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

Abstract

The present disclosure provides methods of making personalized cancer vaccines. The invention further provides a method for training the machine learning HLA-peptide presentation prediction model.

Description

[0001] This application is a divisional application of Chinese patent application No. 201980092864.1, filed on December 20, 2019, entitled "Method and System for Predicting HLA Class II Specific Epitopes and Characterizing CD4+ T Cells" (the corresponding PCT application was filed on December 20, 2019, and has the application number PCT / US2019 / 068084).

[0002] Cross-references

[0003] This application claims the benefits 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 Technology

[0004] The major histocompatibility complex (MHC) is a gene complex encoding the human leukocyte antigen (HLA) gene. The HLA gene is expressed as a protein heterodimer displayed on the surface of human cells to circulating T cells. The HLA gene is highly polymorphic, allowing for fine-tuning of the adaptive immune system. The adaptive immune response depends in part on the ability of T cells to identify and eliminate cells displaying disease-associated peptide antigens that bind to the HLA heterodimer.

[0005] In humans, endogenous and exogenous proteins can be processed into peptides by the proteasome and 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 called human leukocyte antigens (HLA class I and II), and a group of peptides that bind to them and trigger an immune response are called HLA epitopes. HLA epitopes are key components enabling the immune system to detect danger signals such as pathogen infection and autotransformation. 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. The endogenous processing and presentation of HLA class II ligands is a complex process involving a subset of multiple molecular chaperones and enzymes, not all of which have been fully characterized. HLA class II peptide presentation activates helper T cells, subsequently promoting B cell differentiation and antibody production, as well as CTL responses. Activated helper T cells also secrete cytokines and chemokines that activate and induce differentiation of other T cells.

[0006] Understanding the peptide binding preferences of each HLA class II heterodimer is crucial for successfully predicting which cancer or tumor-specific antigens may 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., neoantigen peptides). Such methods and isolated molecules could be used, for example, to develop therapeutics, including but not limited to immunotherapy-based agents. Summary of the Invention

[0007] The methods and compositions described herein can be used in a wide range of applications. For example, they can be used to identify immunogenic antigenic peptides and to develop pharmaceuticals, such as personalized medicines, as well as to isolate and characterize antigen-specific T cells.

[0008] CD4+ T cell responses can have antitumor activity. High CD4+ T cell response rates can be shown without the use of class II prediction (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 be unclear whether these epitopes are typically presented naturally (by tumors or phagocytic dendritic cells). It may be desirable to translate high CD4+ T cell response rates into therapeutic effects by improving the identification of truly presented HLA class II binding epitopes.

[0009] The roles of gene expression, enzyme cleavage, and pathway / localization biases may not yet be strongly 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 a more relevant pathway, although most existing MS data are likely presumed to originate from autophagy. NetMHCIIpan may be the current predictive standard, but it may be considered inaccurate. Of the three HLA class II loci (DR, DP, and DQ), data may only exist for certain common alleles of HLA-DR.

[0010] Different data generation methods may be used to learn the rules of HLA class II presentation, including field criteria and proposed approaches. Field criteria may include affinity measurements, which could form the basis of the NetMHCIIpan predictor, offering low throughput and requiring radioactive reagents, and it omits the role of treatment. Proposed approaches may include mass spectrometry, where data from cell lines / tissues / tumors can help determine the rules of autophagy treatment, while monoallelic MS could allow the determination of allele-specific binding rules (assuming that multiallelic MS data is too complex for effective learning (Bassani-Sternberg. MCP. 2018)).

[0011] There may be different methods to validate new HLA class II predictors: validation using preserved MS data, which is likely the default setting; reviews of vaccine studies (e.g., NT-001), where immune surveillance data can assess vaccine peptide loading on APCs rather than tumor presentation, and the data can be sparsely expanded across many different alleles; biochemical affinity measurements, which can be configured to obtain measurements of peptides with inconsistent predictions (for only 2-3 alleles); and T cell induction, which can be configured to test the rate of ex vivo T cell responses induced by Neon-preferred and NetMHCIIpan-preferred epitopes.

[0012] For validation via T cell induction, the default approach may include evaluating neoORFs with inconsistent predictions from TCGA, where the inducing materials may include healthy donor APCs and T cells, and induction and readout can be performed via SLP (approximately 15-meric peptides). Random peptides can produce high response rates, while SLP may not adequately address the processing problem. Possible solutions may include induction via mRNA.

[0013] The 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 using a set of quality metrics to improve the quality of LC-MS / MS data to rigorously remove false positives that could increase the performance of the prediction model; identifying allele-specific HLA class II binding cores from HLA-ligandome LC-MS / MS datasets; utilizing machine learning algorithms to improve HLA class II ligand and epitope prediction; and / or identifying biological variables, such as gene expression, cleavability, gene preference, cellular localization, and secondary structure, that influence HLA class II ligand presentation and improve HLA class II epitope prediction.

[0014] This paper provides 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 of the plurality of candidate peptide sequences is encoded by the genome or exome of a subject, wherein the plurality of presentation predictions includes an HLA presentation prediction for each of the plurality of candidate peptide sequences, wherein each HLA presentation prediction indicates the probability that one or more proteins encoded by a class II HLA allele of the subject's cells can present a given candidate peptide sequence of the plurality of candidate peptide sequences, wherein the machine learning HLA peptide presentation prediction model is trained using training data containing sequence information of training peptide sequences, the training peptides being identified by mass spectrometry as being to be presented by an HLA protein expressed in the training cells; and (b) identifying peptide sequences of the plurality of peptide sequences as being to be presented by at least one of one or more proteins encoded by a class II HLA allele of the subject's cells, based at least on the plurality of presentation predictions; wherein the machine learning HLA peptide presentation prediction model has a PPV of at least 0.07 according to a presentation positive predictive value (PPV) determination method.

[0015] This paper provides a method comprising: (a) processing amino acid information of a plurality of peptide sequences encoded by the 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 include an HLA binding prediction for each of the plurality of candidate peptide sequences, each binding prediction indicating the probability that one or more proteins encoded by a class II HLA allele of the subject's cells will bind to a given candidate peptide sequence among 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 HLA class II proteins or HLA class II protein analogs; and (b) identifying, at least based on the plurality of binding predictions, a peptide sequence among the plurality of peptide sequences that has a probability greater than a threshold binding prediction probability value of binding to at least one of one or more proteins encoded by a class II HLA allele of the subject's cells; wherein the machine learning HLA peptide binding prediction model has a PPV of at least 0.1 according to a binding positive predictive value (PPV) determination method.

[0016] In some implementations, the machine learning HLA peptide presentation prediction model is trained using training data containing sequence information of training peptide sequences, which are identified by mass spectrometry as HLA proteins to be presented by the training cells.

[0017] In some embodiments, the method includes ranking at least two peptides presented by at least one of one or more proteins encoded by class II HLA alleles identified as being from the subject's cells, based on the presentation prediction.

