a variety of immunoaffinity techniques are suitable for enriching soluble proteins, such as soluble HLA-peptide complexes or membrane-bound HLA-associated polypeptides, for example, which have been cleaved from a membrane by proteolysis. These include techniques in which (1) one or more antibodies capable of specifically binding to soluble proteins are immobilized on a fixed or removable substrate (e.g., a plastic well or resin, latex, or paramagnetic beads), and (2) a solution containing soluble proteins from a biological sample is passed through the antibody-coated substrate, thereby binding the soluble proteins to the antibodies. The substrate with the antibodies and bound soluble proteins is separated from the solution, and the antibodies and soluble proteins are optionally dissociated, for example, by changing the pH and / or ionic strength and / or ionic composition of the solution in which the antibodies are bathed. Alternatively, immunoprecipitation techniques can be used, in which antibodies and soluble proteins are combined and form macromolecular aggregates. The macromolecular aggregates can be separated from the solution by size exclusion techniques or by centrifugation.

[0408] The term "immunopurification (IP)" (or immunoaffinity purification or immunoprecipitation) is a method well known in the art and is widely used in separating desired antigens from samples. Typically, the method includes contacting a sample containing the desired antigen with an affinity matrix comprising antibodies against the antigen covalently attached to a solid phase. The antigen in the sample is bound to the affinity matrix via immunochemical bonds. The affinity matrix is ​​then washed to remove any unbound material. The antigen is removed from the affinity matrix by changing the chemical composition of the solution in contact with the affinity matrix. Immunopurification can be performed on a column containing an affinity matrix, in which case the solution is an eluent. Alternatively, immunopurification can be a batch process, in which case the affinity matrix remains as a suspension in the solution. An important step in the process is removing the antigen from the matrix. This is typically achieved by increasing the ionic strength of the solution in contact with the affinity matrix, such as by adding inorganic salts. Changes in pH can also effectively dissociate the immunochemical bonds between the antigen and the affinity matrix.

[0409] An "agent" is any small molecule compound, antibody, nucleic acid molecule, or polypeptide, or a fragment thereof.

[0410] A "change" or "variation" is an increase or decrease. The change may be as little as 1%, 2%, 3%, 4%, 5%, 10%, 20%, 30%, or 40%, 50%, 60%, or as much as 70%, 75%, 80%, 90%, or 100%.

[0411] A "biological sample" is any tissue, cell, fluid, or other substance derived from an organism. As used herein, the term "sample" includes biological samples, such as any tissue, cell, fluid, or other substance derived from an organism. "Specific binding" refers to a compound (e.g., a peptide) that recognizes and binds to a molecule (e.g., a polypeptide) but does not substantially recognize and bind to other molecules in a sample (e.g., a biological sample).

[0412] A "capture reagent" refers to a reagent that specifically binds to a molecule (eg, a nucleic acid molecule or a polypeptide) to select or isolate that molecule (eg, a nucleic acid molecule or a polypeptide).

[0413] As used herein, the terms "determine," "assess," "determine," "measure," "detect," and their grammatical equivalents refer to both quantitative and qualitative determinations, and thus, the terms "determine" and "determine," "measure," and the like are used interchangeably herein. Where quantitative determination is intended, the phrase "determine the amount of an analyte, etc." is used. Where qualitative and / or quantitative determination is intended, the phrase "determine the level of an analyte" or "detect" an analyte is used.

[0414] A "fragment" is a portion of a protein or nucleic acid that is substantially identical to a reference protein or nucleic acid. In some embodiments, the portion retains at least 50%, 75%, or 80%, or 90%, 95%, or even 99% of the biological activity of the reference protein or nucleic acid described herein.

[0415] The terms "isolated," "purified," "biologically pure," and grammatical equivalents thereof refer to a substance that is free to varying degrees from components with which it normally accompanies its native state. "Isolated" refers to a degree of separation from the original source or environment. "Purity" refers to a degree of separation greater than isolated. A "purified" or "biologically pure" protein is sufficiently free of other substances such that any impurities do not materially affect the biological properties of the protein or cause other adverse consequences. That is, a nucleic acid or peptide of the present disclosure is purified if it is substantially free of cellular material, viral material, or culture medium when produced by recombinant DNA techniques, or substantially free of chemical precursors or other chemicals when chemically synthesized. Purity and homogeneity are typically determined using analytical chemistry techniques, such as polyacrylamide gel electrophoresis or high performance liquid chromatography. The term "purified" can mean that a nucleic acid or protein produces essentially a single band in an electrophoretic gel. For proteins that can be modified, such as phosphorylation or glycosylation, different modifications can produce different isolated proteins, which can be purified separately.

[0416] An "isolated" polypeptide (e.g., a peptide from an HLA-peptide complex) or polypeptide complex (e.g., an HLA-peptide complex) is a polypeptide or polypeptide complex of the present disclosure that has been separated from naturally associated components. Generally, a polypeptide or polypeptide complex is isolated when it is at least 60% free by weight from the proteins and naturally occurring organic molecules with which it is naturally associated. The preparation can be at least 75%, at least 90%, or at least 99% by weight of the polypeptide or polypeptide complex of the present disclosure. An isolated polypeptide or polypeptide complex of the present disclosure can be obtained, for example, by extraction from a natural source, by expression of a recombinant nucleic acid encoding one or more components of the polypeptide or polypeptide complex, or by chemical synthesis of one or more components of the polypeptide or polypeptide complex. Purity can be measured by any suitable method, such as column chromatography, polyacrylamide gel electrophoresis, or by HPLC analysis. In some cases, the MHC class II protein (i.e., MHC class II peptide) encoded by an HLA allele is referred to interchangeably herein as an HLA class II protein (or HLA class II peptide).

[0417] The term "vector" refers to a nucleic acid molecule capable of transporting or mediating the expression of a heterologous nucleic acid. A plasmid is one of the types encompassed by the term "vector". A vector generally refers to a nucleic acid sequence containing an origin of replication and other entities necessary for replication and / or maintenance in a host cell. A vector capable of directing the expression of a gene and / or nucleic acid sequence to which it is operably linked is referred to herein as an "expression vector". Typically, useful expression vectors are typically in the form of "plasmids", which refer to circular double-stranded DNA molecules that are not bound to chromosomes in vector form and typically contain entities or encoding DNA for stable or transient expression. Other expression vectors that can be used in the methods disclosed herein include, but are not limited to, plasmids, episomes, bacterial artificial chromosomes, yeast artificial chromosomes, phage or viral vectors, and such vectors can be integrated into the host's genome or replicate autonomously in the cell. The vector can be a DNA or RNA vector. Other forms of expression vectors known to those skilled in the art that function equivalently, for example, self-replicating extrachromosomal vectors or vectors capable of integrating into the host genome, can also be used. Exemplary vectors are vectors capable of autonomous replication and / or expression of nucleic acids to which they are linked.

[0418] The term "spacer" or "connector" used in connection with fusion proteins refers to a peptide that connects the proteins comprising the fusion protein. Typically, spacers have no specific biological activity other than connecting or maintaining a certain minimum distance or other spatial relationship between protein or RNA sequences. However, in some embodiments, the constituent amino acids of the spacer can be selected to affect certain properties of the molecule, such as the folding, net charge, or hydrophobicity of the molecule. Suitable connectors for use in the embodiments of the present disclosure are well known to those skilled in the art and include, but are not limited to, straight or branched carbon connectors, heterocyclic carbon connectors, or peptide connectors. Connectors are used to separate two antigenic peptides by a certain distance that is sufficient to ensure that each antigenic peptide folds correctly in some embodiments. Exemplary peptide connector sequences adopt a flexible extended conformation and do not show a tendency to develop an ordered secondary structure. Typical amino acids in flexible protein regions include Gly, Asn, and Ser. In fact, it is expected that any arrangement of amino acid sequences containing Gly, Asn, and Ser will meet the above-mentioned criteria for connector sequences. Other nearly neutral amino acids, such as Thr and Ala, can also be used in connector sequences. Other amino acid sequences that can be used as linkers are disclosed in Maratea et al. (1985), Gene 40:39-46, Murphy et al. (1986) Proc. Nat'l. Acad. Sci. USA 83:8258-62, US Patent 4,935,233, and US Patent 4,751,180.

[0419] The term "neoplasia" refers to any disease that results in or is caused by inappropriately high levels of cell division, inappropriately low levels of apoptosis, or both. Glioblastoma is a non-limiting example of a neoplasia or cancer. The terms "cancer" or "tumor" or "hyperproliferative disorder" refer to the presence of cells that have typical characteristics of oncogenic cells, such as uncontrolled proliferation, unlimited proliferation, metastatic potential, rapid growth and proliferation rates, and certain distinctive morphological features. Cancer cells are often in the form of tumors, but such cells can exist alone in an animal or can be non-tumorigenic cancer cells, such as leukemia cells. Cancers include, but are not limited to, B-cell cancers, e.g., multiple myeloma, Waldenstrom's macroglobulinemia, heavy chain diseases (e.g., alpha chain disease, gamma chain disease, and mu chain disease), benign monoclonal gammopathy and immune cell amyloidosis, melanoma, breast cancer, lung cancer, bronchogenic cancer, colorectal cancer, prostate cancer (e.g., metastatic, hormone-refractory prostate cancer), pancreatic cancer, gastric cancer, ovarian cancer, bladder cancer, brain or central nervous system cancer, peripheral nervous system cancer, esophageal cancer, cervical cancer, uterine cancer or endometrial cancer, cancer of the oral cavity or pharynx, liver cancer, kidney cancer, testicular cancer, biliary tract cancer, small intestine or appendix cancer, salivary gland cancer, thyroid cancer, adrenal cancer, osteosarcoma, chondrosarcoma, cancer of blood tissue, etc. Other non-limiting examples of cancer types suitable for the methods encompassed by the present disclosure include human sarcomas and carcinomas, e.g., fibrosarcoma, myxosarcoma, liposarcoma, chondrosarcoma, osteosarcoma, chordoma, angiosarcoma, endotheliosarcoma, lymphangiosarcoma, lymphangioendotheliosarcoma, synovioma, mesothelioma, Ewing's tumor, and sarcoma. tumor), leiomyosarcoma, rhabdomyosarcoma, colon cancer, colorectal cancer, pancreatic cancer, breast cancer, ovarian cancer, squamous cell carcinoma, basal cell carcinoma, adenocarcinoma, sweat gland cancer, sebaceous gland cancer, papillary carcinoma, papillary adenocarcinoma, cystadenocarcinoma, medullary carcinoma, bronchogenic carcinoma, renal cell carcinoma, hepatoma, bile duct cancer, liver cancer, choriocarcinoma, seminoma, embryonal carcinoma, Wilms' tumor, cervical cancer, bone cancer, brain tumor, testicular cancer, lung cancer, small cell lung cancer, bladder cancer, epithelial carcinoma, glioma, astrocytoma, medulloblastoma, craniopharyngioma, ependymoma, pineal carcinoma leukemias, e.g., acute lymphoblastic leukemia and acute myeloid leukemia (myeloblastic, promyelocytic, myelomonocytic, monocytic, and erythroleukemia); chronic leukemias (chronic myeloid (granulocytic) leukemia and chronic lymphocytic leukemia); as well as polycythemia vera, lymphomas (Hodgkin's disease and non-Hodgkin's disease), multiple myeloma, Waldenstrom's macroglobulinemia, and heavy chain disease.In some embodiments, the cancer is an epithelial cancer, such as but not limited to bladder cancer, breast cancer, cervical cancer, colon cancer, gynecological cancer, kidney cancer, laryngeal cancer, lung cancer, oral cancer, head and neck cancer, ovarian cancer, pancreatic cancer, prostate cancer or skin cancer. In other embodiments, the cancer is breast cancer, prostate cancer, lung cancer or colon cancer. In other embodiments, the epithelial cancer is non-small cell lung cancer, non-papillary renal cell carcinoma, cervical cancer, ovarian cancer (e.g., serous ovarian cancer) or breast cancer. Epithelial cancer can be characterized in various other ways, including but not limited to serous, endometrioid, mucinous, clear cell, Brenner type (brenner) or undifferentiated. In some embodiments, the present disclosure is used for the treatment, diagnosis and / or prognosis of lymphoma or its subtypes (including but not limited to mantle cell lymphoma). Lymphoproliferative disorders are also considered to be proliferative diseases.

[0420] The term "vaccine" should be understood to refer to a composition used to generate immunity to prevent and / or treat a disease (e.g., neoplasia / tumor / infectious agent / autoimmune disease). Therefore, a vaccine is a drug comprising an antigen and is intended to be used in humans or animals to produce specific defenses and protective substances by vaccination. A "vaccine composition" may comprise a pharmaceutically acceptable excipient, carrier, or diluent. Aspects of the present disclosure relate to the application of this technology in the preparation of antigen-based vaccines. In these embodiments, a vaccine refers to one or more disease-specific antigenic peptides (or their corresponding nucleic acids encoding them). In some embodiments, the antigen-based vaccine contains at least two, at least three, at least four, at least five, at least six, at least seven, at least eight, at least nine, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 16, at least 17, at least 18, at least 19, at least 20, at least 21, at least 22, at least 23, at least 24, at least 25, at least 26, at least 27, at least 28, at least 29, at least 30 or more antigenic peptides. In some embodiments, the antigen-based vaccine contains 2 to 100, 2 to 75, 2 to 50, 2 to 25, 2 to 20, 2 to 19, 2 to 18, 2 to 17, 2 to 16, 2 to 15, 2 to 14, 2 to 13, 2 to 12, 2 to 10, 2 to 9, 2 to 8, 2 to 7, 2 to 6, 2 to 5, 2 to 4, 3 to 100, 3 to 75, 3 to 50, 3 to 25, 3 to 20, 3 to 19, 3 to 18, 3 to 17, 3 to 16, 3 to 15, 3 to 14, 3 to 13, 3 to 12, 3 to 10, 3 to 9, 3 to 8 5 to 7, 4 to 6, 5 to 100, 5 to 75, 5 to 50, 5 to 25, 5 to 20, 5 to 19, 5 to 18, 4 to 17, 4 to 16, 4 to 15, 4 to 14, 4 to 13, 4 to 12, 4 to 10, 4 to 9, 4 to 8, 4 to 7, 4 to 6, 5 to 100, 5 to 75, 5 to 50, 5 to 25, 5 to 20, 5 to 19, 5 to 18, 5 to 17, 5 to 16, 5 to 15, 5 to 14, 5 to 13, 5 to 12, 5 to 10, 5 to 9, 5 to 8, or 5 to 7 antigenic peptides. In some embodiments, the antigen-based vaccine contains 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 antigenic peptides. In some cases, the antigenic peptide is a neoantigenic peptide. In some cases, the antigenic peptide comprises one or more neoepitopes.