[0018] In some implementations, the method includes selecting one or more peptides from the two or more ranked peptides.

[0019] In some embodiments, the method includes selecting one or more peptides from the plurality of peptides, the peptides being identified as being presented by at least one of one or more proteins encoded by a class II HLA allele of the subject's cells.

[0020] In some implementations, the method includes selecting one or more peptides from two or more peptides based on the presentation prediction ranking.

[0021] In some embodiments, 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 the probability that one or more proteins encoded by a class II HLA allele of the subject's cells can present a given test peptide sequence from the plurality of test peptide sequences, the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.07, 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 presentable by an HLA protein expressed in the cell, and (ii) at least 499 decoy peptide sequences contained in proteins encoded by the 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 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, the top percentage of the plurality of test peptide sequences are predicted to be presented by an HLA protein expressed in the cell.

[0022] In some embodiments, when processing amino acid information of a plurality of test peptide sequences to generate a plurality of test binding predictions, each test binding prediction indicating the probability that one or more proteins encoded by a class II HLA allele of the subject's cells will bind to a given test peptide sequence among the plurality of test peptide sequences, the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.1, wherein the plurality of test peptide sequences comprises at least 20 test peptide sequences, the sequences comprising (i) at least one that has been identified by mass spectrometry as a candidate for presenting an HLA protein expressed in the cells. (ii) at least 19 decoy peptide sequences contained within a protein comprising at least one peptide sequence identified by mass spectrometry as being presented by a single 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 decoy peptide sequences in the plurality of test peptide sequences is 1:19, and the top percentage of the plurality of test peptide sequences are predicted to bind to the HLA protein expressed in the cell according to a machine learning HLA peptide presentation prediction model.

[0023] In some embodiments, there is no amino acid sequence overlap between the at least one hit peptide sequence and the decoy peptide sequence.

[0024] In some implementations, 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.

[0025] 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.

[0026] In some embodiments, the at least 499 bait peptide sequences include 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, and 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 50,000, 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, 800,000900,000 or 1,000,000 bait peptide sequences. Those skilled in the art will recognize that altering the hit:decoy ratio changes the PPV.

[0027] In some embodiments, the at least 500 test peptide sequences include 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, 5 100, 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, 800000900,000 or 1,000,000 test peptide sequences.

[0028] In some implementations, the top-ranked percentages are 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%, and 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%.

[0029] 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.

[0030] In some embodiments, the at least 19 bait 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, and 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, 8250085,000, 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.

[0031] 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, 1 400, 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, 8000082,500, 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 test peptide sequences.

[0032] In some implementations, the top percentage 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%.

[0033] In some embodiments, for proteins encoded by the corresponding HLA alleles in Table 11, the PPV is greater than the corresponding PPV in column 2 of Table 11. In some embodiments, for proteins encoded by the corresponding HLA alleles in Table 11, the PPV is at least equal to the corresponding PPV in column 3 of Table 11.

[0034] In some implementations, for proteins encoded by HLA class II alleles, the PPV is equal to or greater than the corresponding PPV in column 2 of Table 12.

[0035] In some implementations, for proteins encoded by HLA class II alleles, the PPV is greater than the corresponding PPV in column 2 of Table 16.

[0036] In some implementations, the subject is a single subject.

[0037] In some implementations, the subject is a mammal.

[0038] In some implementations, the subject is a human being.

[0039] In some implementations, the training cells are cells that express a single protein encoded by a class II HLA allele from the subject cells.

[0040] In some embodiments, the training cells are monoallelic HLA cells, or cells expressing HLA alleles with affinity tags.

[0041] In some implementations, the subject's cells include cancer cells.

[0042] In some implementations, the method is used to identify peptide sequences.

[0043] In some implementations, the method is used to select peptide sequences.

[0044] In some implementations, the method is used to prepare for cancer treatment.

[0045] In some implementations, the method is used to prepare a subject-specific cancer treatment.

[0046] In some implementations, the method is used to prepare cancer-specific treatments for cancer cells.

[0047] In some implementations, each of the plurality of peptide sequences is associated with cancer.

[0048] In some embodiments, at least one of the plurality of peptide sequences is overexpressed by the subject's cancer cells.

[0049] In some implementations, each of the plurality of peptide sequences is overexpressed by the subject's cancer cells.

[0050] In some embodiments, at least one of the plurality of peptide sequences is a cancer cell-specific peptide.

[0051] In some embodiments, each of the plurality of peptide sequences is a cancer cell-specific peptide.

[0052] In some implementations, each of the plurality of peptide sequences is expressed by the subject's cancer cells.

[0053] In some implementations, at least one of the plurality of peptide sequences is not encoded by a non-cancer cell in the subject.

[0054] In some implementations, each of the plurality of peptide sequences is not encoded by a non-cancer cell in the subject.

[0055] In some implementations, at least one of the plurality of peptide sequences is not expressed by the subject's non-cancer cells.

[0056] In some implementations, each of the plurality of peptide sequences is not expressed by the subject's non-cancer cells.

[0057] In some implementations, the method includes obtaining the plurality of peptide sequences of the subject.

[0058] In some implementations, the method includes obtaining multiple polynucleotide sequences of the subject.

[0059] In some embodiments, the method includes obtaining multiple polynucleotide sequences of a subject, the polynucleotide sequences encoding multiple peptide sequences encoded by the subject's genome or exome or by a pathogen or virus in the subject.

[0060] In some implementations, the method includes obtaining multiple polynucleotide sequences of a subject via a computer processor, the polynucleotide sequences encoding multiple peptide sequences encoded by the subject's genome or exome.

[0061] In some implementations, the method includes obtaining multiple polynucleotide sequences of the subject through genome or exome sequencing.

[0062] In some implementations, the method includes obtaining multiple polynucleotide sequences of the subject through whole-genome sequencing or whole-exome sequencing.

[0063] In some implementations, the processing includes processing with a computer processor.

[0064] In some implementations, the processing includes generating multiple predictive variables based at least on amino acid information from the multiple peptide sequences.

[0065] In some implementations, machine learning HLA-peptide presentation prediction models are used to process the multiple predictor variables.

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

[0067] In some implementations, one or more proteins encoded by the class II HLA alleles of the subject's cells are one or more proteins encoded by the class II HLA alleles expressed by the subject's cancer cells.

[0068] In some implementations, one or more proteins encoded by the class II HLA alleles of the subject's cells are single proteins encoded by the class II HLA alleles of the subject's cells.

[0069] In some implementations, one or more proteins encoded by class II HLA alleles of the subject's cells are two, three, four, five, six or more proteins encoded by class II HLA alleles of the subject's cells.

[0070] In some implementations, one or more proteins encoded by the class II HLA alleles of the subject's cells are each of the proteins encoded by the class II HLA alleles of the subject's cells.

[0071] In some embodiments, the method further includes administering to a subject a composition comprising one or more peptide sequences from a selected subset of peptide sequences.

[0072] In some embodiments, identifying the plurality of peptide sequences includes comparing DNA, RNA, or protein sequences from a subject's cancer cells with DNA, RNA, or protein sequences from a subject's normal cells, wherein each of the plurality of peptides contains at least one mutation present in the subject's cancer cells but not in the subject's normal cells.

[0073] In some implementations, the machine learning HLA-peptide presentation prediction model includes at least a plurality of predictive variables identified based on training data, wherein the training data includes training peptide sequence information containing amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing the relationship between the amino acid position information and presentation probabilities generated as output based on the amino acid position information and the plurality of predictive variables.

[0074] In some implementations, identification includes, at least based on the plurality of presentation predictions, identifying a peptide sequence among the plurality of peptide sequences whose probability of presentation by at least one of one or more proteins encoded by a class II HLA allele of the subject's cells is greater than a threshold presentation prediction probability value.

[0075] In some implementations, a machine learning HLA peptide presentation prediction model is used to predict that one or more of the plurality of test peptide sequences to be presented are at least one of one or more proteins encoded by class II HLA alleles in the subject's cells, with a probability greater than a threshold presentation prediction probability value.

[0076] In some implementations, a machine learning HLA peptide presentation prediction model is used to predict that each of the plurality of test peptide sequences to be presented has a greater than a threshold presentation prediction probability value than at least one of one or more proteins encoded by a class II HLA allele in the subject's cells.

[0077] In some implementations, the number of positives is limited to the number of hits.

[0078] In some implementations, the mass spectrometry method is monoallelic mass spectrometry.

[0079] In some embodiments, the peptide is presented via autophagy by HLA proteins expressed in the cell.

[0080] In some embodiments, the peptide is presented by HLA proteins expressed in the cell via phagocytosis.

[0081] In some implementations, the plurality of predictive variables includes predicted expression levels of source proteins containing peptides.

[0082] In some implementations, the plurality of predictor variables includes a stability prediction value for the source protein containing the peptide.

[0083] In some implementations, the plurality of predictor variables includes predicted values ​​for the degradation rate of the source protein containing the peptide.

[0084] In some implementations, the plurality of predictor variables includes a predicted value for the protein cleavability of the source protein containing the peptide.

[0085] In some implementations, the plurality of predictive variables includes predicted values ​​for the cellular or tissue localization of the source protein containing the peptide.

[0086] In some implementations, the plurality of predictive variables includes predicted values ​​for the intracellular processing patterns of the source protein containing the peptide, wherein the processing patterns of the source protein particularly include predicted values ​​for whether the source protein undergoes autophagy, phagocytosis, and intracellular transport.

[0087] In some implementations, the quality of training data is improved by using multiple quality metrics.

[0088] In some implementations, the multiple quality metrics include removal of common contaminant peptides, high score peak intensity, high score, and high quality accuracy.

[0089] In some implementations, the peak intensity of the score is at least 50%.

[0090] In some implementations, the peak intensity of the score is at least 60%.

[0091] In some implementations, the score is at least 7.

[0092] In some implementations, the quality accuracy is at most 5 ppm.

[0093] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single immunoprecipitated HLA protein expressed in the cell.

[0094] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single exogenous HLA protein expressed in the cell.

[0095] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single recombinant HLA protein expressed in the cell.

[0096] In some implementations, the plurality of predictor variables includes peptide-HLA affinity predictor variables.

[0097] In some implementations, the peptides presented by HLA proteins include peptides identified by searching a database of enzyme-specific, unmodified peptides.

[0098] In some implementations, peptides presented by HLA proteins include peptides identified by searching peptide databases using a reverse database search strategy.

[0099] In some embodiments, the HLA protein includes HLA-DR, HLA-DQ, or HLA-DP protein.

[0100] 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*07:0!, 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*0!02, HLA-DRB3*03:01, HLA-DRB4*01:01, HLA-DRB5*01:01.

[0101] In some embodiments, the HLA-DR pairs with DRA*01:01.

[0102] 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.

[0103] In some embodiments, the HLA-DR protein comprises a dimer form of DRA*01:01.

[0104] In some embodiments, the HLA protein includes HLA-DP proteins 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.

[0105] In some embodiments, the HLA-DP protein pairing comprises DPA1*01:03.

[0106] 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.

[0107] In some embodiments, the peptides presented by HLA proteins include peptides identified by comparing the MS / MS profiles of HLA-peptides with the MS / MS profiles of one or more peptides or proteins in a peptide or protein database.

[0108] In some implementations, the mutation is selected from point mutations, splice site mutations, frameshift mutations, readthrough mutations, and gene fusion mutations.

[0109] In some embodiments, the peptide presented by the HLA protein has a length of 15-40 amino acids.

[0110] In some embodiments, peptides presented by HLA proteins include peptides identified by comparing the MS / MS spectra of HLA-peptides with the MS / MS spectra of one or more peptides or proteins in a peptide or protein database.

[0111] In some implementations, the personalized cancer treatment further includes an adjuvant.

[0112] In some implementations, the personalized cancer treatment further includes immune checkpoint inhibitors.

[0113] In some implementations, the training data includes structured data, time-series data, unstructured data, relational data, or any combination thereof.

[0114] In some implementations, the unstructured data includes image data.

[0115] In some implementations, the relationship data includes data from client systems, enterprise systems, operating systems, websites, web-accessible application interfaces (APIs), or any combination thereof.

[0116] In some implementations, the training data is uploaded to a cloud-based database.

[0117] In some implementations, the training is performed using a convolutional neural network.

[0118] In some implementations, the convolutional neural network comprises at least two convolutional layers.

[0119] In some implementations, the convolutional neural network includes at least one batch normalization step.

[0120] In some implementations, the convolutional neural network includes at least one spatial dropout step.

[0121] In some implementations, the convolutional neural network includes at least one global max pooling step.

[0122] In some implementations, the convolutional neural network includes at least one dense layer.

[0123] In some implementations, identifying peptide sequences includes identifying mutated peptide sequences expressed in the subject's cancer cells.

[0124] In some implementations, identifying peptide sequences includes identifying peptide sequences that are not expressed in the subject's normal cells.

[0125] In some implementations, identifying peptide sequences includes identifying viral peptide sequences.

[0126] In some implementations, identifying peptide sequences includes identifying overexpressed peptide sequences.

[0127] This article provides a method for identifying HLA class II specific peptides for immunotherapy in subjects, comprising: obtaining candidate peptides containing epitopes and multiple peptide sequences via a computer processor, each peptide sequence containing the epitope; processing the amino acid information of the multiple peptide sequences via a computer processor using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the multiple peptide sequences to immune cells, each presentation prediction indicating the probability that one or more proteins encoded by HLA class II alleles can present a given peptide sequence among the multiple peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing sequence information of peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; and obtaining HLA from subject cells... A protein is selected from one or more proteins encoded by class II alleles, the protein being predicted by a machine learning HLA-peptide presentation prediction model to bind to the candidate peptide, wherein the probability of the protein presenting the candidate peptide to immune cells is greater than a threshold presentation prediction probability value; the candidate peptide is contacted with the selected protein such that the candidate peptide competes with a placeholder peptide associated with the selected protein; and the candidate peptide is identified as a peptide for immunotherapy that is specific to the selected protein, based on whether the candidate peptide replaces the placeholder peptide.