[0421] The term "pharmaceutically acceptable" refers to an agent that is approved or approvable by a regulatory agency of the federal or state government, or is listed in the U.S. Pharmacopoeia or other generally recognized pharmacopoeia for use in animals, including humans. A "pharmaceutically acceptable excipient, carrier, or diluent" refers to an excipient, carrier, or diluent that can be administered to a subject together with a medicament and does not destroy its pharmacological activity and is non-toxic when administered in a dose sufficient to deliver a therapeutic amount of the medicament. As described herein, a "pharmaceutically acceptable salt" of the combined disease-specific antigens can be an acid salt or basic salt that is generally considered in the art to be suitable for contact with human or animal tissues without excessive toxicity, irritation, allergic reaction, or other problems or complications. Such salts include inorganic and organic acid salts of basic residues such as amines, and alkali metal or organic salts of acidic residues such as carboxylic acids. Specific pharmaceutical salts include, but are not limited to, salts of acids such as hydrochloric acid, phosphoric acid, hydrobromic acid, malic acid, glycolic acid, fumaric acid, sulfuric acid, p-sulfamic acid, aminobenzenesulfonic acid, formic acid, toluenesulfonic acid, methanesulfonic acid, benzenesulfonic acid, ethanedisulfonic acid, 2-hydroxyethylsulfonic acid, nitric acid, benzoic acid, 2-acetoxybenzoic acid, citric acid, tartaric acid, lactic acid, stearic acid, salicylic acid, glutamic acid, ascorbic acid, pamoic acid, succinic acid, fumaric acid, maleic acid, propionic acid, hydroxymaleic acid, hydroiodic acid, phenylacetic acid, alkanoic acids such as acetic acid, HOOC-(CH2)n-COOH, where n is 0-4, etc. Similarly, pharmaceutically acceptable cations include, but are not limited to, sodium, potassium, calcium, aluminum, lithium, and ammonium. Those skilled in the art will recognize from this disclosure and the knowledge in the art that other pharmaceutically acceptable salts of the combined disease-specific antigens provided herein include those listed in Remington's Pharmaceutical Sciences, 17th edition, Mack Publishing Company, Easton, PA, p. 1418 (1985). In general, pharmaceutically acceptable acid salts or base salts can be synthesized from parent compounds containing a basic or acidic moiety by any conventional chemical method. Briefly, such salts can be prepared by reacting the free acid or base form of these compounds with a stoichiometric amount of the appropriate base or acid in a suitable solvent.

[0422] Nucleic acid molecules that can be used in the methods of the present invention include any nucleic acid molecules encoding polypeptides of the present invention or fragments thereof. Such nucleic acid molecules do not have to be 100% identical to the endogenous nucleic acid sequence, but generally exhibit substantial identity. Polynucleotides having substantial identity to the endogenous sequence are generally capable of hybridizing to at least one strand of a double-stranded nucleic acid molecule. "Hybridization" refers to the pairing of nucleic acid molecules under various stringent conditions to form double-stranded molecules between complementary polynucleotide sequences or portions thereof. (See, e.g., Wahl, GM and SL Berger (1987) Methods Enzymol. 152: 399; Kimmel, AR (1987) Methods Enzymol. 152: 507). For example, stringent salt concentrations can generally be less than about 750 mM NaCl and 75 mM trisodium citrate, less than about 500 mM NaCl and 50 mM trisodium citrate, or less than about 250 mM NaCl and 25 mM trisodium citrate. In the absence of an organic solvent such as formamide, low stringency hybridization can be obtained, while in the presence of at least about 35% formamide or at least about 50% formamide, high stringency hybridization can be obtained. Stringent temperature conditions can generally include a temperature of at least about 30°C, at least about 37°C, or at least about 42°C. It is well known to those skilled in the art to vary other parameters, such as hybridization time, the concentration of detergents such as sodium dodecyl sulfate (SDS), and the inclusion or exclusion of carrier DNA. Various stringency levels are achieved by combining these various conditions as needed. In an exemplary embodiment, hybridization can occur at 30°C in 750mM NaCl, 75mM trisodium citrate, and 1% SDS. In another exemplary embodiment, hybridization can be carried out at 37°C in 500mM NaCl, 50mM trisodium citrate, 1% SDS, 35% formamide, and 100 μg / ml denatured salmon sperm DNA (ssDNA). In another exemplary embodiment, hybridization can occur at 42°C in 250 mM NaCl, 25 mM trisodium citrate, 1% SDS, 50% formamide, and 200 μg / ml ssDNA. Useful variations on these conditions will be apparent to those skilled in the art. For most applications, the post-hybridization wash step may also vary in stringency. Wash stringency conditions can be defined by salt concentration and temperature. As described above, wash stringency can be increased by lowering the salt concentration or by increasing the temperature. For example, the stringent salt concentration for the wash step can be less than about 30 mM NaCl and 3 mM trisodium citrate, or less than about 15 mM NaCl and 1.5 mM trisodium citrate. Stringent temperature conditions for the wash step can include a temperature of at least about 25°C, at least about 42°C, or at least about 68°C. In an exemplary embodiment, the wash step can be performed at 25°C in 30 mM NaCl, 3 mM trisodium citrate, and 0.1% SDS.In other exemplary embodiments, the wash step can be performed in 15 mM NaCl, 1.5 mM trisodium citrate, and 0.1% SDS at 42° C. In another exemplary embodiment, the wash step can be performed in 15 mM NaCl, 1.5 mM trisodium citrate, and 0.1% SDS at 68° C. Other variations on these conditions will be apparent to those skilled in the art. Hybridization techniques are well known to those skilled in the art and are described, for example, in Benton and Davis (Science 196:180, 1977); Grunstein and Hogness (Proc. Natl. Acad. Sci., USA 72:3961, 1975); Ausubel et al. (Current Protocols in Molecular Biology, Wiley Interscience, New York, 2001); Berger and Kimmel (Guide to Molecular Cloning Techniques, 1987, Academic Press, New York); and Sambrook et al., Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory Press, New York.

[0423] "Substantially identical" refers to a polypeptide or nucleic acid molecule that exhibits at least 50% identity to a reference amino acid sequence (e.g., any amino acid sequence described herein) or a nucleic acid sequence (e.g., any nucleic acid sequence described herein). Such a sequence can be at least 60%, 80%, or 85%, 90%, 95%, 96%, 97%, 98% or even 99% or higher levels identical to the sequence used for comparison at the amino acid level or nucleic acid level. Sequence identity is typically measured using sequence analysis software (e.g., Sequence Analysis Software Package of the Genetics Computer Group, University of Wisconsin Biotechnology Center, 1710 University Avenue, Madison, Wis. 53705, BLAST, BESTFIT, GAP, or PILEUP / PRETTYBOX programs). Such software matches identical or similar sequences by assigning degrees of homology to various substitutions, deletions, and / or other modifications. Conservative substitutions generally include substitutions within the following groups: glycine, alanine; valine, isoleucine, leucine; aspartic acid, glutamic acid, asparagine, glutamine; serine, threonine; lysine, arginine; and phenylalanine, tyrosine. In an exemplary method for determining the degree of identity, the BLAST program can be used, where a probability score between e-3 and emin° indicates closely related sequences. A "reference" is a comparison standard.

[0424] The term "subject" or "patient" refers to an animal that is the object of treatment, observation, or experiment. By way of example only, a subject includes, but is not limited to, a mammal, including, but not limited to, a human or a non-human mammal, such as a non-human primate, mouse, cow, horse, dog, sheep, or cat.

[0425] The terms "treat", "treat" and the like mean to reduce, prevent or improve a condition and / or the symptoms associated therewith (e.g., neoplasia or tumor or infectious agent or autoimmune disease). "Treatment" may refer to administering treatment to a subject after an attack or suspected attack of a disease (e.g., cancer or infection with an infectious agent or autoimmune disease). "Treatment" includes the concept of "relief", which refers to reducing the frequency or severity of any symptoms or other adverse effects associated with the disease and / or side effects associated with the treatment. The term "treatment" also encompasses the concept of "management", which refers to reducing the severity of the disease or condition in a patient, for example, prolonging the lifespan of a patient with the disease or prolonging their survival, or delaying their recurrence, for example, extending the remission period of a patient already suffering from the disease. It should be understood that, although not excluded, treating a disease or condition does not require the complete elimination of the condition, condition or symptoms associated therewith.

[0426] As used herein, the terms "prevent," "preventing," and grammatical equivalents thereof refer to avoiding or delaying the onset of symptoms associated with a disease or condition in a subject who has not already developed such symptoms when administration of an agent or compound is initiated.

[0427] The term "therapeutic effect" refers to a certain degree of alleviation of one or more symptoms of a disease (e.g., neoplasia, infection of a tumor or infectious agent or autoimmune disease) or its associated pathology. "Therapeutically effective amount" as used herein refers to the dosage of the degree that effectively prolongs the survival of patients with such disease, alleviates one or more signs or symptoms of the disease, prevents or delays, etc., beyond what would be expected in the absence of such treatment, after a single or multiple dose is administered to a cell or subject. "Therapeutically effective amount" is intended to limit the amount required for the therapeutic effect. A physician or veterinarian with ordinary skill in the art can easily determine and prescribe a "therapeutically effective amount" (e.g., ED50) of the desired pharmaceutical composition. For example, a physician or veterinarian can start the dosage of the compounds of the present invention used in a pharmaceutical composition at a level lower than the level required to obtain the desired therapeutic effect, and gradually increase the dosage until the desired effect is obtained. Disease, condition, and disease are used interchangeably herein.

[0428] Those of ordinary skill in the art will recognize that the terms "peptide tag," "affinity tag," "epitope tag," or "affinity receptor tag" are used interchangeably herein. As used herein, the term "affinity receptor tag" refers to an amino acid sequence that allows for easy detection or purification of the tagged protein, for example, by affinity purification. Affinity receptor tags are typically (but not necessarily) placed at or near the N- or C-terminus of the HLA allele. Various peptide tags are well known in the art. Non-limiting examples include polyhistidine tags (e.g., 4 to 15 consecutive His residues, such as 8 consecutive His residues); polyhistidine-glycine tags; HA tags (e.g., Field et al., Mol. Cell. Biol., 8:2159, 1988); c-myc tags (e.g., Evans et al., Mol. Cell. Biol., 5:3610, 1985); herpes simplex virus glycoprotein D (gD) tags (e.g., Paborsky et al., Protein Engineering, 3:547, 1990); FLAG tag (e.g., Hopp et al., BioTechnology, 6:1204, 1988; U.S. Patent Nos. 4,703,004 and 4,851,341); KT3 epitope tag (e.g., Martine et al., Science, 255:192, 1992); tubulin epitope tag (e.g., Skinner, Biol. Chem., 266:15173, 1991); T7 gene 10 protein peptide tag (e.g., Lutz-Freyemuth et al., P . Natl. Acad. Sci. USA, 87:6393, 1990); a streptavidin tag (StrepTag™ or StrepTagII™; see, e.g., Schmidt et al., J. Mol. Biol., 255(5):753-766, 1996 or U.S. Patent No. 5,506,121; also commercially available from Sigma-Genosys); or a VSV-G epitope tag derived from the glycoprotein of vesicular stomatitis virus; or a V5 tag derived from a small epitope (Pk) found on the P and V proteins of simian virus 5 (SV5) paramyxovirus. In some embodiments, the affinity receptor tag is an "epitope tag," a type of peptide tag that adds a recognizable epitope (antibody binding site) to an HLA protein to provide binding of a corresponding antibody, thereby allowing identification or affinity purification of the tagged protein. Non-limiting examples of epitope tags are protein A or protein G, which can bind to IgG. In some embodiments, the matrix of IgG Sepharose 6 Fast Flow chromatography resin is covalently coupled to human IgG. This resin allows for high flow rates, rapid, and convenient purification of proteins labeled with Protein A.Many other tag moieties are known and contemplated by those of ordinary skill and are contemplated herein. Any peptide tag can be used so long as it is capable of being expressed as a component of an affinity receptor-tagged HLA-peptide complex.

[0429] As used herein, the term "affinity molecule" refers to a molecule or ligand that binds to an affinity receptor peptide with chemical specificity. Chemical specificity is the ability of a protein binding site to bind a specific ligand. The fewer ligands a protein can bind, the higher its specificity. Specificity describes the strength of the binding between a given protein and a ligand. This relationship can be described by the dissociation constant (KD), which characterizes the equilibrium between the bound and unbound states of a protein-ligand system.

[0430] The term "affinity receptor-tagged HLA-peptide complex" refers to a complex comprising an HLA class I or class II associated peptide or portion thereof that specifically binds to a monoallelic recombinant HLA class I or class II peptide comprising an affinity receptor peptide.

[0431] The terms "specific binding" or "specific binding" when applied to the interaction of an affinity molecule and an affinity receptor tag or epitope with an HLA peptide means that the interaction is dependent on the presence of a particular structure (e.g., an antigenic determinant or epitope) on the protein; in other words, the affinity molecule recognizes and binds to a specific affinity receptor peptide structure rather than to the protein as a whole.

[0432] As used herein, the term "affinity" refers to a measure of the strength of binding between two members of a binding pair (e.g., an "affinity receptor tag" and an "affinity molecule," and an HLA binding peptide and an HLA class I or class II molecule). KD is the dissociation constant and has units of molar concentration. The affinity constant is the reciprocal of the dissociation constant. The affinity constant is sometimes used as a general term to describe the chemical entity. It is a direct measure of the binding energy. Affinity can be determined experimentally by surface plasmon resonance (SPR), for example, using a commercially available Biacore SPR unit. Affinity can also be expressed as an inhibitory concentration 50 (IC50), i.e., the concentration at which 50% of the peptide is displaced. Similarly, lnIC50 refers to the natural logarithm of the IC50. K off Refers to the dissociation rate constant, for example, the rate constant for dissociation of an affinity molecule from an affinity receptor-tagged HLA-peptide complex.

[0433] In some embodiments, affinity receptor-tagged HLA-peptide complexes contain a biotin acceptor peptide (BAP) and are immunopurified from the complex cell mixture using streptavidin / NeutrAvidin beads. Biotin-avidin / streptavidin binding is the strongest non-covalent interaction known in nature. This property is used in a wide range of applications as a biological tool, such as immunopurification of proteins covalently linked to biotin. In an exemplary embodiment, a nucleic acid sequence encoding an HLA allele uses a biotin acceptor peptide (BAP) as an affinity receptor tag for immunopurification. BAP can be specifically biotinylated in vivo or in vitro at a single lysine residue within the tag (e.g., U.S. Patents 5,723,584; 5,874,239; and 5,932,433; and British Patent GB2370039). BAP is typically 15 amino acids long and contains a lysine as a biotin acceptor residue. In some embodiments, the BAP is placed at or near the N- or C-terminus of a monoallelic HLA peptide. In some embodiments, the BAP is positioned between the heavy chain domain and the β2 microglobulin domain of an HLA class I peptide. In some embodiments, the BAP is positioned between the β chain domain and the α chain domain of an HLA class II peptide. In some embodiments, the BAP is positioned in the loop region between the α1, α2, and α3 domains of an HLA class I heavy chain, or between the α1 and α2 and β1 and β2 domains of an HLA class II α and β chain, respectively. Figure 2 Exemplary constructs designed for HLA class I and II expression using BAP for biotinylation and immunopurification are described in .

[0434] As used herein, the term "biotin" refers to the compound biotin itself and its analogs, derivatives, and variants. Thus, the term "biotin" includes biotin (cis-hexahydro-2-oxo-1H-thieno[3,4]imidazole-4-pentanoic acid) and any derivatives and analogs thereof, including biotin-like compounds. Such compounds include, for example, amino or sulfhydryl derivatives of biotin-eN-lysine, biocytin hydrazide, 2-iminobiotin, and biotinyl-E-aminohexanoic acid-N-hydroxysuccinimide ester, sulfosuccinimidyl iminobiotin, biotin bromoacetyl hydrazide, p-diazonium benzoyl biocytin, 3-(N-maleimidopropionyl) biocytin, desthiobiotin, and the like. The term "biotin" also includes biotin variants that can specifically bind to one or more of Rhizavidin, avidin, streptavidin, tamavidin moieties, or other avidin-like peptides.