[0128] In some implementations, obtaining the candidate peptide includes identifying the candidate peptide, wherein identifying the candidate peptide includes comparing DNA, RNA, or protein sequences from the subject's cancer cells with DNA, RNA, or protein sequences from the subject's normal cells.

[0129] In some implementations, the processing includes determining multiple predictor variables based at least on amino acid information of the multiple peptide sequences, and processing the multiple predictor variables using a machine learning HLA-peptide presentation prediction model.

[0130] In some implementations, the machine learning HLA-peptide presentation prediction model includes at least a plurality of predictive variables identified based on training data, wherein the training data includes training peptide sequence information containing amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing the relationship between the amino acid position information and presentation probabilities generated as output based on the amino acid position information and the plurality of predictive variables.

[0131] In some implementations, the number of positives is limited to the number of hits.

[0132] In some implementations, the mass spectrometry method is monoallelic mass spectrometry.

[0133] In some implementations, the plurality of predictive variables includes any one or more of the following: predicted expression levels of source proteins containing peptides, predicted stability, predicted degradation rates, predicted cleavability, predicted cell or tissue localization, and predicted intracellular processing patterns including autophagy, phagocytosis, and intracellular transport.

[0134] In some implementations, the quality of training data is improved by using multiple quality metrics.

[0135] In some implementations, the multiple quality metrics include removal of common contaminant peptides, high score peak intensity, high score, and high quality accuracy.

[0136] In some implementations, the peak intensity of the score is at least 50%.

[0137] In some implementations, the peak intensity of the score is at least 60%.

[0138] In some embodiments, the placebo peptide is a CLIP peptide.

[0139] In some implementations, the placebo peptide is a CMV peptide.

[0140] In some embodiments, the method further includes measuring the IC50 of the occupant peptide being replaced by the target peptide.

[0141] In some embodiments, the IC50 of the placebo peptide being replaced by the target peptide is less than 500 nM.

[0142] In some implementations, at least one of the proteins encoded by the HLA class II allele from the subject's cells is an HLA class II tetramer or multimer.

[0143] In some implementations, the target peptide is further identified by mass spectrometry.

[0144] In some embodiments, at least one protein encoded by the HLA class II allele of the subject's cells is a recombinant protein.

[0145] In some embodiments, at least one protein encoded by the HLA class II allele of the subject's cells is expressed in eukaryotic cells.

[0146] In some embodiments, the peptide is presented via autophagy by HLA proteins expressed in the cell.

[0147] In some embodiments, the peptide is presented by HLA proteins expressed in the cell via phagocytosis.

[0148] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single immunoprecipitated HLA protein expressed in the cell.

[0149] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single exogenous HLA protein expressed in the cell.

[0150] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single recombinant HLA protein expressed in the cell.

[0151] In some implementations, the plurality of predictor variables includes peptide-HLA affinity predictor variables.

[0152] In some implementations, the peptides presented by HLA proteins include peptides identified by searching a database of enzyme-specific, unmodified peptides.

[0153] In some implementations, peptides presented by HLA proteins include peptides identified by searching peptide databases using a reverse database search strategy.

[0154] In some embodiments, the HLA protein includes HLA-DR, HLA-DQ, or HLA-DP protein.

[0155] In some implementations, the immunotherapy is cancer immunotherapy.

[0156] In some implementations, the epitope is a cancer-specific epitope.

[0157] In some embodiments, the at least one protein encoded by the HLA class II allele comprises at least the α1 and β1 subunits of the HLA protein in dimer form.

[0158] In some implementations, the identity of the peptide is known.

[0159] In some implementations, the identity of the peptide is unknown.

[0160] In some implementations, the identity of the peptide is determined by mass spectrometry.

[0161] In some implementations, peptide exchange assays include detecting peptide fluorescent probes or tags.

[0162] In some embodiments, the placebo peptide is a CLIP peptide. In some embodiments, the placebo peptide has the amino acid sequence PVSKMRMATPLLMQA.

[0163] In some embodiments, the polynucleotide construct comprises an expression vector, which further comprises one or more of the following: a promoter, a secretion signal, a dimerizing factor, a ribosome jumping sequence, and one or more tags for purification and / or detection.

[0164] In some implementations, the placeholder peptide sequence is encoded by a nucleic acid sequence within the vector.

[0165] In some implementations, the sequence encoding a cleavable domain is placed between the sequence encoding a placeholder peptide and the sequence encoding an HLAβ1 peptide.

[0166] This article provides a method for determining the immunogenicity of MHC class II binding peptides, comprising: selecting a protein encoded by an HLA class II allele that is predicted by a machine learning HLA-peptide presentation prediction model to bind to an MHC class II binding peptide, 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 the probability that one or more proteins encoded by the HLA class II allele can present the given peptide sequence, and wherein the probability of the protein presenting 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 replaces a vacancy 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 CD4+ T cells; and measuring one or more activation parameters of the CD4+ T cells, the parameters being selected from: cytokine induction, chemokine induction, and expression of cell surface markers.

[0167] In some implementations, the HLA class II allele is a tetramer or a multimer.

[0168] In some implementations, the cytokine is IL-2.

[0169] This article provides a method for inducing CD4+ T cell activation in a subject for cancer immunotherapy. The method includes: identifying a peptide sequence associated with cancer and containing a cancer mutation, wherein identifying the peptide sequence includes comparing a DNA, RNA, or protein sequence from the subject's cancer cells with a DNA, RNA, or protein sequence from the subject's normal cells; selecting a protein encoded by an HLA class II allele, which is normally expressed by the subject's cells 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 up to 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 HLA-mediated binding of the selected protein. The placeholder peptide associated with the selected protein encoded by the class II allele is replaced with an IC50 value of less than 500 nM; optionally, the identified peptide is purified; and an effective amount of a polypeptide containing the sequence of the identified peptide or a polynucleotide encoding the polypeptide is administered to the subject.

[0170] This article provides a method for screening drugs containing peptide sequences for immunogenicity in a subject, comprising: obtaining a plurality of peptide sequences of the peptide sequence by a computer processor; processing the amino acid information of the plurality of peptide sequences by a computer processor 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 the probability that one or more proteins encoded by a class I or class II MHC allele of the subject's cells can present an epitope sequence of a given peptide sequence of the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing sequence information associated with HLA proteins expressed in the cells; determining or predicting, based on the plurality of presentation predictions, that each of the plurality of peptide sequences of the peptide sequence is not immunogenic to the subject; and administering a composition containing the drug to the subject.