[0435] As used herein, a "PPV determination method" may refer to a presentation PPV determination method. For example, a "PPV determination method" may refer to a method comprising the steps of: (a) processing amino acid information for a plurality of test peptide sequences using an HLA peptide presentation prediction model, such as a machine learning HLA peptide presentation prediction model, to generate a plurality of test presentation predictions, each test presentation prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell (such as a class II HLA allele of a cell of a subject) are capable of presenting a given test peptide sequence of the plurality of test peptide sequences, wherein the plurality of test peptide sequences comprises at least 500 test peptide sequences, the test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, and (ii) at least 499 hit peptide sequences included in a peptide sequence expressed by an organism (e.g., a cell of the same species as the subject) (b) identifying or determining a top percentage of the plurality of test peptide sequences, such as the top 0.2% of the plurality of test peptide sequences, as presented by the class II HLA alleles of the cell; and (c) calculating the PPV of the HLA peptide presentation prediction model, wherein the PPV is the fraction of the plurality of test peptide sequences that are identified or determined as presented by the class II HLA alleles of the cell, where these peptides are peptides observed to be presented by the class II HLA alleles of the cell by mass spectrometry. In some embodiments, the bait peptides have the same length, i.e., contain the same number of amino acids as the hit peptides. In some embodiments, the bait peptides may contain one more or one less amino acid than the hit peptides. In some embodiments, the bait peptides are peptides that are endogenous peptides. In some embodiments, the bait peptide is a synthetic peptide. In some embodiments, the bait peptide is an endogenous peptide that has been identified by mass spectrometry as binding to a first MHC class I or class II protein, wherein the first MHC class I or class II protein is different from a second MHC class I or class II protein that binds to the hit peptide. In some embodiments, the bait peptide can be a scrambled peptide, for example, the bait peptide can comprise an amino acid sequence wherein the positions of the amino acids are rearranged relative to the amino acid positions of the hit peptides over the length of the peptide. In some embodiments, the PPV determination method can be a presentation PPV determination method. In some embodiments, the ratio of the number of hit peptide sequences to the number of bait peptide sequences is about 1:10, 1:20, 1:50, 1:100, 1:250, 1:500, 1:1000, 1:1500, 1:2000, 1:2500, 1:5000, 1:7500, 1:10000, 1:25000,1:50000 or 1:100000. In some embodiments, the at least one hit peptide sequence comprises at least 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, 94, 95, 96, 97, 98, 99, or 100 hit peptide sequences. In some embodiments, the at least 499 decoy peptide sequences comprise at least 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700 700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 8700, 8800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 10 000、11000、12000、13000、14000、15000、16000、17000、18000、19000、20000、21000、22000、23000、24000、25000、26000、27000、28000、29000、30000、31000、32000、33000、34000、35000、36000、37000、38000、39000、40000、41000、42000、43000、44000、45000、46000、47000、48000、49000、50000、52500、55000、57500、60000、62500、65000、67500、70000、72500、75000、77500、80000、82500、85000、87500、90000、92 or 1,000,000, 500, 95,000, 97,500, 100,000, 125,000, 150,000, 175,000, 200,000, 225,000, 250,000, 275,000, 300,000, 325,000, 350,000, 375,000, 400,000, 425,000, 450,000, 475,000, 500,000, 600,000, 700,000, 800,000, 900,000, or 1,000,000 bait peptide sequences. In some embodiments, the at least 500 test peptide sequences comprise at least 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600 00, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 8700, 8800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800 , 9900, 10000, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000, 19000, 20000, 21000, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000,36000、37000、38000、39000、40000、41000、42000、43000、44000、45000、46000、47000、48000、49000、50000、52500、55000、57500、60000、62500、65000、67500、70000、72500、75000、77500、80000、82500、85000、87500、90 000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 test peptide sequences. In some embodiments, identifying or calling a top percentage of the plurality of test peptide sequences as presented by the class II HLA alleles of the cell comprises identifying or calling a top 0.20%, 0.30%, 0.40%, 0.50%, 0.60%, 0.70%, 0.80%, 0.90%, 1.00%, 1.10%, 1.20%, 1.30%, 1.40%, 1.50%, 1.60%, 1.70%, 1.80%, 1.90%, 1.01%, 1.12%, 1.13%, 1.14%, 1.15%, 1.16%, 1.17%, 1.18%, 1.19% or more of the plurality of test peptide sequences as presented by the class II HLA alleles of the cell. 0%, 2.00%, 2.10%, 2.20%, 2.30%, 2.40%, 2.50%, 2.60%, 2.70%, 2.80%, 2.90%, 3.00%, 3.10%, 3.20%, 3.30%, 3.40%, 3.50%, 3.60%, 3.70%, 3.80%, 3.90%, 4.00%, 4.10%, 4.20%, 4.30%, 4.40%, 4.50%, 4.60%, 4.70%, 4.80%, 4.90%, 5.00%, 5.10%, 5.20%, 5.30%, 5.40%, 5.50%, 5.60%, 5.70%, 5.80%, 5.90%, 6.00%, 6.10%, 6.20%, 6.30%, 6.40%, 6.50%, 6.60%, 6.70%, 6.80%, 6.90%, 7.00%, 7.10%, 7.2 0%, 7.30%, 7.40%, 7.50%, 7.60%, 7.70%, 7.80%, 7.90%, 8.00%, 8.10%, 8.20%, 8.30%, 8.40%, 8.50%, 8.60%, 8.70%, 8.80%, 8.90%, 9.00%, 9.10%, 9.20%, 9.30%, 9.40%, 9.50%, 9.60%, 9.70%, 9.80%,9. 90%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19% or 20% are identified or determined to be presented by the cell's class II HLA allele. In some embodiments, the cell is a monoallelic cell.

[0436] As used herein, a "PPV determination method" may refer to a binding PPV determination method. For example, a "PPV determination method" may refer to a method comprising the following steps: (a) processing amino acid information of a plurality of test peptide sequences using an HLA peptide binding prediction model, such as a machine learning HLA peptide binding prediction model, to generate a plurality of binding presentation predictions, each binding presentation prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell (such as a class II HLA allele of a subject's cell) will bind to a given test peptide sequence of the plurality of test peptide sequences, wherein the plurality of test peptide sequences comprises at least 20 test peptide sequences, the test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, and (ii) at least 19 decoy peptide sequences contained within a protein, the protein (a) comprising at least one peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in a cell, wherein the ratio of the number of hit peptide sequences to the number of bait peptide sequences in the plurality of test peptide sequences is less than 1, for example, the ratio of the at least one hit peptide sequence to the at least 19 bait peptide sequences is 1:19; (b) identifying or determining a top percentage of the plurality of test peptide sequences, such as the top 5% of the plurality of test peptide sequences, as binding to the HLA protein; and (c) calculating the PPV of the HLA peptide binding prediction model, wherein the PPV is the fraction of the test peptide sequences in the plurality of test peptide sequences identified or determined as binding to a class II HLA allele of the cell, which peptides are peptides observed to be presented by the class II HLA allele of the cell by mass spectrometry. In some embodiments, the ratio of the number of hit peptide sequences to the number of bait peptide sequences is about 1:2, 1:3, 1:4, 1:5, 1:10, 1:20, 1:25, 1:30, 1:40, 1:50, 1:75, 1:100, 1:200, 1:250, 1:500, or 1:1000. In some embodiments, the at least one hit peptide sequence comprises at least 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 or 100 hit peptide sequences. In some embodiments, the at least 19 decoy peptide sequences comprise at least 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800 400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800 , 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 630 0, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 8700, 8 800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 10000, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000, 19000, 20000 , 21000, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41 000、42000、43000、44000、45000、46000、47000、48000、49000、50000、52500、55000、57500、60000、62500、65000、67500、70000、72500、75000、77500、00,000, 800,000, 825,000, 850,000, 875,000, 900,000, 925,000, 950,000, 975,000, 100,000, 125,000, 150,000, 175,000, 200,000, 225,000, 250,000, 275,000, 300,000, 325,000, 350,000, 375,000, 400,000, 425,000, 450,000, 475,000, 500,000, 600,000, 700,000, 800,000, 900,000 or 1,000,000 bait peptide sequences. In some embodiments, the at least 20 test peptide sequences comprise at least 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470 , 480, 490, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300 , 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8 300、8400、8500、8600、8700、8800、8900、9000、9100、9200、9300、9400、9500、9600、9700、9800、9900、10000、11000、12000、13000、14000、15000、16000、17000、18000、19000、20000、21000、22000、23000、24000、25000、26000、27000、28000、29000, 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41000, 42000, 43000, 44000, 45000, 46000, 47000, 48000, 49000, 50000, 52500, 55000, 57500, 60000, 62500, 65000, 67500, 70000, 72500, 75000, 77500, 80000 , 82500, 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 10000000 test peptide sequences. In some embodiments, identifying or calling a top percentage of the plurality of test peptide sequences as being presented by an HLA class II allele of the cell comprises identifying or calling the top 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%, or 40% as being presented by an HLA class II allele of the cell. In some embodiments, the cell is a monoallelic cell.

[0437] Human Leukocyte Antigen (HLA) System

[0438] The immune system can be divided into two functional subsystems: the innate immune system and the adaptive immune system. The innate immune system is the first line of defense against infection, rapidly neutralizing most potential pathogens before they can cause an overt infection. The adaptive immune system responds to molecular structures (called antigens) in invading organisms. Unlike the innate immune system, the adaptive immune system is highly specific to pathogens. Adaptive immunity can also provide long-lasting protection; for example, a person who has recovered from measles is now protected from measles for life. There are two types of adaptive immune responses: humoral and cell-mediated. In humoral responses, antibodies secreted into body fluids by B cells bind to pathogen-derived antigens, leading to pathogen elimination through various mechanisms, such as complement-mediated lysis. In cell-mediated responses, T cells, which are capable of destroying other cells, are activated. For example, if disease-associated proteins are present in cells, they are broken down into peptides by proteolysis within the cell. Specific cellular proteins then attach themselves to the antigens or peptides thus formed and transport them to the cell surface, where they are presented to the body's molecular defense mechanisms in T cells. Cytotoxic T cells recognize these antigens and kill cells bearing the antigen.

[0439] The terms "major histocompatibility complex (MHC)", "MHC molecule" or "MHC protein" refer to proteins that are capable of binding peptides generated by proteolytic cleavage of protein antigens and representing potential T cell epitopes, transporting them to the cell surface, and presenting the peptides to specific cells, such as cytotoxic T lymphocytes or T helper cells. The human MHC is also known as the HLA complex. Thus, the terms "human leukocyte antigen (HLA) system", "HLA molecule" or "HLA protein" refer to the gene complex that encodes the human MHC protein. The term MHC is referred to as the "H-2" complex in murine species. One of ordinary skill in the art will recognize that the terms "major histocompatibility complex (MHC)", "MHC molecule", "MHC protein" and "human leukocyte antigen (HLA) system", "HLA molecule", "HLA protein" are used interchangeably herein.

[0440] HLA proteins are classified into two types, known as HLA class I and HLA class II. The structures of the proteins of the two HLA classes are very similar; however, they have very different functions. HLA class I proteins are found on the surface of almost all cells in the body, including most tumor cells. HLA class I proteins are loaded with antigens, which usually originate from endogenous proteins or pathogens present within the cell, and are then presented to naive or cytotoxic T lymphocytes (CTLs). HLA class II proteins are found on antigen-presenting cells (APCs), which include but are not limited to dendritic cells, B cells, and macrophages. They primarily present peptides from external antigen sources, such as peptides processed outside the cell, to helper T cells. Most peptides bound by HLA class I proteins originate from cytoplasmic proteins produced in the organism's own healthy host cells and generally do not stimulate an immune response.

[0441] HLA class I molecules (Figure 1) consist of two non-covalently linked polypeptide chains: an HLA-encoded α chain (heavy chain, 44 to 47 kD) and a non-HLA-encoded subunit, termed β2-microglobulin (or β2m), (12 kD). The α chain has three extracellular domains—α1, α2, and α3—and a transmembrane region, with the α1 and α2 regions capable of binding peptides of approximately 7 to 13 amino acids (e.g., approximately 8 to 11 amino acids, or 9 or 10 amino acids). HLA class 1 molecules bind to peptides with appropriate binding motifs and present them to cytotoxic T lymphocytes. HLA class 1 heavy chains can be the protein product of an HLA-A allele, also known as an HLA-A monomer, or the protein product of an HLA-B allele (also known as an HLA-B monomer), or the protein product of an HLA-C allele (HLA-C monomer), each of which is complexed with β2-microglobulin. α1 is dependent on the non-HLA protein β2m; β2m is encoded by the β-2-microglobulin gene located on human chromosome 15. The α3 domain is linked to the transmembrane region, anchoring the HLA class I molecule to the cell membrane. Presented peptides are held in the central region of the α1 / α2 heterodimer (a molecule composed of two different subunits) by the base of the peptide-binding groove. HLA class IA, HLA class IB, or HLA class IC are highly polymorphic. The HLA class 1-A gene (referred to as the HLA-A gene), the HLA class 1-B gene (referred to as the HLA-B gene), and the HLA class 1-C gene (referred to as the HLA-C gene) each contain eight exons: exon 1 encodes the leader peptide, exons 2 and 3 encode the α1 and α2 domains, exon 5 encodes the transmembrane region, and exons 6 and 7 encode the cytoplasmic tail. Polymorphisms in exons 2 and 3 determine the peptide-binding specificity of each class 1 molecule. HLA-B class genes (HLA-B) have many possible variations, expression patterns, and antigens presented. This group is further divided into those encoded within the HLA locus, such as HLA-E, HLA-F, HLA-G, and those not encoded therein, such as stress ligands, such as ULBP, Rae1, and H60. The antigen / ligand of many of these molecules remains unknown, but they can interact with CD8+ T cells, NKT cells, and NK cells.

[0442] In some embodiments, the present disclosure utilizes non-classical HLA class IE alleles. HLA-E molecules are recognized by natural killer (NK) cells and CD8+ T cells. HLA-E is expressed in almost all tissues, including lung, liver, skin, and placental cells. HLA-E expression has also been detected in solid tumors (e.g., osteosarcoma and melanoma). HLA-E molecules bind to TCRs expressed on CD8+ T cells, thereby leading to T cell activation. HLA-E is also known to bind to CD94 / NKG2 receptors expressed on NK cells and CD8+ T cells. CD94 can be paired with several different isoforms of NKG2 to form receptors with the potential to inhibit (NKG2A, NKG2B) or promote (NKG2C) cell activation. HLA-E can bind to peptides derived from amino acid residues 3-11 of the leader sequences of most HLA-A, -B, -C, and -G molecules, but cannot bind to its own leader peptide. It has also been shown that HLA-E presentation is derived from peptides of endogenous proteins similar to HLA-A, -B and -C alleles. Under physiological conditions, CD94 / NKG2A and the engagement of the HLA-E loaded with peptides from HLA class I leader sequences usually induce inhibitory signals. Cytomegalovirus (CMV) utilizes the mechanism of escaping NK cell immune surveillance by expressing UL40 glycoprotein (simulating HLA-A leader sequence). However, it has also been reported that CD8+ T cells can identify the HLA-E loaded with the UL40 peptide derived from the CMV Toledo strain, and work in defense CMV. A large number of studies have revealed several important functions of HLA-E in infectious diseases and cancer.

[0443] Before presentation on the cell surface, peptide antigens attach themselves to HLA class I molecules through competitive affinity binding within the endoplasmic reticulum. The affinity of an individual peptide antigen is directly related to its amino acid sequence and the presence of specific binding motifs at defined positions within the amino acid sequence. If the sequence of such a peptide is known, peptide vaccines, for example, can be used to manipulate the immune system against diseased cells.

[0444] MHC molecules are highly polymorphic, meaning that many MHC variants exist. Each variant is encoded by a variant of the protein-encoding gene, and each such variant gene is called an allele. In humans, MHC is known as human leukocyte antigen (HLA), which involves three types of HLA class II molecules: DP, DQ, and DR. HLA class II peptides (Figure 1) have two chains: α and β, each with two domains—α1 and α2, and β1 and β2—each with a transmembrane domain: α2 and β2, respectively, that anchor the HLA class II molecule to the cell membrane. The peptide-binding groove is formed by heterodimers of α1 and β1. The most widely studied HLA-DR molecule has a DRA and DRB, corresponding to the α and β domains, respectively. While the DRB is diverse, the DRA is nearly identical. Therefore, the binding specificity of a DRB allele dictates the binding specificity of the corresponding HLA-DR. Each MHC protein has its own binding specificity, meaning that the set of peptides bound to an MHC molecule may differ from the peptides bound to another MHC molecule. Classical molecules present peptides to CD4+ lymphocytes. Non-classical molecules with intracellular functions, annexins, are not exposed on the cell membrane but rather in the inner membrane of lysosomes, typically loading antigenic peptides onto classical HLA class II molecules.

[0445] In the HLA class II system, phagocytic cells such as macrophages and immature dendritic cells ingest entities through phagocytosis into phagosomes—although B cells more commonly exhibit endocytosis into endosomes. The phagosomes fuse with lysosomes, where acidic enzymes cleave the ingested proteins into a variety of peptides. Autophagy is another source of HLA class II peptides. Through physicochemical interactions with host-borne HLA class II variants (encoded in the host genome), specific peptides become immunodominant and loaded onto HLA class II molecules. They are then transported to and externalized at the cell surface. The most studied subclasses of HLA class II genes are: HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA-DQB1, HLA-DRA, and HLA-DRB1.

[0446] Presentation of peptides by HLA class II molecules to CD4+ helper T cells is essential for immune responses to foreign antigens (Roche and Furuta, 2015). Once activated, CD4+ T cells promote B cell differentiation and antibody production, as well as CD8+ T cell (CTL) responses. CD4+ T cells also secrete cytokines and chemokines that activate and induce the differentiation of other immune cells. HLA class II molecules are heterodimers of α and β chains that interact to form a peptide binding groove that is more open than the HLA class I peptide binding groove (Unanue et al., 2016). Peptides that bind to HLA class II molecules are thought to have a 9-amino acid binding core with flanking residues on the N-terminal or C-terminal side that protrude from the binding groove (Jardetzky et al., 1996; Stern et al., 1994). These peptides are typically 12-16 amino acids in length and usually contain 3-4 anchor residues at positions P1, P4, P6 / 7, and P9 of the binding moiety (Rossjohn et al., 2015).

[0447] HLA alleles are expressed in a codominant manner, meaning that the alleles (variants) inherited from both parents are expressed equally. For example, each person carries two alleles for each of the three class I genes (HLA-A, HLA-B, and HLA-C), so six different types of HLA class II can be expressed. In the HLA class II locus, each person inherits a pair of HLA-DP genes (DPA1 and DPB1, encoding the α and β chains), HLA-DQ (DQA1 and DQB1 for the α and β chains), one HLA-DRα gene (DRA1), and one or more HLA-DRβ genes (DRB1 and DRB3, -4, or -5). For example, HLA-DRB1 has more than nearly 400 known alleles. This means that a heterozygous individual can inherit six or eight functional HLA class II alleles, three or more from each parent. Therefore, HLA genes are highly polymorphic; many different alleles exist in different individuals within a population. The genes encoding the HLA proteins have many possible variations, allowing each person's immune system to respond to a wide variety of foreign invaders. Some HLA genes have hundreds of identified forms (alleles), each of which is given a specific number. In some embodiments, the HLA class I alleles are HLA-A*02:01, HLA-B*14:02, HLA-A*23:01, and HLA-E*01:01 (nonclassical). In some embodiments, the HLA class II alleles are HLA-DRB*01:01, HLA-DRB*01:02, HLA-DRB*11:01, HLA-DRB*15:01, and HLA-DRB*07:01.

[0448] The subject's subject-specific HLA alleles or HLA genotype can be determined by any method known in the art. In an exemplary embodiment, the HLA genotype is determined by any method described in International Patent Application PCT / US2014 / 068746 (published as WO2015085147 on June 11, 2015), which is incorporated herein by reference in its entirety. Briefly, the method includes determining a polymorphic gene type, which may include generating an alignment of reads extracted from a sequencing data set to a gene reference set comprising allelic variants of the polymorphic gene, determining a first posterior probability or a score derived from the posterior probability for each allelic variant in the alignment, identifying the allelic variant with the largest first posterior probability or score derived from the posterior probability as a first allelic variant, identifying one or more overlapping reads aligned to the first allelic variant and one or more other allelic variants, determining a second posterior probability or a score derived from the posterior probability for the one or more other allelic variants using a weighting factor, determining a second allelic variant by selecting the allelic variant with the largest second posterior probability or score derived from the posterior probability, the first and second allelic variants defining the gene type of the polymorphic gene, and providing an output of the first and second allelic variants.

[0449] In some embodiments, MHC II class peptide described herein: antigen peptide combination and presentation prediction method has the ability of predicting binding substances from the large group library MHC II class peptides encoded by independent HLA allele.In some embodiments, the large database of HLA matching peptides verified by MAPTAC technology is trained with mass spectrometry.In some embodiments, the large database of HLA matching peptides verified by mass spectrometry comprises more than 1.2x10^6 such HLA matching peptides.In some embodiments, the large database of HLA matching peptides verified by mass spectrometry covers more than 150 HLA alleles, including MHC I class and II class allele hypotypes.In some embodiments, this database covers at least 95% U.S. population for HLA-I and HLA-II (DR hypotype).

[0450] As described in this article, there is ample evidence in animals and humans that mutated epitopes can effectively induce immune responses, that spontaneous tumor regression or long-term survival is associated with CD8+ T cell responses to mutated epitopes, and that "immunoediting" can track changes in the expression of dominant mutant antigens in mice and humans.

[0451] Sequencing techniques reveal that each tumor contains multiple patient-specific mutations that alter the protein-coding content of genes. Such mutations produce altered proteins ranging from single amino acid changes (caused by missense mutations) to long regions that add new amino acid sequences due to frameshifts, readthrough of stop codons, or translation of intronic regions (new open reading frame mutations; neoORFs). These mutant proteins are valuable targets for the host's immune response to tumors because, unlike natural proteins, they are not subject to the immunosuppressive effects of self-tolerance. Therefore, mutated proteins are more likely to be immunogenic and more specific to tumor cells than to the patient's normal cells. In essence, short peptides (8-24 amino acids long) containing cancer-associated mutations are candidates for cancer immunotherapy.

[0452] In some embodiments, the algorithm of the driver prediction method can be further used to perform mutation determination on the peptide. In some embodiments, the prediction method can be used to determine the driver mutation status, and / or RNA expression status, and / or cleavage prediction within the peptide.

[0453] The term "T cell" includes CD4+ T cells and CD8+ T cells. The term T cell also includes T helper type 1 T cells and T helper type 2 T cells. As used herein, T cells are generally classified into two major categories: helper T (TH) cells and cytotoxic T lymphocytes (CTLs) based on their function and the cell surface antigens (cluster differentiation antigens or CD) that also help the T cell receptor bind to the antigen.

[0454] Mature helper T (TH) cells express the surface protein CD4 and are called CD4+ T cells. After T cell development, mature immature T cells leave the thymus and begin to spread throughout the body, including lymph nodes. Immature T cells are T cells that have never been exposed to the antigens to which they are programmed to respond. Like all T cells, they express the T cell receptor-CD3 complex. The T cell receptor (TCR) consists of a constant region and a variable region. The variable region determines to which antigens the T cell can respond. CD4+ T cells have a TCR with affinity for MHC class II proteins, and CD4 participates in determining MHC affinity during thymic maturation. MHC class II proteins are usually only found on the surface of specialized antigen presenting cells (APCs). Specialized antigen presenting cells (APCs) are mainly dendritic cells, macrophages, and B cells, although dendritic cells are the only group of cells that constitutively (always) express MHC class II. Some APCs also bind native (or unprocessed) antigens to their surfaces, such as follicular dendritic cells, but unprocessed antigens do not interact with T cells and do not participate in their activation. Peptide antigens bound to HLA class I proteins are generally shorter than those bound to HLA class II proteins.

[0455] Cytotoxic T lymphocytes (CTLs), also known as cytotoxic T cells, cytolytic T cells, CD8+ T cells or killer T cells, refer to lymphocytes that induce apoptosis in targeted cells. CTLs interact with antigens (Ag) processed on the surface of TCR and target cells to form antigen-specific conjugates with target cells, thereby causing apoptosis of target cells. Apoptotic bodies are eliminated by macrophages. The term "CTL response" is used to refer to the primary immune response mediated by CTL cells. Cytotoxic T lymphocytes have both T cell receptors (TCR) and CD8 molecules on their surface. The T cell receptor can recognize and bind to peptides complexed with HLA class I molecules. Each cytotoxic T lymphocyte expresses a unique T cell receptor that can bind to specific MHC / peptide complexes. Most cytotoxic T cells express T cell receptors (TCRs) that can recognize specific antigens. In order for TCR to bind to HLA class I molecules, the former must be accompanied by a glycoprotein called CD8, which binds to the constant part of HLA class I molecules. Therefore, these T cells are called CD8+ T cells. The affinity between CD8 and MHC molecules brings T cells and target cells together during antigen-specific activation. Once activated, CD8+ T cells are recognized as T cells and are generally classified as having a predetermined cytotoxic role in the immune system. However, CD8+ T cells also have the ability to produce certain cytokines.

[0456] "T cell receptor (TCR)" is a cell surface receptor involved in the activation of T cells in response to antigen presentation. TCR is usually composed of two chains, α and β, which are assembled to form a heterodimer and associate with the CD3 transduction subunit to form a T cell receptor complex present on the cell surface. Each α and β chain of TCR consists of immunoglobulin-like N-terminal variable (V) and constant (C) regions, a hydrophobic transmembrane domain and a short cytoplasmic region. As for immunoglobulin molecules, the variable regions of α and β chains are produced by V(D)J recombination, thereby producing a great diversity of antigen specificity in T cell populations. However, compared to immunoglobulins that recognize complete antigens, T cells are activated by processed peptide fragments associated with MHC molecules, thereby introducing an additional dimension to T cell recognition of antigens, which is called MHC restriction. Recognition of MHC differences between donor and recipient by T cell receptors can lead to T cell proliferation and the potential development of GVHD. It has been shown that normal surface expression of the TCR depends on the coordinated synthesis and assembly of all seven components of the complex (Ashwell and Klusner 1990). Inactivation of TCRα or TCRβ can lead to the elimination of the TCR from the T cell surface, thereby preventing recognition of alloantigens and thus preventing GVHD. However, TCR disruption usually leads to the elimination of CD3 signaling components and alters the pattern of further T cell expansion.

[0457] The term "HLA peptidome" refers to a set of peptides that specifically interact with a particular HLA class and can contain thousands of different sequences. The HLA peptidome includes a diverse set of peptides derived from normal and abnormal proteins expressed in cells. Therefore, the HLA peptidome can be studied to identify cancer-specific peptides for the development of tumor immunotherapy and as a source of information about protein synthesis and degradation programs within cancer cells. In some embodiments, the HLA peptidome is a set of soluble HLA peptides (sHLA). In some embodiments, the HLA peptidome is a set of membrane-bound HLA (mHLA).

[0458] "Antigen presenting cells" or "APCs" include professional antigen presenting cells (e.g., B lymphocytes, macrophages, monocytes, dendritic cells, Langerhans cells), as well as other antigen presenting cells (e.g., keratinocytes, endothelial cells, astrocytes, fibroblasts, oligodendrocytes, thymic epithelial cells, thyroid epithelial cells, glial cells (brain), pancreatic beta cells, and vascular endothelial cells). "Antigen presenting cells" or "APCs" are cells that express major histocompatibility complex (MHC) molecules and can display foreign antigens complexed with MHC on their surface.

[0459] Monoallelic HLA cell lines

[0460] Monoallelic cell lines expressing a single HLA class I allele, a pair of HLA class II alleles, or a single HLA class I allele and a pair of HLA class II alleles can be generated by transducing or transfecting a suitable cell population with a polynucleic acid (e.g., a vector) encoding a single HLA allele. Figure 2). Suitable cell populations include, for example, HLA class I-deficient cell lines in which a single HLA class I allele is exogenously expressed, HLA class II-deficient cell lines in which a pair of HLA class II alleles is exogenously expressed, or class I and class I type II-deficient cell lines in which a single HLA class I and / or a pair of class II alleles are exogenously expressed. As an exemplary embodiment, the HLA class I-deficient B cell line is B721.221. However, it is clear to the skilled person that other cell populations with HLA class I and / or HLA class II defects can be generated. Exemplary methods for deleting / inactivating endogenous HLA class I or HLA class II genes include CRISPR-Cas9-mediated genome editing, for example in THP-1 cells. In some embodiments, the cell population is a professional antigen-presenting cell, such as a macrophage, a B cell, and a dendritic cell. The cell may be a B cell or a dendritic cell. In some embodiments, the cell is a tumor cell or a cell from a tumor cell line. In some embodiments, the cell is isolated from a patient. In some embodiments, the cells contain an infectious agent or a portion thereof. In some embodiments, the cell population comprises at least 10 7 cells. In some embodiments, the cell population is further modified, for example, by increasing or decreasing the expression and / or activity of at least one gene. In some embodiments, the gene encodes a member of the immunoproteasome. It is known that the immunoproteasome is involved in the processing of HLA class I binding peptides and includes LMP2 (β1i), MECL-1 (β2i) and LMP7 (β5i) subunits. The immunoproteasome can also be induced by interferon-γ. Therefore, in some embodiments, the cell population can be contacted with one or more cytokines, growth factors or other proteins. The cells can be stimulated with inflammatory cytokines such as interferon-γ, IL-10, IL-6 and / or TNF-α. The cell population can also be subjected to various environmental conditions, such as stress (heat stress, hypoxia, glucose starvation, DNA damaging agents, etc.). In some embodiments, the cells are contacted with one or more of chemotherapeutic drugs, radiotherapy, targeted therapy or immunotherapy. Therefore, the methods disclosed herein can be used to study the effects of various genes or conditions on HLA peptide processing and presentation. In some embodiments, the conditions used are selected to match the condition of the patient whose HLA-peptide population is to be identified.

[0461] A virus-based system (e.g., an adenovirus system, an adeno-associated virus (AAV) vector, a poxvirus, or a lentivirus) can be used to encode and express a single HLA allele of the present disclosure. Plasmids that can be used for adeno-associated virus, adenovirus, and lentivirus delivery have been previously described (see, e.g., U.S. Patents 6,955,808 and 6,943,019, and U.S. Patent Application 20080254008, incorporated herein by reference). In vectors that can be used in the practice of the present disclosure, integration into the host genome of the cell can be achieved using retroviral gene transfer methods, typically resulting in long-term expression of the inserted transgene. In an exemplary embodiment, the retrovirus is a lentivirus. In addition, high transduction efficiencies have been observed in many different cell types and target tissues. The tropism of a retrovirus can be altered by incorporating foreign envelope proteins and expanding the potential target population of target cells. Retroviruses can also be engineered to allow conditional expression of the inserted transgene so that only certain cell types are infected by the lentivirus. Cell type-specific promoters can be used to target expression in specific cell types. Lentiviral vectors are retroviral vectors (thus both lentiviral and retroviral vectors can be used in the practice of the present disclosure). In addition, lentiviral vectors are able to transduce or infect non-dividing cells and typically produce high viral titers.