[0171] This article provides a method for preparing HLA class II tetramers or multimers by conjugating four separate HLA protein α1 and β1 heterodimers. The method includes: expressing a vector in eukaryotic cells containing nucleic acid sequences encoding the α and β chains of HLA proteins, a secretion signal, a biotinylated 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 placebo peptide; purifying the secreted heterodimers from cell culture medium; verifying peptide-binding activity using a peptide exchange assay; adding streptavidin to conjugate the heterodimers into tetramers; purifying the tetramers in a yield greater than 1 mg / L. Multimers, such as pentamers, hexamers, or octamers, can also be produced in the same manner, and these are also considered herein.

[0172] In some implementations, the vector contains a CMV promoter.

[0173] In some embodiments, the vector comprises a sequence encoding a placebo peptide linked to the β1 chain via a cleavable site.

[0174] In some implementations, peptide exchange assays involve pre-cleaving the occupier peptide from the β chain.

[0175] In some implementations, the cleavable site is a thrombin cleavage site.

[0176] In some implementations, the peptide exchange assay is a FRET assay.

[0177] 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.

[0178] This article provides an HLAII tetramer or polymer comprising HLA-DR, HLA-DP, or HLA-DQ heterodimers, each heterodimer containing α and β chains, wherein the heterodimer is purified and present at a concentration greater than 1 mg / L.

[0179] In some implementations, the HLA class II tetramer is selected from Tables 8A-8C.

[0180] In some embodiments, the HLA class II tetramer comprises heterodimer pairs selected from the group consisting of HLA-DR, HLA-DP, and HLA-DQ proteins.

[0181] 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, and 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).

[0182] In some implementations, the heterodimer pairs are expressed in eukaryotic cells.

[0183] In some implementations, the heterodimer pair is encoded by a carrier.

[0184] This article provides a vector comprising nucleic acid sequences encoding the α and β chains of the HLA protein described herein, a secretion signal, a biotinylated 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 placebo peptide.

[0185] This article provides a cell that includes the vector described herein.

[0186] In some embodiments, the HLAII heterodimer is secreted from eukaryotic cells into a cell culture medium and further purified by any of the following methods: column chromatography, ion exchange chromatography, size exclusion chromatography, affinity chromatography, or LC-MS.

[0187] This article provides a method for screening drugs containing peptide sequences for immunogenicity in a subject, comprising: obtaining a plurality of peptide sequences of the peptide sequence by a computer processor; processing the amino acid information of the plurality of peptide sequences by a computer processor 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 the probability that one or more proteins encoded by class I or class II MHC alleles of the subject's cells can present an epitope sequence of a given peptide sequence of the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing 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 peptide sequence is immunogenic to the subject.

[0188] This article provides a method for screening drugs containing peptide sequences for immunogenicity in subjects. The method includes: inputting amino acid information of the peptide sequence of the peptide 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 the probability that one or more proteins encoded by class I or class II MHC alleles of the subject's cells will present an epitope sequence of a given peptide sequence; wherein the machine learning HLA-peptide presentation prediction model includes: a plurality of predictor variables determined at least based on training data; wherein the training data includes: sequence information of peptide sequences 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 the relationship between the amino acid position information received as input and the 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 peptide sequence is not immunogenic to the subject; and administering a composition containing the drug to the subject.

[0189] This article provides a method for screening drugs containing peptide sequences for immunogenicity in subjects. The method includes: inputting amino acid information of the peptide 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 the probability that one or more proteins encoded by class I or class II MHC alleles of the subject's cells will present an epitope sequence of a given peptide sequence; wherein the machine learning HLA-peptide presentation prediction model includes: at least a plurality of predictor variables determined based on training data; wherein the training data includes: sequence information of peptide sequences 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 the relationship between the amino acid position information received as input and the presentation probability generated as output based on the amino acid position information and the predictor variables; and, based on the set of presentation predictions, determining or predicting that at least one of the peptide sequences of the peptide sequence is immunogenic to the subject.

[0190] This article provides a method for screening drugs containing peptide sequences for immunogenicity in a subject, comprising: obtaining a plurality of peptide sequences of the peptide sequence by a computer processor; processing the amino acid information of the plurality of peptide sequences by a computer processor 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 the probability that one or more proteins encoded by a class I or class II MHC allele of the subject's cells can present an epitope sequence of a given peptide sequence of the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing sequence information associated with HLA proteins expressed in the cells; determining or predicting, based on the plurality of presentation predictions, that each of the plurality of peptide sequences of the peptide sequence is not immunogenic to the subject; and administering a composition containing the drug to the subject.

[0191] In some implementations, the method further includes deciding not to administer the drug to the subject.

[0192] In some embodiments, the drug comprises an antibody or a binding fragment thereof.

[0193] 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 the class I or class II MHC allele of the subject cell is the protein encoded by the class I MHC allele of the subject cell.

[0194] 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 the class I or class II MHC allele of the subject cell is the protein encoded by the class II MHC allele of the subject cell.

[0195] This article provides a method for treating a subject with an autoimmune disease or condition, comprising: (a) identifying or predicting an epitope of an expressed protein presented by class I or class II MHCs from the subject's cells, wherein a complex of the identified or predicted epitope and class I or class II MHCs is targeted by the subject's CD8 or CD4 T cells; (b) identifying a T cell receptor (TCR) that binds to the complex; (c) expressing the TCR in regulatory T cells derived from or allogeneic from the subject; and (d) administering the regulatory T cells expressing the TCR to the subject.

[0196] In some implementations, the autoimmune disease or condition is diabetes.

[0197] In some implementations, the cells are pancreatic islet cells.

[0198] This article provides a method for treating a subject with 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 a class I or class II MHC of the subject's cells, and (ii) a class I or class II MHC, wherein the complex is targeted by the subject's CD8 or CD4 T cells.

[0199] This document provides 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 operatively 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 the probability that one or more proteins encoded by class II MHC alleles of the subject's cells can present a given peptide sequence of 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, at least based on the plurality of presentation predictions, select a subset of the plurality of peptide sequences for personalized cancer treatment of the subject.

[0200] This document provides a computer system for identifying HLA class II specific peptides for immunotherapy in a subject, comprising: a database configured to store candidate peptides containing epitopes and a plurality of peptide sequences, each peptide sequence containing the epitope; and one or more computer processors operatively coupled to the database, wherein the one or more computer processors are individually and jointly 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 the probability that one or more proteins encoded by an HLA class II allele can present a given peptide sequence of 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 peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; and HLA from subject cells. A protein is selected from one or more proteins encoded by class II alleles, the protein being predicted by a machine learning HLA-peptide presentation prediction model to bind to the candidate peptide, wherein the probability that the protein will present the candidate peptide to immune cells is greater than a threshold presentation prediction probability value; and when the candidate peptide comes into contact with the selected protein such that the candidate peptide competes with a placeholder peptide associated with the selected protein, the candidate peptide is identified as a peptide for immunotherapy that is specific to the selected protein, based on whether the candidate peptide replaces the placeholder peptide.