[0462] The choice of retroviral gene transfer system can depend on the target tissue. Retroviral vectors are composed of cis-acting long terminal repeats (LTRs) with the capacity to package up to 6-10 kb of foreign sequence. The minimal cis-acting LTRs are sufficient for replication and packaging of the vector, which is then used to integrate the desired nucleic acid into the target cell to provide permanent expression. Widely used retroviral vectors that can be used in the practice of the present disclosure include vectors based on murine leukemia virus (MuLV), gibbon ape leukemia virus (GaLV), simian immunodeficiency virus (SIV), human immunodeficiency virus (HIV), and combinations thereof (see, e.g., Buchscher et al. (1992) J. Virol. 66:2731-2739; Johann et al. (1992) J. Virol. 66:1635-1640; Sommnerfelt et al. (1990) Virol. 176:58-59; Wilson et al. (1998) J. Virol. 63:2374-2378; Miller et al. (1991) J. Virol. 65:2220-2224; PCT / US94 / 05700). Additionally, minimal non-primate lentiviral vectors, such as lentiviral vectors based on equine infectious anemia virus (EIAV), are useful in the practice of the present disclosure (see, e.g., Balagaan, (2006) J Gene Med; 8:275-285, published online on November 21, 2005, Wiley InterScience DOI: 10.1002 / jgm.845). The vectors may have a cytomegalovirus (CMV) promoter driving target gene expression. Thus, the present disclosure relates to one or more vectors useful in practicing the present disclosure: viral vectors, including retroviral vectors and lentiviral vectors.

[0463] Any HLA allele can be expressed in a cell population. In an exemplary embodiment, the HLA allele is an HLA class I allele. In some embodiments, the HLA class I allele is an HLA-A allele or an HLA-B allele. In some embodiments, the HLA allele is an HLA class II allele. Sequences of HLA class I and class II alleles can be found in the IPD-IMGT / HLA database. Exemplary HLA alleles include, but are not limited to, HLA-A*02:01, HLA-B*14:02, HLA-A*23:01, HLA-E*01:01, HLA-DRB*01:01, HLA-DRB*01:02, HLA-DRB*11:01, HLA-DRB*15:01, and HLA-DRB*07:01.

[0464] In some embodiments, an HLA allele is selected to correspond to a genotype of interest. In some embodiments, the HLA allele is a mutant HLA allele, which can be an allele that does not naturally occur in a patient with the disease or a naturally occurring allele. The methods disclosed herein have the further advantage of identifying HLA binding peptides for HLA alleles associated with various diseases, as well as alleles that occur at low frequencies. Thus, in some embodiments, the methods provided herein can identify an HLA allele even if it is present at a frequency of less than 1% in a population, such as a Caucasian population.

[0465] In some embodiments, the nucleic acid sequence encoding the HLA allele further comprises an affinity receptor tag that can be used to immunopurify the HLA protein. Suitable tags are well known in the art. In some embodiments, the affinity receptor tag is a polyhistidine tag, a polyhistidine-glycine tag, a polyarginine tag, a polyaspartic acid tag, a polycysteine ​​tag, a polyphenylalanine, a c-myc tag, a herpes simplex virus glycoprotein D (gD) tag, a FLAG tag, a KT3 epitope tag, a microtubule epitope tag, a T7 gene 10 protein peptide tag, a streptavidin tag, a streptavidin binding peptide (SPB) tag, a Strep-tag, a Strep-tag II, an albumin binding protein (ABP) tag, an alkaline phosphatase (AP) tag, a blue tongue virus tag (B-tag), a calmodulin binding peptide (CBP) tag, a chloramphenicol acetyltransferase (CAT) tag, a choline Binding domain (CBD) tag, chitin binding domain (CBD) tag, cellulose binding domain (CBP) tag, dihydrofolate reductase (DHFR) tag, galactose binding protein (GBP) tag, maltose binding protein (MBP), glutathione-S-transferase (GST), Glu-Glu (EE) tag, human influenza hemagglutinin (HA) tag, horseradish peroxidase (HRP) tag, NE-tag, HSV tag, ketosteroid isomerase (KSI) tag, KT3 tag, LacZ tag, luciferase tag, NusA tag, PDZ domain tag, AviTag, calmodulin tag, E-tag, S-tag, SBP-tag, Softag 1, Softag 3, TC tag, VSV-tag, Xpress tag, Isopeptag, SpyTag, SnoopTag, Profinity eXact tag, Protein C tag, S1-tag, S-tag, biotin-carboxyl carrier protein (BCCP) tag, green fluorescent protein (GFP) tag, small ubiquitin-like modifier (SUMO) tag, tandem affinity purification (TAP) tag, HaloTag, Nus-tag, thioredoxin tag, Fc-tag, CYD tag, HPC tag, TrpE tag, ubiquitin tag, VSV-G epitope tag derived from the glycoprotein of vesicular stomatitis virus, or V5 tag derived from a small epitope (Pk) found on the P and V proteins of simian virus 5 (SV5) paramyxovirus. In some embodiments, the affinity receptor tag is an "epitope tag," which is a type of peptide tag that adds a recognizable epitope (antibody binding site) to the HLA protein to provide binding of the corresponding antibody, thereby allowing identification or affinity purification of the tagged protein. Non-limiting examples of epitope tags are protein A or protein G, which can bind to IgG. In some embodiments, the affinity receptor tag comprises a biotin acceptor peptide (BAP) or a human influenza hemagglutinin (HA) peptide sequence.Many other tag moieties are known and contemplated by those of ordinary skill and are contemplated herein. Any peptide tag can be used so long as it is capable of being expressed as a component of an affinity receptor-tagged HLA-peptide complex.

[0466] The methods provided herein comprise isolating HLA-peptide complexes from cells transfected or transduced with affinity pull-down of HLA constructs ( Figure 3 In some embodiments, the complex can be isolated using standard immunoprecipitation techniques known in the art with commercially available antibodies. The cells can be lysed first. HLA class I-peptide complexes can be isolated using an HLA class I-specific antibody, such as the W6 / 32 antibody, while HLA class II-peptide complexes can be isolated using an HLA class II-specific antibody, such as the M5 / 114.15.2 monoclonal antibody. In some embodiments, a single (or a pair of) HLA alleles is expressed as a fusion protein with a peptide tag, and the HLA-peptide complex is isolated using a binding molecule that recognizes the peptide tag.

[0467] The method further comprises isolating the peptide from the HLA-peptide complex and sequencing the peptide. The peptide is isolated from the complex by any method known to those skilled in the art, such as acid elution. Although any sequencing method can be used, in some embodiments, a method using mass spectrometry is employed, such as liquid chromatography-mass spectrometry (LC-MS or LC-MS / MS, or HPLC-MS or HPLC-MS / MS). These sequencing methods are well known to those skilled in the art and are reviewed in Medzihradszky KF and Chalkley RJ. Mass Spectrom Rev. 2015 Jan-Feb; 34(1):43-63.

[0468] In some embodiments, the cell population expresses one or more endogenous HLA alleles. In some embodiments, the cell population is an engineered cell population lacking one or more endogenous HLA class I alleles. In some embodiments, the cell population is an engineered cell population lacking endogenous HLA class I alleles. In some embodiments, the cell population is an engineered cell population lacking one or more endogenous HLA class II alleles. In some embodiments, the cell population is an engineered cell population lacking endogenous HLA class II alleles or an engineered cell population lacking endogenous HLA class I alleles and endogenous HLA class II alleles. In some embodiments, the cell population comprises cells that have been enriched or sorted, for example, by fluorescence-activated cell sorting (FACS). In some embodiments, fluorescence-activated cell sorting (FACS) is used to sort the cell population. In some embodiments, the cell population is pre-FACS sorted for cell surface expression of HLA class I or class II or both HLA class I and class II. For example, FACS can be used to sort a population of cells for cell surface expression of HLA class I alleles, HLA class II alleles, or a combination thereof.

[0469] Methods for preparing personalized cancer vaccines

[0470] Once a specific mutation for cancer is identified, such that the mutation is present in the DNA of cancer cells but not in normal cells of the same human subject, and the mutation causes a change in one or more amino acids in the protein encoded by the DNA, the mutation can be a target for the host immune response. The natural immune response can be directed against the mutated protein, resulting in the destruction of cancer cells expressing the protein. Due to the natural tolerance response and immunocompromised environment in cancer tissues, immunotherapy is a clinical approach that attempts to enhance this immune response to overcome the body's tolerance and immunosuppression. Therefore, proteins or peptides containing the above mutations are suitable candidates for immunotherapy.

[0471] The mutated protein is taken up by professional phagocytes, which act as antigen-presenting cells (APCs), fragmented, and displayed on the cell surface as an antigen for T cell activation within an antigen-presenting complex containing major histocompatibility complex (MHC) proteins. Human MHC proteins are known as human leukocyte antigens (HLA). MHC proteins can be either MHC class I or class II, with some functional differences attributed to peptide presentation by class I or class II MHC proteins (HLA class I and HLA class II proteins). A notable distinction is that HLA class I-peptide complexes present antigens to cytotoxic CD8+ T cells, while HLA class II-peptide complexes can also activate CD4+ T cells, leading to a prolonged immune response. CD8+ T cells are essential for the cell-by-cell elimination of pathogenic cells, such as infected or tumor cells. CD4+ T cells have a more lasting effect after activation, the most important of which is the generation of immunological memory. CD4+ T cells are recruited differently depending on the type of immune threat, and multiple subsets with overlapping or distinct functions can be recruited together. This helps balance the immune response to pathogen threats. In these respects, HLA class II peptide-mediated antigen presentation enables sustained and tailored immune responses. On the other hand, HLA class II binding to peptides can be promiscuous, so nonspecific peptide binding and presentation to the immune system can lead to abnormal immune responses, such as autoimmunity.

[0472] In one aspect, the present disclosure provides methods for predicting peptides that can accurately pair or bind to specific HLA class II α and β chain heterodimers, such that high-fidelity binding of the peptide to HLA class II proteins (including α and β chain heterodimers) ensures that the specific peptide is presented to T lymphocytes, thereby eliciting a specific immune response and avoiding any cross-reactivity or immune promiscuity.

[0473] In one aspect, the present disclosure provides a method for predicting a peptide that can accurately bind to a specific HLA class II protein, so that when the peptide is therapeutically administered to a subject expressing a specific cognate HLA class II protein, the peptide can be used to activate a more sustained and potent immune response by virtue of the ability of the HLA class II protein to activate CD4+ T cells and stimulate immune memory. In some embodiments, a given peptide predicted to bind to an HLA class II protein with high specificity is a peptide comprising a mutation, wherein the mutation is prevalent in the subject's cancer or tumor cells; and the same HLA class II protein predicted to bind to the mutated peptide either (a) does not bind, or (b) binds to the corresponding non-mutated wild-type peptide with a significantly lower affinity than the mutated peptide in the subject. Preferential binding of HLA to mutated peptides is advantageous in the development of immunotherapeutics because cells expressing the wild-type peptide will be protected from immune attack by T cells reactive to the HLA-presented peptide. In some embodiments, the predicted peptide that specifically binds to an HLA class II protein is a peptide with a post-translational modification. Exemplary post-translational modifications include, but are not limited to, phosphorylation, ubiquitination, dephosphorylation, glycosylation, methylation, or acetylation. In some embodiments, the predicted peptide is post-translationally modified prior to use in immunotherapy.

[0474] In some embodiments, the immunotherapy approaches and strategies disclosed herein can also be applied to suppress unwanted immune activation, such as in autoimmune reactions. Specifically, peptides identified as potential binders to specific HLA subtypes can be tailored to bind to specific HLA molecules and induce tolerance rather than eliciting an immunogenic response.

[0475] In one aspect, the present invention provides a customized or personalized immunotherapy approach for a specific subject. Each subject or patient expresses a specific set of HLA class I and HLA class II proteins. HLA typing is a well-known technique that allows determination of the specific repertoire of HLA proteins expressed by a subject. Once the HLA heterodimers expressed by a particular subject are known, improved, sophisticated, and reliable methods as described herein for predicting peptides that bind to specific HLA class II α and β chain heterodimers with high fidelity can ensure that a specific immune response tailored specifically for that subject can be generated.

[0476] The genes encoding HLA heterodimers are highly polymorphic, with over 4,000 HLA class II allelic variants identified throughout the human population. For each of the HLA class II loci, an individual can inherit different alleles from the maternal and paternal HLA haplotypes, and each HLA class II heterodimer is composed of an α chain and a β chain. Due to the large number of α and β chain pairing combinations, particularly for the HLA-DP and HLA-DQ alleles, the population of possible HLA heterodimers is very complex. HLA class II heterodimers are translated in the endoplasmic reticulum (ER) and assembled into a stable complex with the invariant chain (Ii) derived from the protein CD74. Ii stabilizes the class II complex by allowing correct protein folding and enables HLA class II heterodimers to be exported to the endosomal / lysosomal compartment. Within these HLA class II loading compartments, Ii is proteolytically cleaved by cathepsins into placeholder peptides known as CLIPs. CLIP is then exchanged for a higher-affinity peptide by the molecular chaperone HLA-DM (a non-classical HLA class II heterodimer) in a low-pH environment. The HLA class II complex loaded with the high-affinity peptide then travels to the trans-Golgi apparatus and ultimately to the cell surface for display by CD4+ T cells.

[0477] Each HLA heterodimer is estimated to bind thousands of peptides with allele-specific binding preferences. In fact, each HLA allele is estimated to bind and present approximately 1,000–10,000 unique peptides to T cells. Given this diversity in HLA binding, accurate prediction of whether a peptide is likely to bind to a specific HLA allele is very challenging. The allele-specific peptide binding properties of HLA class II molecules are poorly understood due to the heterogeneity of α and β chain pairing, data complexity that limits the ability to confidently assign core binding epitopes, and the lack of allele-specific antibodies of the immunoprecipitation grade required for high-resolution biochemical analysis. Furthermore, analysis of peptide epitopes derived from a given HLA allele introduces uncertainty when multiple HLA alleles are presented on the cell surface.

[0478] Prediction of candidate neoantigens has been primarily performed for HLA class I epitopes (given the availability of experimental data for class I prediction algorithms compared to class II), but CD4+ T cell responses have been frequently observed in preclinical and clinical personalized neoantigen vaccination studies. These observations suggest that HLA class II epitope processing and presentation may also play a key role in cancer therapy. Although HLA class II prediction algorithms exist, they are inaccurate because the open peptide-binding groove on the HLA class II heterodimer allows for the binding of longer peptides (typically 15–40 amino acids), which increases the heterogeneity and complexity of epitope presentation. Therefore, further work is needed to better understand the properties of the HLA class II peptide-binding core and the cellular processes involved in class II epitope processing and presentation. The field of proteomics is currently limited by the complexity of HLA class II heterodimer formation and the availability of immunoprecipitation-grade antibodies for isolation of HLA class II–peptide complexes. To overcome these challenges, a monoallelic HLA profiling workflow was developed that relies on LC-MS / MS to characterize allele-specific HLA class II ligand panels and class II epitope prediction methods. The following definitions supplement those in the art and are specific to this application and should not be attributed to any related or unrelated context, such as any commonly owned patents or applications. Exemplary materials and methods are described herein, but any methods and materials similar or equivalent to those described herein can be used in the practice of testing the present disclosure. Therefore, the terminology used herein is for the purpose of describing specific embodiments only and is not intended to be limiting.

[0479] Disclosed herein are methods for preparing personalized cancer vaccines. The methods for preparing personalized cancer vaccines may include identifying peptide sequences with mutations that are expressed in cancer cells of a subject; inputting amino acid position information of the identified peptide sequences into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the identified peptide sequences, each presentation prediction representing a probability that one or more proteins encoded by a class II MHC allele of the subject's cancer cells will present a given sequence of the identified peptide sequence; and selecting a subset of the peptide sequences identified based on the set of presentation predictions for use in preparing the personalized cancer vaccine.

[0480] In some embodiments, one or more results obtained from the methods described herein can provide one or more quantitative values ​​indicating one or more of the following: the likelihood of diagnostic accuracy, the likelihood of the presence of a condition in a subject, the likelihood of a subject developing a condition, the likelihood of success of a particular treatment, or any combination thereof. In some embodiments, the methods described herein can predict the risk or likelihood of developing a condition. In some embodiments, the methods described herein can provide an early diagnostic indicator of the development of a condition. In some embodiments, the methods described herein can confirm the diagnosis or presence of a condition. In some embodiments, the methods described herein can monitor the progression of a condition. In some embodiments, the methods described herein can monitor the efficacy of a treatment for a condition in a subject.

[0481] Methods for identification of MHC-II peptides

[0482] In one aspect, the present invention provides a method for identifying one or more peptides presented by an MHC-II protein for immune activation. In some embodiments, the one or more peptides comprise an epitope. In some embodiments, the method relates to a computational prediction of the likelihood that a particular epitope will be presented by an MHC-II protein. In some embodiments, the method relates to a computational prediction of the specificity of an epitope for MHC-II presentation. In some embodiments, the computational prediction method relates to an assessment of peptide-MHC interactions. In some embodiments, the computational prediction method relates to predicting the allele specificity of a peptide for antigen presentation.

[0483] In some embodiments, the computational prediction method involves the integration of bioinformatics information, such as nucleotide sequence, structural motifs of biomolecules, protein-protein interaction characteristics and functional efficacy such as immunogenicity. In some embodiments, the computational prediction method involves machine learning. Based on machine learning methods, such as simple pattern motifs, support vector machines (SVM), hidden Markov models (HMM), neural network (NN) models, quantitative structure-activity relationship (QSAR) analysis, structure-based methods and biophysical methods, many immunoinformatics methods for predicting peptide-MHC interactions have been developed for MHC class I and class II. These methods can be divided into two categories, i.e., intra-allelic (allele-specific) and trans-allelic (pan-specific) methods. The intra-allelic method is trained on a limited set of experimental peptide binding data for a specific MHC molecule and is applied to predict peptides that bind to the molecule. Due to the extreme polymorphism of MHC molecules, the existence of thousands of allelic variants, and the lack of sufficient experimental binding data, it is impossible to establish a prediction model for each allele. Therefore, cross-allele and universal methods such as NetMHCIIpan (Karosiene E et al., NetMHCIIpan-3.0, a common pan-specific MHC class II prediction method including all three human MHC class II isotypes, HLA-DR, HLA-DP and HLADQ. Immunogenetics (2013) 65(10): 711–24) and TEPITOPEpan (Zhang L et al., TEPITOPEpan: extending TEPITOPE for peptide binding prediction covering over 700 HLA-DR molecules. PLoS One (2012) 7(2): e30483) have been developed using peptide binding data that extend across many alleles or across species. Similar methods are also available for MHC-I, such as NetMHCpan and KISS.

[0484] In some embodiments, the peptide sequence may not be expressed in normal cells of the subject. In some embodiments, each and every cell of the subject may not be a cancer cell. The cancer cell may be produced by different cancers, including but not limited to thyroid cancer, adrenocortical cancer, anal cancer, aplastic anemia, bile duct cancer, bladder cancer, bone cancer, bone metastasis, central nervous system (CNS) cancer, peripheral nervous system (PNS) cancer, breast cancer, Castleman disease, cervical cancer, childhood non-Hodgkin lymphoma, lymphoma, colorectal cancer, endometrial cancer, esophageal cancer, Ewing's family of tumors (e.g., Ewing's sarcoma), eye cancer, gallbladder cancer, gastrointestinal carcinoid tumors, gastrointestinal stromal tumors, gestational trophoblastic disease, hairy cell leukemia, Hodgkin's disease, Kaposi's sarcoma, kidney cancer, laryngeal and hypopharyngeal cancer, acute lymphocytic leukemia, acute myeloid leukemia, leukemia, childhood leukemia, chronic lymphocytic leukemia, chronic myeloid leukemia, liver cancer, lung cancer, lung carcinoid tumor, non-Hodgkin lymphoma, male breast cancer, malignant mesothelioma, multiple myeloma, myelodysplastic syndrome, myeloproliferative disorders, nasal and paranasal cancer, nasopharyngeal cancer, neuroblastoma, oral cavity and oropharyngeal cancer, osteosarcoma, ovarian cancer, pancreatic cancer, penile cancer, pituitary tumor, prostate cancer, retinoblastoma, rhabdomyosarcoma, salivary gland cancer, sarcoma (soft tissue cancer in adults), melanoma skin cancer, non-melanoma skin cancer, stomach cancer, testicular cancer, thymic cancer, uterine cancer (such as uterine sarcoma), vaginal cancer, vulvar cancer, or Waldenstrom's macroglobulinemia.

[0485] The identification can include comparing the DNA, RNA or protein sequence from the subject's cancer cell with the DNA, RNA or protein sequence from the subject's normal cell. The DNA, RNA or protein sequence from the subject's cancer cell can be different from the DNA, RNA or protein sequence from the subject's normal cell. The identification can identify nucleic acid variants with high sensitivity.

[0486] The machine learning HLA-peptide presentation prediction model may include a plurality of predictor variables identified based at least on training data. The training data may include sequence information of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information including amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables.

[0487] In some embodiments, the training data may further include structured data, time series data, unstructured data, and relational data. Unstructured data may include audio data, image data, video, mechanical data, electrical data, chemical data, and any combination thereof, for accurately simulating or training a robot or simulation. Time series data may include data from one or more of smart meters, smart appliances, smart devices, monitoring systems, telemetry devices, or sensors. Relational data includes data from client systems, enterprise systems, operating systems, websites, network-accessible application programming interfaces (APIs), or any combination thereof. This can be accomplished by the user through any method of inputting files or other data formats into the software or system.

[0488] In some embodiments, the training data can be stored in a database. The database can be stored in a computer-readable format. The computer processor can be configured to access data stored in a computer-readable memory. In some embodiments, a computer system can be used to analyze the data to obtain results. The results can be stored remotely or internally on a storage medium and transmitted to personnel such as a pharmaceutical expert. In some embodiments, the computer system can be operably coupled to a component for transmitting the results. The components for transmission can include wired and wireless components. Examples of wired communication components can include universal serial bus (USB) connections, coaxial cable connections, Ethernet cables such as Cat5 or Cat6 cables, fiber optic cables, or telephone lines. Examples of wireless communication components can include Wi-Fi receivers, components for accessing mobile data standards such as 3G or 4G LTE data signals, or Bluetooth receivers. In some embodiments, all of this data in the storage medium is collected and archived to build a data warehouse.

[0489] In some embodiments, the database includes an external database. The external database can be a medical database, such as, but not limited to, an adverse drug reaction database, an AHFS supplement file, an allergen selection list file, an average WAC pricing file, a brand probability file, a Canadian drug file v2, a comprehensive price history, a controlled substance file, a drug allergy cross reference file, a drug application file, a drug administration and administration database, a drug image database v2.0 / drug imprint database v2.0, a drug expiration date file, a drug indication database, a drug laboratory conflict database, a drug therapy monitoring system (DTMS) v2.2 / DTMS consumer monograph, a duplicate treatment database, a federal government pricing file, a healthcare common procedure coding system code (HCPCS) database, an ICD-10 mapping file, an immunization cross reference file, a comprehensive A to Z drug facts module, a comprehensive patient education, a master parameter database, a half span electronic drug file (MED-File) v2, a Medicaid Rebate file, care plan file, condition selection list file, condition master database, medication order management database (MOMD), monitoring parameter database, patient safety program file, payment limits-part B (PAF-B) v2.0, precautions database, RxNorm cross-reference file, standard drug identifier database, substitution group file, supplemental name file, harmonized classification cross-reference file, or warning label database.

[0490] In some embodiments, training data may also be obtained through other data sources. Data sources may include sensors or smart devices, such as appliances, smart meters, wearable devices, monitoring systems, data stores, customer systems, billing systems, financial systems, crowd-sourced data, weather data, social networks, or any other sensor, enterprise system, or data store. Examples of smart meters or sensors may include meters or sensors located at customer locations, or meters or sensors located between customers and generation or source locations. By integrating data from a wide range of sources, the system may be able to perform complex and detailed analysis. In some embodiments, data sources may include, but are not limited to, sensors or databases used in other healthcare platforms.

[0491] HLA typing is conventionally performed by serological methods using antibodies or by PCR-based methods such as sequence-specific oligonucleotide probe hybridization (SSOP) or sequence-based typing (SBT). The first is hampered by a potentially high degree of cross-reactivity and a limited resolving power, while the second has difficulties related to PCR efficiency, since the possibilities for positioning primers are very limited due to the polymorphic positions.

[0492] In some embodiments, sequence information is identified by sequencing methods or methods using mass spectrometry, such as liquid chromatography-mass spectrometry (LC-MS or LC-MS / MS, or HPLC-MS or HPLC-MS / MS). These sequencing methods may be well known to those skilled in the art and are reviewed in Medzihradszky KF and Chalkley RJ. Mass Spectrom Rev. 2015 Jan-Feb; 34(1):43-63. In some embodiments, the mass spectrometry is monoallelic mass spectrometry. In some embodiments, the mass spectrometry may be MS analysis, MS / MS analysis, LC-MS / MS analysis, or a combination thereof. In some embodiments, MS analysis may be used to determine the mass of the intact peptide. For example, the determination may include determining the mass of the intact peptide (e.g., MS analysis). In some embodiments, MS / MS analysis may be used to determine the mass of a peptide fragment. For example, the determination may include determining the mass of a peptide fragment, which may be used to determine the amino acid sequence of the peptide or a portion thereof (e.g., MS / MS analysis). In some embodiments, the mass of a peptide fragment may be used to determine the amino acid sequence within the peptide. In some embodiments, LC-MS / MS analysis can be used to separate complex peptide mixtures. For example, the determination can include, for example, separating complex peptide mixtures by liquid chromatography, and determining the mass of the intact peptide, the mass of the peptide fragments, or a combination thereof (e.g., LC-MS / MS analysis). This data can be used, for example, for peptide sequencing.

[0493] In some embodiments, the training peptide sequence information includes amino acid position information of the training peptides. In some embodiments, the training peptide sequence information includes sequence information of up to about 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20%, 10% or less of the sequence of the peptide presented by the HLA protein expressed in the cell and identified by mass spectrometry. In some embodiments, the training peptide sequence information may include sequence information of at least about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% or more of the sequence of the peptide presented by the HLA protein expressed in the cell and identified by mass spectrometry.

[0494] Any information and data can be paired with the subject who originated it. A subject or healthcare professional can retrieve information and data from storage or a server using the subject's identity. The subject's identity can include a patient's photo, name, address, Social Security number, birthday, phone number, ZIP code, or any combination thereof. The subject's identity can be encrypted and encoded into a visual graphic code. The visual graphic code can be a one-time barcode that can be uniquely associated with the subject's identity. The barcode can be a UPC barcode, EAN barcode, Code 39 barcode, Code 128 barcode, ITF barcode, CodaBar barcode, GS1 DataBar barcode, MSI Plessey barcode, QR barcode, Datamatrix code, PDF417 code, or Aztec barcode. The visual graphic code can be configured to be displayed on a display screen. The barcode can include a QR code that can be optically captured and read by a machine. The barcode can define elements such as the barcode version, format, position, alignment, or timing to facilitate reading and decoding. Barcodes can encode various types of information in any suitable format, such as binary or alphanumeric information. QR codes can have various symbol sizes, as long as the QR code can be scanned from a reasonable distance by an imaging device. QR codes can be in any image file format (such as EPS or SVG vector graphics, PNG, TIF, GIF, or JPEG raster graphics formats).

[0495] In some embodiments, the function representing the relationship between the amino acid position information received as input and the presentation likelihood generated as output based on the amino acid position information and the predictor variable comprises a linear or nonlinear function. The function can be, for example, a rectified linear unit (ReLU) activation function, a leaky ReLu activation function, or other functions such as saturated hyperbolic tangent, identity, binary step function, logistic function, arcTan, softsign, parametric rectified linear unit, exponential linear unit, softPlus, bent identity, softExponential, sinusoid, Sinc, Gaussian, or sigmoid function, or any combination thereof.

[0496] In some embodiments, the linear function is obtained by linear regression. In some embodiments, linear regression is a method for predicting a target variable by fitting the best linear relationship between the dependent variable and the independent variable. The best fit may mean that the sum of all distances between the shape at each point and the actual observed value is minimized. Linear regression may include simple linear regression or multiple linear regression. Simple linear regression can use a single independent variable to predict the dependent variable. Multiple linear regression can use more than one independent variable to predict the dependent variable by fitting the best linear relationship. A nonlinear function can be obtained by nonlinear regression. Nonlinear regression can be a form of regression analysis in which the observed data is modeled by a function that is a nonlinear combination of model parameters and depends on one or more independent variables. Nonlinear regression can include step functions, piecewise functions, splines, and generalized additive models.

[0497] In some embodiments, the presentation probability is presented by a one-dimensional value (e.g., probability). In some embodiments, probability is configured to measure the likelihood that an event may occur. In some embodiments, the probability range is about 0 to 1, 0.1 to 0.9, 0.2 to 0.8, 0.3 to 0.7, or 0.4 to 0.6. The higher the probability of an event occurring, the more likely the event is to occur. In some embodiments, an event includes any type of situation, including, as a non-limiting example, whether an HLA-peptide will present certain peptides with specific amino acid position information, and whether a person will become ill based on the amino acid position information. In some embodiments, the likelihood can be presented by a multidimensional value. The multidimensional value can be presented by a multidimensional space, a heat map, or a spreadsheet.

[0498] In one embodiment, a subset of peptide sequences identified based on a set of presentation predictions is selected to prepare a personalized cancer vaccine. In some embodiments, the subset comprises up to about 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20%, 10% or less of peptide sequences identified based on the set of presentation predictions. In other cases, the subset may comprise at least about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% or more of peptide sequences identified based on the set of presentation predictions. A cancer vaccine may be a vaccine for treating an existing cancer or preventing the development of a cancer. The vaccine may be prepared from a sample taken from a patient and may be specific to the patient.

[0499] In some embodiments, poxviruses are used in disease vaccines or immunogenic compositions. These include orthopoxviruses, avipox, cowpox, MVA, NYVAC, canarypox, ALVAC, fowlpox, TROVAC, and the like. Advantages of such vectors may include simple construction, the ability to accommodate large amounts of foreign DNA, and high expression levels. With respect to poxviruses that may be used to practice the present disclosure, poxviruses of the Chordopoxvirinae subfamily (vertebrate poxviruses), for example, orthopoxviruses and avipox viruses, such as cowpox viruses (e.g., Wyeth strain, WR strain (e.g., VR-1354), Copenhagen strain, NYVAC, NYVAC.1, NYVAC.2, MVA, MVA-BN), canarypox virus (e.g., Wheatley C93 strain, ALVAC), fowlpox virus (e.g., FP9 strain, Webster strain, TROVAC), dovepox, pigeonpox, quailpox, and raccoonpox, particularly information on their synthetic or non-naturally occurring recombinants, their uses, and methods for making and using such recombinants can be found in the scientific and patent literature.

[0500] In some embodiments, vaccinia viruses are used in disease vaccines or immunogenic compositions to express antigens. Recombinant vaccinia viruses can replicate in the cytoplasm of infected host cells, so that the polypeptide of interest can induce an immune response.

[0501] In some embodiments, ALVAC is used as a vector in disease vaccines or immunogenic compositions. ALVAC can be a canarypox virus that can be modified to express foreign transgenes and has been used as a vaccination method against prokaryotic and eukaryotic antigens.

[0502] In some embodiments, a modified Ankara vaccinia (MVA) virus is used as a viral vector for an antigen vaccine or immunogenic composition. MVA can be a member of the Orthopoxviridae family and has been produced by approximately 570 serial passages of the Ankara strain of vaccinia virus (CVA) in chicken embryo fibroblasts. Due to these passages, the resulting MVA virus contains 31 kilobases less genomic information than CVA and is highly host cell restricted. MVA can be characterized by its extreme attenuation, i.e., weakened virulence or infectivity, but still having excellent immunogenicity. When tested in a variety of animal models, MVA can be shown to be non-toxic, even in immunosuppressed individuals. In addition, -HER2 may be a candidate immunotherapy for the treatment of HER-2-positive breast cancer and is currently undergoing clinical trials.