[0201] This document provides a computer system for screening drugs comprising peptide sequences for immunogenicity in a subject, comprising: a database configured to store a plurality of peptide sequences of the peptide sequences; and one or more computer processors operatively coupled to the database, wherein the one or more computer processors are individually and jointly 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 the probability that one or more proteins encoded by class I or class II MHC alleles of a subject's cells can present an epitope sequence of a given peptide sequence of the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing sequence information associated with HLA proteins expressed in the cells; and based on the plurality of presentation predictions, determine or predict that each of the plurality of peptide sequences of the peptide sequences is not immunogenic to the subject, wherein a composition comprising the drug is administered to the subject.

[0202] This document provides a computer system for screening drugs comprising peptide sequences for immunogenicity in a subject, comprising: a database configured to store a plurality of peptide sequences of the peptide sequences; and one or more computer processors operatively 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 the probability that one or more proteins encoded by class I or class II MHC alleles of a subject's cells can present an epitope sequence of a given peptide sequence of 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 based on the plurality of presentation predictions, to determine or predict that at least one of the plurality of peptide sequences of the peptide sequences is immunogenic to the subject.

[0203] This document provides a non-transitory computer-readable medium containing 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 of 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 the probability that one or more proteins encoded by class II MHC alleles of the subject's cells can present a given peptide sequence of 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 selecting a subset of the plurality of peptide sequences for personalized cancer treatment of the subject, at least based on the plurality of presentation predictions.

[0204] This document provides a non-transitory computer-readable medium containing machine-executable code that, when executed by one or more computer processors, implements a method for identifying HLA class II-specific peptides for immunotherapy in a subject. The method includes: obtaining candidate peptides containing epitopes and a plurality of peptide sequences, each peptide sequence containing the epitope; processing the 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 the probability that one or more proteins encoded by an HLA class II allele can present a given peptide sequence from the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing sequence information of peptide sequences presented by HLA proteins expressed in cells and identified by mass spectrometry; and obtaining HLA from a subject's cells. A protein is selected from one or more proteins encoded by class II alleles, the protein being predicted by a machine learning HLA-peptide presentation prediction model to bind to the candidate peptide, wherein the probability of the protein presenting the candidate peptide to immune cells is greater than a threshold presentation prediction probability value; and when the candidate peptide comes into contact with the selected protein such that the candidate peptide competes with the occupant peptide, the candidate peptide is identified as a peptide for immunotherapy that is specific to the selected protein, based on whether the candidate peptide replaces the occupant peptide.

[0205] This document provides a non-transitory computer-readable medium containing machine-executable code that, when executed by one or more computer processors, implements a method for screening for immunogenicity of a drug comprising a peptide sequence in a subject, the method comprising: obtaining a plurality of peptide sequences of the peptide 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 the probability that one or more proteins encoded by a class I or class II MHC allele of a subject's cells can present an epitope sequence of a given peptide sequence of the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing sequence information associated with HLA proteins expressed in the cells; and based on the plurality of presentation predictions, determining or predicting that each of the plurality of peptide sequences of the peptide sequence is not immunogenic to the subject, wherein a composition comprising the drug is administered to the subject.

[0206] This document provides a non-transitory computer-readable medium containing machine-executable code that, when executed by one or more computer processors, implements a method for screening for immunogenicity of a drug comprising a peptide sequence in a subject, the method comprising: obtaining a plurality of peptide sequences of the peptide sequence; processing the 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 the probability that one or more proteins encoded by a class I or class II MHC allele of a subject's cells can present an epitope sequence of a given peptide sequence of the plurality of peptide sequences, wherein the machine learning HLA-peptide presentation prediction model is trained using training data containing sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry; and, based on the plurality of presentation predictions, determining or predicting that at least one of the plurality of peptide sequences of the peptide sequence is immunogenic to the subject.

[0207] This paper provides 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 of the plurality of candidate peptide sequences is encoded by the genome or exome of a subject, wherein the plurality of presentation predictions includes an HLA presentation prediction for each of the plurality of candidate peptide sequences, wherein each presentation prediction indicates the probability that one or more proteins encoded by a class II HLA allele of the subject's cells can present a given candidate peptide sequence of the plurality of candidate peptide sequences, wherein the machine learning HLA peptide presentation prediction model is trained using training data containing sequence information of training peptide sequences, the training peptides being identified by mass spectrometry as being presentable by an HLA protein expressed in the training cells; and identifying, at least based on the plurality of presentation predictions, peptide sequences among the plurality of peptide sequences whose probability of presentation by at least one of one or more proteins encoded by a class II HLA allele of the subject's cells is greater than a threshold presentation prediction probability. 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 multiple test peptide sequences to generate multiple test presentation predictions, each test presentation prediction indicating the probability that one or more proteins encoded by class II HLA alleles of the subject's cells can present a given test peptide sequence among the multiple test peptide sequences. The multiple test peptide sequences comprise at least 500 test peptide sequences, the sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as presentable by an HLA protein expressed in the cell, and (ii) at least 499 decoy peptide sequences contained within proteins encoded by the 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 to the at least 499 decoy peptide sequences among the multiple test peptide sequences is 1:499, and according to the machine learning HLA peptide presentation prediction model, 0.2% of the multiple test peptide sequences are predicted to be presented by an HLA protein expressed in the cell.

[0208] This article provides a method comprising: processing amino acid information of a plurality of peptide sequences encoded by the 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 include an HLA binding prediction for each of the plurality of candidate peptide sequences, each binding prediction indicating the probability that one or more proteins encoded by class II HLA alleles of the subject's cells will bind to a given candidate peptide sequence among the plurality of candidate peptide sequences, wherein the machine learning HLA peptide binding prediction model is trained using training data, the training data comprising proteins identified as binding to HLA class II proteins or HLA The sequence information of peptide sequences that bind to class II protein analogs; and, based at least on the plurality of binding predictions, identifying a peptide sequence among the plurality of peptide sequences whose probability of binding to at least one of one or more proteins encoded by class II HLA alleles of the subject's cells is greater than a threshold binding prediction probability value; wherein, when processing the amino acid information of the plurality of test peptide sequences to generate a plurality of test binding predictions, each test binding prediction indicating the likelihood that one or more proteins encoded by class II HLA alleles of the subject's cells will bind to a given test peptide sequence among the plurality of test peptide sequences, the machine learning HLA peptide binding prediction model has a positive predictive value (PPV) of at least 0.1. 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 presenting an HLA protein expressed in a cell, and (ii) at least 19 decoy peptide sequences contained within a protein that comprises peptide sequences identified by mass spectrometry as presenting an HLA protein expressed in a cell, wherein the organism and the subject are of the same species, 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 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.

[0209] In some implementations, the machine learning HLA peptide presentation prediction model is trained using training data containing sequence information of training peptide sequences, which are identified by mass spectrometry as HLA proteins to be presented by the training cells.

[0210] In some implementations, a machine learning HLA peptide presentation prediction model is used to predict that one or more of the plurality of test peptide sequences to be presented are at least one of one or more proteins encoded by class II HLA alleles of the subject's cells with a probability greater than a threshold presentation prediction probability value.