[0503] In some embodiments, a positive predictive value (PPV) is used as part of a prediction model. PPV, also referred to as a precision measurement, is the probability that an individual diagnosed with a disease or condition by, for example, a test or model actually suffers from the disease or condition. It can be calculated by dividing the number of true positive results by the total number of positive results (including false positive results) returned. PPV=true positive / (true positive+false positive). For example, if in a group of 100 patients, the model determines a positive result in 50 patients, 25 of whom are true positives, then the PPV will be 25 / 50=0.5. The closer the PPV is to 1, the more accurate the diagnostic method, such as a test or model, is. PPV can be used to determine the accuracy of a prediction model. PPV can be used to adjust a prediction model to adapt to the false positive results that the model may generate.

[0504] The decision rate can be used as part of a predictive model. The decision rate can be considered as the percentage of true positive results to the total number of positives in the sample set. Decision rate = true positive / (true positive + false negative). For example, if in a group of 100 patients, the model determined a positive result in 50 patients, 25 of which were true positives, and there were 75 positives in the group of patients, then the decision rate would be {25 / (25+25)}x100=50%. The decision rate can be used to determine the accuracy of the predictive model. The decision rate can be used to adjust the predictive model to adapt to the false positive results or false negative results that the model may generate.

[0505] In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 0.1%-10%. In some embodiments, the prediction model can have a positive predictive value of at most 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1 or less at a decision rate of 0.1%-10%. The prediction model can have a positive predictive value of at least 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate below 0.1%. In some embodiments, the prediction model can have a positive predictive value of at most 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1 or less at a decision rate below 0.1%. The prediction model can have a positive predictive value of at least 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate above 10%. In some embodiments, the prediction model can have a positive predictive value of at most 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1 or less at a decision rate above 10%.

[0506] In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 0.1% to 10%. In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 0.1% to 10%. In some embodiments, the prediction model has a positive predictive value of at least 0.1% to 0.5%, 0.1% to 1%, 0.1% to 2%, 0.1% to 3%, 0.1% to 4%, 0.1% to 5%, 0.1% to 6%, 0.1% to 7%, 0.1% to 8%, 0.1% to 9%, 0.1% to 10%, 0.5% to 1%, 0.5% to 2%, 0.5% to 3%, 0. 0.5% to 4%, 0.5% to 5%, 0.5% to 6%, 0.5% to 7%, 0.5% to 8%, 0.5% to 9%, 0.5% to 10%, 1% to 2%, 1% to 3%, 1% to 4%, 1% to 5%, 1% to 6%, 1% to 7%, 1% to 8%, 1% to 9%, 1% to 10%, 2% to 3%, 2% to 4%, 2% to 5%, 2% to 6 %, 2% to 7%, 2% to 8%, 2% to 9%, 2% to 10%, 3% to 4%, 3% to 5%, 3% to 6%, 3% to 7%, 3% to 8%, 3% to 9%, 3% to 10%, 4% to 5%, 4% to 6%, 4% to 7%, 4% to 8%, 4% to 9%, 4% to 10%, 5% to 6%, 5% to 7%, 5% to 8%, 5% to 9%, 5% to 10%, In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9% or 10%. In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of at least 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, or 9%. In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of at most 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9% or 10%.

[0507] In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 10% to 20%. In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 10% to 20%. In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 10% to 20%. 1% to 16%, 11% to 17%, 11% to 18%, 11% to 19%, 11% to 20%, 12% to 13%, 12% to 14%, 12% to 15%, 12% to 16%, 12% to 17%, 12% to 18%, 12% to 19%, 12% to 20%, 13% to 14%, 13% to 15%, 13% to 16%, 13% to 17%, 13% to 18%, 13% to 19%, 13% to 20%, 14% to 15%, 14% to 16%, 14% to 17%, 14% to 18%, 14% to 19%, 14% to 20%, 15% to 16%, 15% to 17%, 15% to 18%, 15% to 19%, 15% to 20%, 16% to 17%, 16% to 1 In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19% or 20%. In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of at least 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, or 19%. In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of at most 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19% or 20%.

[0508] In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of at least 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19% or 20%. For example, the prediction model can have a positive predictive value of at least 0.1 at a decision rate of at least 10%. For example, the prediction model can have a positive predictive value of at least 0.2 at a decision rate of at least 10%. For example, the prediction model can have a positive predictive value of at least 0.3 at a decision rate of at least 10%. For example, the prediction model can have a positive predictive value of at least 0.4 at a decision rate of at least 10%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of at least 10%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of at least 10%. For example, the prediction model may have a positive predictive value of at least 0.7 at a decision rate of at least 10%. For example, the prediction model may have a positive predictive value of at least 0.8 at a decision rate of at least 10%. For example, the prediction model may have a positive predictive value of at least 0.9 at a decision rate of at least 10%. For example, the prediction model may have a positive predictive value of at least 0.1 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.2 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.3 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.4 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.7 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.8 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.9 at a decision rate of at least 5%. For example, the prediction model may have a positive predictive value of at least 0.1 at a decision rate of at least 20%. For example, the prediction model may have a positive predictive value of at least 0.2 at a decision rate of at least 20%. For example, the prediction model may have a positive predictive value of at least 0.3 at a decision rate of at least 20%. For example, the prediction model may have a positive predictive value of at least 0.4 at a decision rate of at least 20%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of at least 20%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of at least 20%.For example, the prediction model can have a positive predictive value of at least 0.7 at a decision rate of at least 20%. For example, the prediction model can have a positive predictive value of at least 0.8 at a decision rate of at least 20%. For example, the prediction model can have a positive predictive value of at least 0.9 at a decision rate of at least 20%.

[0509] In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of about 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19% or 20%. For example, the prediction model can have a positive predictive value of at least 0.1 at a decision rate of about 10%. For example, the prediction model can have a positive predictive value of at least 0.2 at a decision rate of about 10%. For example, the prediction model can have a positive predictive value of at least 0.3 at a decision rate of about 10%. For example, the prediction model can have a positive predictive value of at least 0.4 at a decision rate of about 10%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of approximately 10%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of approximately 10%. For example, the prediction model may have a positive predictive value of at least 0.7 at a decision rate of approximately 10%. For example, the prediction model may have a positive predictive value of at least 0.8 at a decision rate of approximately 10%. For example, the prediction model may have a positive predictive value of at least 0.9 at a decision rate of approximately 10%. For example, the prediction model may have a positive predictive value of at least 0.1 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.2 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.3 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.4 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.7 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.8 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.9 at a decision rate of approximately 5%. For example, the prediction model may have a positive predictive value of at least 0.1 at a decision rate of approximately 20%. For example, the prediction model may have a positive predictive value of at least 0.2 at a decision rate of approximately 20%. For example, the prediction model may have a positive predictive value of at least 0.3 at a decision rate of approximately 20%. For example, the prediction model may have a positive predictive value of at least 0.4 at a decision rate of approximately 20%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of approximately 20%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of approximately 20%. For example, the predictive model can have a positive predictive value of at least 0.7 at a decision rate of about 20%.For example, the prediction model can have a positive predictive value of at least 0.8 at a decision rate of about 20%.For example, the prediction model can have a positive predictive value of at least 0.9 at a decision rate of about 20%.

[0510] In some embodiments, the prediction model has a positive predictive value of at least 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 or more at a decision rate of less than 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19% or 20%. For example, the prediction model can have a positive predictive value of at least 0.1 at a decision rate of at most 10%. For example, the prediction model can have a positive predictive value of at least 0.2 at a decision rate of at most 10%. For example, the prediction model can have a positive predictive value of at least 0.3 at a decision rate of at most 10%. For example, the prediction model can have a positive predictive value of at least 0.4 at a decision rate of at most 10%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of at most 10%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of at most 10%. For example, the prediction model may have a positive predictive value of at least 0.7 at a decision rate of at most 10%. For example, the prediction model may have a positive predictive value of at least 0.8 at a decision rate of at most 10%. For example, the prediction model may have a positive predictive value of at least 0.9 at a decision rate of at most 10%. For example, the prediction model may have a positive predictive value of at least 0.1 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.2 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.3 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.4 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.6 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.7 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.8 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.9 at a decision rate of at most 5%. For example, the prediction model may have a positive predictive value of at least 0.1 at a decision rate of at most 20%. For example, the prediction model may have a positive predictive value of at least 0.2 at a decision rate of at most 20%. For example, the prediction model may have a positive predictive value of at least 0.3 at a decision rate of at most 20%. For example, the prediction model may have a positive predictive value of at least 0.4 at a decision rate of at most 20%. For example, the prediction model may have a positive predictive value of at least 0.5 at a decision rate of at most 20%. For example, a ...

Claims

1. A method comprising: (a) processing amino acid information of a plurality of candidate peptide sequences using a machine learning HLA peptide presentation prediction model to generate a plurality of presentation predictions, wherein each candidate peptide sequence in the plurality of candidate peptide sequences is encoded by a genome or exome of a subject or a pathogen or virus in the subject, wherein the plurality of presentation predictions comprises an HLA presentation prediction for each of the plurality of candidate peptide sequences, wherein each HLA presentation prediction indicates a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject is capable of presenting a given candidate peptide sequence in the plurality of candidate peptide sequences, wherein the machine learning HLA peptide presentation prediction model is trained using training data comprising sequence information of training peptide sequences, the training peptides being identified by mass spectrometry as to be presented by HLA proteins expressed in training cells; as well as (b) identifying, based at least on the plurality of presentation predictions, a peptide sequence from the plurality of candidate peptide sequences as being presented by at least one of the one or more proteins encoded by a class II HLA allele of a cell of the subject; According to the method for determining the positive predictive value of presentation, the machine learning HLA peptide presentation prediction model has a positive predictive value of presentation of at least 0.

07.

2. The method of claim 1 , wherein the machine learning HLA peptide presentation prediction model has a positive predictive value of at least 0.07 when processing amino acid information of a plurality of test peptide sequences to generate a plurality of test presentation predictions, each test presentation prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject is capable of presenting a given test peptide sequence in the plurality of test peptide sequences, wherein the plurality of test peptide sequences comprises at least 500 test peptide sequences, the sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as to be presented by an HLA protein expressed in a cell, and (ii) at least 499 decoy peptide sequences contained within a protein encoded by a genome of an organism, wherein the organism and the subject are of the same species, wherein the ratio of the at least one hit peptide sequence in the plurality of test peptide sequences to the at least 499 decoy peptide sequences is 1:499, and according to the machine learning HLA peptide presentation prediction model, a top percentage of the plurality of test peptide sequences are predicted to be presented by an HLA protein expressed in a cell.

3. The method of claim 1, wherein the at least one hit peptide sequence comprises at least 5, 10, 20, 50 or 100 hit peptide sequences.

4. The method of claim 1, wherein the at least 499 bait peptide sequences comprise at least 2500, 5000, 10000, 25000, 50000 or 100000 bait peptide sequences.

5. The method of claim 1, wherein the at least 500 test peptide sequences comprise at least 2500, 5000, 10000, 25000, 50000 or 100000 test peptide sequences.

6. The method according to claim 1 or 2, wherein the peptide presented by the HLA protein has a length of 15-40 amino acids.

7. A method comprising: (a) processing amino acid information of a plurality of candidate peptide sequences encoded by a genome or exome of a subject using a machine learning HLA peptide presentation prediction model to generate a plurality of presentation predictions, wherein the plurality of presentation predictions comprises an HLA presentation prediction for each of the plurality of candidate peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject will bind to a given candidate peptide sequence in the plurality of candidate peptide sequences, wherein the machine learning HLA peptide presentation prediction model is trained using training data, wherein the training data comprises sequence information of peptide sequences identified to bind to HLA class II proteins or HLA class II protein analogs; as well as (b) identifying, based at least on the plurality of presentation predictions, a peptide sequence from the plurality of candidate peptide sequences having a probability of binding to at least one of the one or more proteins encoded by the class II HLA alleles of a cell of the subject that is greater than a threshold presentation prediction probability value; According to the method for determining the positive predictive value of presentation, the machine learning HLA peptide presentation prediction model has a positive predictive value of presentation of at least 0.

1.

8. The method of claim 7, wherein the machine learning HLA peptide presentation prediction model is trained using training data comprising sequence information of training peptide sequences, wherein the training peptides are identified by mass spectrometry as being presented by HLA proteins expressed in training cells.

9. The method of claim 1 or 2, wherein the method comprises ranking at least two peptides presented by at least one of the one or more proteins identified as encoded by a class II HLA allele of the subject's cells based on the multiple presentation predictions.

10. The method of claim 9, wherein the method comprises selecting one or more peptides from the at least two ranked peptides.

11. The method of claim 1 or 2, wherein the method comprises selecting one or more peptides from the plurality of candidate peptide sequences, the peptides being identified as presented by at least one of the one or more proteins encoded by a class II HLA allele of the subject's cells.

12. The method of claim 1 or 2, wherein the method comprises selecting one or more peptides from two or more peptides ranked based on the plurality of presentation predictions.

13. The method of claim 2, wherein the machine learning HLA peptide presentation prediction model has a positive predictive value of at least 0.1 when processing amino acid information of a plurality of test peptide sequences to generate a plurality of test presentation predictions, each test presentation prediction indicating a likelihood that one or more proteins encoded by a class II HLA allele of a cell of the subject will bind to a given test peptide sequence in the plurality of test peptide sequences, wherein the plurality of test peptide sequences comprises at least 20 test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as to be presented by an HLA protein expressed in a cell, and (ii) at least 19 decoy peptide sequences contained within a protein comprising at least one peptide sequence identified by mass spectrometry as to be presented by an HLA protein expressed in a cell, such as a single HLA protein expressed in a cell, wherein the ratio of the at least one hit peptide sequence to the at least 19 decoy peptide sequences in the plurality of test peptide sequences is 1:19, and according to the machine learning HLA peptide presentation prediction model, a top percentage of the plurality of test peptide sequences are predicted to bind to an HLA protein expressed in a cell.

14. The method of claim 13, wherein there is no amino acid sequence overlap between the at least one hit peptide sequence and the at least 19 decoy peptide sequences.

15. The method of claim 1 or 2, wherein the machine learning HLA peptide presentation prediction model has a positive predictive value of at least 0.08, 0.09, 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.2, 0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29, 0.3, 0.31, 0.32, 0.33, 0.34, 0.35, 0.36, 0.37, 0.38, 0.39, 0.4, 0.41, 0.42, 0.43, 0.44, 0.45, 0.46, 0.47, 0.48, 0.49, 0.50 49, 0.5, 0.51, 0.52, 0.53, 0.54, 0.55, 0.56, 0.57, 0.58, 0.59, 0.6, 0.61, 0.62, 0.63, 0.64, 0.65, 0.66, 0.67, 0.68, 0.69, 0.7, 0.71, 0.72, 0.73, 0.74 , 0.75, 0.76, 0.77, 0.78, 0.79, 0.8, 0.81, 0.82, 0.83, 0.84, 0.85, 0.86, 0.87, 0.88, 0.89, 0.9, 0.91, 0.92, 0.93, 0.94, 0.95, 0.96, 0.97, 0.98 or 0.

99.

16. The method of claim 2 or 13, wherein the top percentage is the top 0.2%, 0.3%, 0.4%, 0.5%, 0.6%, 0.7%, 0.8%, 0.9%, 1.00% or 2.00%.

17. The method of claim 13, wherein the at least one hit peptide sequence comprises at least 5, 10, 20, 50 or 100 hit peptide sequences.