[0211] In some implementations, a machine learning HLA peptide presentation prediction model is used to predict that each of the plurality of test peptide sequences to be presented has a greater than a threshold presentation prediction probability value than at least one of one or more proteins encoded by a class II HLA allele in the subject's cells.

[0212] In some embodiments, for proteins encoded by the corresponding HLA alleles in Table 13, the PPV is greater than the corresponding PPV in column 2 of Table 11. In some embodiments, for proteins encoded by the corresponding HLA alleles in Table 11, the PPV is at least equal to the corresponding PPV in column 3 of Table 11.

[0213] In some implementations, for proteins encoded by HLA class II alleles, the PPV is greater than the corresponding PPV in column 2 of Table 12.

[0214] In some implementations, for proteins encoded by the corresponding HLA alleles in Table 16, the PPV is at least equal to the corresponding PPV in column 2 of Table 16.

[0215] This article provides a method for preparing personalized cancer treatment, comprising: identifying peptide sequences, wherein the peptide sequences are associated with cancer, wherein the identification includes comparing DNA, RNA, or protein sequences from a subject's cancer cells with DNA, RNA, or protein sequences from a subject's normal cells; and using a computer processor to input the amino acid position information of the identified peptide sequences into a machine learning HLA-peptide presentation prediction model to generate a set of presentation predictions for the identified peptide sequences, each presentation prediction representing the HLA sequence from the subject's cells. The probability of a given peptide sequence being presented by one or more proteins encoded by class II alleles; wherein the machine learning HLA-peptide presentation prediction model comprises: at least a plurality of predictor variables determined 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 containing amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing the relationship between the amino acid position information received as input and the presentation probability generated as output based on the amino acid position information and the predictor variables; and selecting a subset of identified peptide sequences based on the set of presentation predictions for preparation of personalized cancer treatment; wherein the prediction model has a positive predictive value of at least 0.1 at a decision rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0216] This paper provides a method comprising training a machine learning HLA-peptide presentation prediction model, wherein training includes inputting amino acid position information sequences of HLA-peptides isolated from one or more HLA-peptide complexes from cells expressing HLA class II alleles into the HLA-peptide presentation prediction model using a computer processor; the machine learning HLA-peptide presentation prediction model includes: at least a plurality of predictor variables determined based 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 containing 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 the relationship between the amino acid position information received as input and the presentation probability generated as output based on the amino acid position information and the predictor variables.

[0217] In some implementations, the presentation model has a positive predictive value of at least 0.25 at a decision rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0218] In some implementations, the presentation model has a positive predictive value of at least 0.4 at a decision rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0219] In some implementations, the presentation model has a positive predictive value of at least 0.6 at a decision rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0220] In some implementations, the mass spectrometry method is monoallelic mass spectrometry.

[0221] In some embodiments, the peptide is presented via autophagy by HLA proteins expressed in the cell.

[0222] In some embodiments, the peptide is presented by HLA proteins expressed in the cell via phagocytosis.

[0223] In some implementations, the quality of training data is improved by using multiple quality metrics.

[0224] In some implementations, the multiple quality metrics include removal of common contaminant peptides, high score peak intensity, high score, and high quality accuracy.

[0225] In some implementations, the peak intensity of the score is at least 50%.

[0226] In some implementations, the peak intensity of the score is at least 60%.

[0227] In some implementations, the score is at least 7.

[0228] In some implementations, the quality accuracy is at most 5 ppm.

[0229] In some implementations, the quality accuracy is at most 2 ppm.

[0230] In some implementations, the skeleton cutting score is at least 5.

[0231] In some implementations, the skeleton cutting score is at least 8.

[0232] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single immunoprecipitated HLA protein expressed in the cell.

[0233] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single exogenous HLA protein expressed in the cell.

[0234] In some implementations, the peptide presented by the HLA protein expressed in the cell is a peptide presented by a single recombinant HLA protein expressed in the cell.

[0235] In some implementations, the plurality of predictor variables includes peptide-HLA affinity predictor variables.

[0236] In some implementations, the plurality of predictor variables includes a predictor variable for the expression level of the source protein.

[0237] In some implementations, the plurality of predictor variables includes a peptide cleavability predictor variable.

[0238] In some embodiments, the training peptide sequence information includes sequences of peptides presented by HLA proteins, including peptides identified by searching an enzyme-specific, unmodified peptide database. In some embodiments, peptides presented by HLA proteins include peptides identified by searching a de novo peptide sequencing tool.

[0239] In some implementations, peptides presented by HLA proteins include peptides identified by searching peptide databases using a reverse database search strategy.

[0240] In some embodiments, the HLA protein includes HLA-DR and HLA-DP or HLA-DQ proteins. In some embodiments, the HLA protein includes an HLA-DR protein selected from HLA-DR and HLA-DP or HLA-DQ proteins. In some embodiments, the HLA protein includes an HLA-DR 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 ...4:02 / HLA-DPA1*01:03, HLA-DPB1*04:02 / HLA-DPA1*04:02 / HLA-DPA1*04:02 / 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).

[0241] In some embodiments, the peptides presented by HLA proteins include peptides identified by comparing the MS / MS spectra of HLA-peptides with the MS / MS spectra of one or more HLA-peptides in a peptide database.

[0242] In some implementations, the mutation is selected from point mutations, splice site mutations, frameshift mutations, readthrough mutations, and gene fusion mutations.

[0243] In some implementations, the peptides presented by HLA proteins have a length of 15-40 amino acids.

[0244] In some embodiments, the 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 said one or more isolated HLA complexes; (c) obtaining MS / MS profiles of said one or more isolated HLA-peptides; and (d) obtaining peptide sequences corresponding to the MS / MS profiles of said one or more isolated HLA-peptides from a peptide database; wherein the sequence of said one or more isolated HLA-peptides is identified from the one or more sequences obtained in step (d).

[0245] In some implementations, the personalized cancer treatment further includes an adjuvant.

[0246] In some implementations, the personalized cancer treatment further includes immune checkpoint inhibitors.

[0247] In some implementations, the training data includes structured data, time-series data, unstructured data, relational data, or any combination thereof.

[0248] In some implementations, the unstructured data includes image data.

[0249] In some implementations, the relationship data includes data from client systems, enterprise systems, operating systems, websites, web-accessible application interfaces (APIs), or any combination thereof.

[0250] In some implementations, the training data is uploaded to a cloud-based database.

[0251] In some implementations, the training is performed using a convolutional neural network.

[0252] In some implementations, the convolutional neural network comprises at least two convolutional layers.

[0253] In some implementations, the convolutional neural network (CNN) includes at least one batch normalization step.

[0254] In some implementations, the convolutional neural network includes at least one spatial dropout step.

[0255] In some implementations, the convolutional neural network includes at least one global max pooling step.

[0256] In some implementations, the convolutional neural network includes at least one dense layer.