18. The method of claim 9, wherein the at least 19 decoy peptide sequences comprise at least 500, 1000, 2000, 5000 or 10000 decoy peptide sequences.

19. The method of claim 9, wherein the at least 20 test peptide sequences comprise at least 500, 1000, 2000, 5000 or 10000 test peptide sequences.

20. The method of claim 2 or 13, wherein the top percentage is the top 5%, 6%, 7%, 8%, 9%, 10%, 15% or 20%.

21. The method of claim 1 or 2, wherein for a protein encoded by a corresponding HLA allele of Table 11, the positive predictive value is greater than the positive predictive value of column 2 of Table 11.

22. The method of claim 1 or 7, wherein for a protein encoded by an HLA class II allele, the positive predictive value is greater than the positive predictive value of column 2 of Table 16.

23. The method of claim 1 or 2, wherein the subject is a single subject.

24. The method of claim 1 or 2, wherein the cell is a cell that expresses a single protein encoded by a class II HLA allele of the subject's cell.

25. The method of claim 1 or 2, wherein the cell is a monoallelic HLA cell, or a cell expressing an HLA allele with an affinity tag.

26. The method of claim 1 or 2, wherein each peptide sequence in the plurality of candidate peptide sequences is associated with cancer.

27. The method of claim 1 or 2, wherein at least one peptide sequence in the plurality of candidate peptide sequences is overexpressed by a cancer cell in the subject.

28. The method of claim 1 or 2, wherein each peptide sequence in the plurality of candidate peptide sequences is overexpressed by a cancer cell of the subject.

29. The method of claim 1 or 2, wherein at least one peptide sequence among the plurality of candidate peptide sequences is a cancer cell specific peptide.

30. The method of claim 1 or 2, wherein each peptide sequence in the plurality of candidate peptide sequences is a cancer cell specific peptide.

31. The method of claim 1 or 2, wherein each peptide sequence in the plurality of candidate peptide sequences is expressed by a cancer cell in the subject.

32. The method of claim 1 or 2, wherein at least one peptide sequence in the plurality of candidate peptide sequences is not encoded by a non-cancer cell of the subject.

33. The method of claim 1 or 2, wherein each peptide sequence in the plurality of candidate peptide sequences is not encoded by a non-cancer cell of the subject.

34. The method of claim 1 or 2, wherein at least one peptide sequence in the plurality of candidate peptide sequences is not expressed by non-cancerous cells of the subject.

35. The method of claim 1 or 2, wherein each peptide sequence in the plurality of candidate peptide sequences is not expressed by non-cancerous cells of the subject.

36. The method of claim 1 or 2, wherein the method comprises obtaining the plurality of candidate peptide sequences of the subject.

37. The method of claim 1 or 2, wherein the method comprises obtaining a plurality of polynucleotide sequences of the subject.

38. The method of claim 1 or 2, wherein the method comprises obtaining a plurality of polynucleotide sequences of the subject, the polynucleotide sequences encoding a plurality of candidate peptide sequences encoded by the genome or exome of the subject.

39. The method of claim 1 or 2, wherein the method comprises obtaining, by a computer processor, a plurality of polynucleotide sequences of the subject, the polynucleotide sequences encoding a plurality of candidate peptide sequences encoded by the genome or exome of the subject.

40. The method of claim 1 or 2, wherein the method comprises obtaining a plurality of polynucleotide sequences of the subject by genome or exome sequencing.

41. The method of claim 1 or 2, wherein the method comprises obtaining multiple polynucleotide sequences of the subject by whole genome sequencing or whole exome sequencing.

42. The method of claim 1 or 2, wherein the method further formulates a composition comprising one or more peptide sequences from the selected subset of the plurality of candidate peptide sequences.

43. The method of claim 1 or 2, wherein the HLA class II protein comprises an HLA-DR, HLA-DQ or HLA-DP protein.

44. The method according to claim 1 or 2, wherein the HLA class II protein is selected from: HLA-DPB1*01:01 / HLA-DPA1*01:03, HLA-DPB1*02:01 / HLA-DPA1*01:03, HLA-DPB1*03:01 / HLA-DPA1*01:03, HLA-DPB1*04:01 / HLA-DPA1*01:03, HLA-DPB1*04:02 / HLA-DPA1*01:03, HLA-DPB1*06:01 / HLA-DPA1*01:03, HLA-DQB1*02:01 / HLA-DQA1*05:01, HLA-DQB1*02:02 / HLA-DQA1*02:01, HLA-DQB1*06:02 / HLA-DQA1*01:02, HLA-DQB1*06:04 / HLA-DQA1*01:02, HLA-DRB1*01:01, HLA-DRB1*01:02, HLA-DRB1*03:01, HLA-DRB1*03:02, HLA-DRB1*04:01, HLA-DRB1*04:02, HLA-DRB1*04:03, HLA-DRB1*04:04, HLA-DRB1*04:05, HLA-DRB1*04:07, HLA-DRB1*07:01, HLA-DRB1*08:01, HLA-DRB1*08:02, HLA-DRB1*08:03, HLA-DRB1*08:04, HLA-DRB1*09:01, HLA-DRB1*10:01, HLA-DRB1*11:01, HLA-DRB1*11:02, HLA-DRB1*11:04, HLA-DRB1*12:01, HLA-DRB1*12:02, HLA-DRB1*13:01, HLA-DRB1*13:02, HLA-DRB1*13:03, HLA-DRB1*14:01, HLA-DRB1*15:01, HLA-DRB1*15:02, HLA-DRB1*15:03, HLA-DRB1*16:01, HLA-DRB3*01:01, HLA-DRB3*02:02, HLA-DRB3*03:01, HLA-DRB4*01:01 and HLA-DRB5*01:01;HLA-DR proteins selected from the following group: DPA*01:03 / DPB*04:01, DRB1*01:01, DRB1*01:02, DRB1*03:01, DRB1*04:01, DRB1*04:02 , DRB1*04:04, DRB1*04:05, DRB1*07:01, DRB1*08:01, DRB1*08:02, DRB1*08:03, DRB1*09:01, DRB1*11:01, DRB1*11:02, DRB1*11:04, DRB1*12:01, DRB1*13:01, DRB1*13:02, DRB1*13:03, DRB1*14:01, DRB1*15:01, DRB1*15:02, DRB1*15:03, DRB1*16:02, DRB3*01:01, DRB3*02:01, DRB3*02:02, DRB3*03:01, DRB4*01:01, D RB4*01:03 and DRB5*01:01; an HLA-DP protein selected from the group consisting of DPB1*01:01, DPB1*02:01, DPB1*02:02, DPB1*03:01, DPB1*04:01, DPB1*04:02, DPB1*05:01, DPB1*06:01, DPB1*11:01, DPB1*13:01, DPB1*17:01; or an HLA-DQ protein complex selected from the group consisting of A1* 01:01+B1*05:01, A1*01:02+B1*06:02, A1*01:02+B1*06:04, A1*01:03+B1*06:03, A1*02:01+B1*02:02, A 1*02:01+B1*03:03, A1*03:01+B1*03:02, A1*03:03+B1*03:01, A1*05:01+B1*02:01 and A1*05:05+B1*03:

01. ; 45. The method of claim 42, wherein the composition further comprises an adjuvant.

46. ​​The method of claim 42, wherein the composition further comprises an immune checkpoint inhibitor.

47. A method for identifying an HLA class II specific peptide for immunotherapy of a subject, comprising: (a) obtaining, by a computer processor, a candidate peptide comprising an epitope and a plurality of peptide sequences, each of which comprises the epitope; (b) using a machine learning HLA-peptide presentation prediction model, processing the amino acid information of the plurality of peptide sequences by the computer processor to generate a presentation prediction for each of the plurality of peptide sequences to an immune cell, each presentation prediction indicating a likelihood that one or more proteins encoded by an HLA class II allele is capable of presenting a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data, the training data comprising sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry, and wherein according to a presentation positive predictive value determination method, the machine learning HLA peptide presentation prediction model has a positive predictive value of at least 0.07; (c) selecting a protein from one or more proteins encoded by HLA class II alleles of cells of the subject, the protein being predicted to bind to the candidate peptide by a machine learning HLA-peptide presentation prediction model, wherein the probability of the protein presenting the candidate peptide to immune cells is greater than a threshold presentation prediction probability value; (d) contacting the candidate peptide with a selected protein such that the candidate peptide competes for a placeholder peptide associated with the selected protein; as well as (e) identifying the candidate peptide as a peptide for immunotherapy that is specific to the selected protein based on whether the candidate peptide replaces the placeholder peptide.

48. A method for determining the immunogenicity of an MHC class II binding peptide, comprising: (a) selecting a protein encoded by an HLA class II allele that is predicted to bind to an MHC class II binding peptide by a machine learning HLA-peptide presentation prediction model, wherein the machine learning HLA-peptide presentation prediction model is configured to generate a presentation prediction for a given peptide sequence, the presentation prediction indicating a likelihood that one or more proteins encoded by the HLA class II allele are capable of presenting the given peptide sequence, and wherein the probability that the protein presents the MHC class II binding peptide is greater than a threshold presentation prediction probability value, and wherein the machine learning HLA peptide presentation prediction model has a positive predictive value of at least 0.07 according to a presentation positive predictive value determination method; (b) contacting the MHC class II binding peptide with a selected protein such that the MHC class II binding peptide competes with and displaces a placeholder peptide associated with the selected protein, thereby forming a complex comprising the HLA class II allele and the MHC class II binding peptide; (c) contacting the complex with CD4+ T cells, and (d) determining one or more activation parameters of the CD4+ T cells, wherein the parameters are selected from the group consisting of: induction of cytokines, induction of chemokines and expression of cell surface markers.

49. A method comprising: (a) identifying a peptide sequence associated with cancer and comprising a cancer mutation, wherein identifying the peptide sequence comprises comparing a DNA, RNA or protein sequence from a cancer cell of the subject with a DNA, RNA or protein sequence from a normal cell of the subject; (b) selecting a protein encoded by an HLA class II allele that is normally expressed by cells of the subject and that is predicted to bind to the peptide sequence by a machine learning HLA-peptide presentation prediction model; wherein the machine learning HLA-peptide presentation prediction model has a positive predictive value of at least 0.1 at a call rate of at least 0.1%, 0.1%-50%, or at most 50%, and wherein the probability that the protein presents the identified peptide sequence is greater than a threshold presentation prediction probability value; (c) contacting the identified peptide sequence with a selected protein encoded by an HLA class II allele to verify whether the identified peptide sequence competes with a placeholder peptide associated with the selected protein encoded by the HLA class II allele, thereby displacing the placeholder peptide with an IC50 value of less than 500 nM; and (d) Optionally, purifying the identified peptide.

50. A method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, comprising: (a) obtaining a plurality of peptide sequences of the polypeptide sequence by a computer processor; (b) using a machine learning HLA-peptide presentation prediction model, processing the amino acid information of the plurality of peptide sequences by a computer processor to generate a presentation prediction for each of the plurality of peptide sequences, each presentation prediction indicating a likelihood that one or more proteins encoded by the class II MHC alleles of a cell of the subject is capable of presenting an epitope sequence of a given peptide sequence in the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data, the training data comprising sequence information associated with HLA proteins expressed in cells; and (c) based on the presentation prediction, determining or predicting that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject, and wherein the machine learning HLA peptide presentation prediction model has a positive predictive value of at least 0.07 according to a presentation positive predictive value determination method.

51. A method for preparing an HLA class II tetramer by conjugating four individual HLA protein α1 and β1 heterodimers, the method comprising: (a) expressing in a eukaryotic cell a vector comprising a nucleic acid sequence encoding an α chain and a β chain of an HLA protein, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, such that each HLA protein α1 and β1 heterodimer is secreted in a dimerized state, wherein the heterodimer is associated with a placeholder peptide, (b) purifying the secreted heterodimer from the cell culture medium, (c) verifying peptide binding activity using a peptide exchange assay, (d) adding streptavidin to conjugate the heterodimer into a tetramer, (e) purifying the tetramer with a yield greater than 1 mg / L.

52. An HLA class II tetramer or multimer comprising HLA-DR or HLA-DP or HLA-DQ heterodimers, each heterodimer comprising an α chain and a β chain, wherein the heterodimers are purified and present at a concentration greater than 1 mg / L.

53. The HLA class II tetramer or multimer according to claim 52, wherein the HLA class II tetramer or multimer is an HLA class II protein selected from the group consisting of: HLA-DPB1*01:01 / HLA-DPA1*01:03, HLA-DPB1*02:01 / HLA-DPA1*01:03, HLA-DPB1*03:01 / HLA-DPA1*01:03, HLA-DPB1*04:01 / HLA-DPA1*01:03, HLA-DPB1*04:02 / HLA-DPA1*01:03, HLA-DPB1*06:01 / HLA-DPA1*01:03, HLA-DQB1*02:01 / HLA-DQA1*05:01, HLA-DQB1*02:02 / HLA-DQA1*02:01, HLA-DQB1*06:02 / HLA-DQA1*01:02, HLA-DQB1*06:04 / HLA-DQA1*01:02, HLA-DRB1*01:01, HLA-DRB1*01:02, HLA-DRB1*03:01, HLA-DRB1*03:02, HLA-DRB1*04:01, HLA-DRB1*04:02, HLA-DRB1*04:03, HLA-DRB1*04:04, HLA-DRB1*04:05, HLA-DRB1*04:07, HLA-DRB1*07:01, HLA-DRB1*08:01, HLA-DRB1*08:02, HLA-DRB1*08:03, HLA-DRB1*08:04, HLA-DRB1*09:01, HLA-DRB1*10:01, HLA-DRB1*11:01, HLA-DRB1*11:02, HLA-DRB1*11:04, HLA-DRB1*12:01, HLA-DRB1*12:02, HLA-DRB1*13:01, HLA-DRB1*13:02, HLA-DRB1*13:03, HLA-DRB1*14:01, HLA-DRB1*15:01, HLA-DRB1*15:02, HLA-DRB1*15:03, HLA-DRB1*16:01, HLA-DRB3*01:01, HLA-DRB3*02:02, HLA-DRB3*03:01, HLA-DRB4*01:01, and HLA-DRB5*01:

01.

54. A vector comprising a nucleic acid sequence encoding the α chain and β chain of the HLA protein of any one of claims 51-53, a secretion signal, a biotinylation motif and at least one tag for identification or purification, such that each HLA protein α1 and β1 heterodimer is secreted in a dimerized state, optionally wherein the secreted heterodimer is associated with a placeholder peptide.

55. A cell comprising the vector of claim 54.

56. A method for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: (a) using a computer processor to input amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that an epitope sequence of a given peptide sequence is presented by one or more proteins encoded by a class II MHC allele of a subject's cell; wherein the machine learning HLA-peptide presentation prediction model comprises: a plurality of predictor variables determined based at least on training data; wherein the training data comprises: sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing a relationship between amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables; (b) determining or predicting, based on the set of presentation predictions, that at least one of the peptide sequences of the polypeptide sequences is immunogenic to the subject, and wherein the machine learning HLA peptide presentation prediction model has a presentation positive predictive value of at least 0.07 according to a presentation positive predictive value determination method.

57. Use of a composition for the preparation of a medicament for treating an autoimmune disease or condition in a subject in need thereof, the composition comprising a regulatory T cell expressing a T cell receptor that binds to a complex comprising: (i) an epitope of an expressed protein identified or predicted by a machine learning HLA peptide presentation prediction model as presented by class II MHC of a cell of the subject, wherein the machine learning HLA peptide presentation prediction model has a presentation positive predictive value of at least 0.07 according to a presentation positive predictive value determination method, and (ii) class II MHC, wherein the complex is targeted by a CD8 or CD4 T cell of the subject.

Images