[0257] In some implementations, identifying peptide sequences includes identifying mutated peptide sequences expressed in the subject's cancer cells.

[0258] In some implementations, identifying peptide sequences includes identifying peptide sequences that are not expressed in the subject's normal cells.

[0259] In some implementations, identifying peptide sequences includes identifying overexpressed peptide sequences.

[0260] In some implementations, identifying peptide sequences includes identifying viral peptide sequences. In one aspect, this document provides a method for identifying HLA class II specific peptides for immunotherapy that are specific to a subject, the method comprising: identifying candidate peptides containing epitopes; inputting amino acid information of a plurality of peptide sequences (each containing 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 sequences to immune cells, each presentation prediction representing the probability that one or more proteins encoded by an HLA class II allele of a subject's cells present a given peptide sequence containing the epitope; wherein the prediction model has a positive prediction value of at least 0.1 at a decision rate of at least 0.1%, 0.1%-50%, or at most 50%; selecting proteins from one or more proteins encoded by an HLA class II allele of the subject's cells, the proteins being predicted by the prediction model to bind to the candidate peptides, wherein the probability that the proteins present the candidate peptides to immune cells is greater than a threshold presentation prediction probability value; and binding the candidate peptides to HLA... Contact with proteins encoded by class II alleles, such that the candidate peptide competes with a placeholder peptide associated with a protein encoded by HLA class II alleles; and the candidate peptide is identified as a peptide for immunotherapy that is specific to proteins encoded by HLA class II alleles, depending on whether the candidate peptide replaces the placeholder peptide.

[0261] In some implementations, the immunotherapy is cancer immunotherapy.

[0262] In some embodiments, identification includes comparing DNA, RNA, or protein sequences from the subject's cancer cells with DNA, RNA, or protein sequences from the subject's normal cells. In some embodiments, the epitope is a cancer-specific epitope.

[0263] In some embodiments, the at least one protein encoded by an HLA class II allele comprises at least the α1 and β1 subunits of an HLA protein or fragment thereof present in a dimer form. In some embodiments, the placebo peptide is a CLIP peptide. In some embodiments, the placebo peptide is a CMV peptide. In some embodiments, the method further includes measuring the IC50 of the placebo peptide being substituted by the target peptide. In some embodiments, the IC50 of the placebo peptide being substituted by the target peptide is less than 500 nM. In some embodiments, at least one protein encoded by an HLA class II allele from a subject cell 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 an HLA class II allele from a subject cell is a recombinant protein. In some embodiments, the at least one protein encoded by an HLA class II allele from a subject cell is expressed in eukaryotic cells.

[0264] In one aspect, this paper provides a method for verifying the specificity of a candidate peptide in binding to an HLA class II protein. The method comprises: expressing in eukaryotic cells a polynucleotide construct containing a nucleic acid sequence encoding an HLA class II protein, the protein comprising an α-chain and a β-chain or portions thereof, capable of binding a peptide containing an MHC-II binding epitope, wherein the expressed HLA class II protein or portions thereof remain associated with a placeholder peptide; isolating the expressed HLA class II protein or portions thereof in the eukaryotic cells; and performing a peptide exchange assay by: (a) adding incremental amounts of the candidate peptide to determine whether the candidate peptide substitutes for the placeholder peptide associated with the HLA class II protein or portions thereof; and (b) calculating the IC50 of the substitution reaction to determine the affinity of the candidate peptide for the HLA class II protein or portions thereof relative to the placeholder peptide, thereby verifying the specificity of the candidate peptide in binding to the HLA class II protein.

[0265] 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.

[0266] In some embodiments, the peptide exchange assay includes detecting a fluorescent probe or tag of the peptide. In some embodiments, the occupier peptide is a CLIP peptide.

[0267] In some embodiments, the polynucleotide 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 dimerizing factor, a ribosome jumping sequence, and one or more tags for purification and / or detection.

[0268] In one aspect, this article provides a method for determining the immunogenicity of MHC class II binding peptides, the method comprising: selecting a protein encoded by an HLA class II allele that is predicted to bind to the peptide by a machine learning HLA-peptide presentation prediction model; wherein the prediction model has a positive prediction value of at least 0.1 at a decision 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 peptide with the selected protein encoded by the HLA class II allele, such that the peptide competes for and replaces a vacancy 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 measuring one or more activation parameters of the CD4+ T cells, the parameters being selected from cytokine induction, chemokine induction, and expression of cell surface markers.

[0269] 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.

[0270] In one aspect, this article 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 containing a cancer mutation, wherein the identification includes comparing a DNA, RNA, or protein sequence from the subject's cancer cells with a DNA, RNA, or protein sequence from the subject's normal cells; selecting a protein encoded by an HLA class II allele, which is normally expressed by the subject's cells 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 decision 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 HLA-mediated binding of the selected protein. The selected protein-associated placeholder peptide encoded by the class II allele is replaced with an IC50 value of less than 500 nM; the identified peptide is purified; and an effective amount of the identified peptide is administered to the subject.

[0271] In one aspect, this article provides a method for preparing HLA class II tetramers or multimers, comprising: expressing in eukaryotic cells a vector containing nucleic acid sequences encoding α and β chains of HLA proteins, a linker, one or more protease cleavage sites, a secretion signal, a biotinylated 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 placebo peptide; purifying the secreted heterodimer from a cell culture medium; verifying peptide-binding activity using a peptide exchange assay; adding streptavidin to conjugate the heterodimer into a tetramer; purifying the tetramer; and achieving a yield greater than 1 mg / L.

[0272] In some embodiments, the vector comprises a CMV promoter. In some embodiments, the vector comprises a sequence encoding a locating peptide linked to the β1 chain via a cleavable site. In some embodiments, the peptide exchange assay involves pre-cleaving the locating 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.

[0273] In one aspect, this document provides a composition comprising an HLA class II tetramer comprising HLA-DR, HLA-DP, or HLA-DQ heterodimers, each heterodimer comprising α and β chains, 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 proteins selected from HLA-DR and HLA-DP or HLA-DQ proteins.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).

[0274] In some embodiments, the heterodimer pairs are expressed in eukaryotic cells. In some embodiments, the heterodimer pairs are encoded by a vector. In some embodiments, the vector comprises: nucleic acid sequences encoding the α and β chains of an HLA protein, a secretion signal, a biotinylated 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 placebo peptide. In some embodiments, the vector comprises: nucleic acid sequences encoding the α and β chains of an HLA protein, a secretion signal, a biotinylated 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 placebo peptide.

[0275] In some implementations, HLA class II heterodimers are secreted from eukaryotic cells into cell culture medium and purified by any of the following methods: column chromatography or batch chromatography, ion exchange chromatography, size exclusion chromatography, affinity chromatography, or LC-MS.

[0276] In one aspect, this article provides a method for screening drugs containing peptide sequences for immunogenicity in a subject, the method comprising: (b) using a computer processor to input amino acid information of the peptide sequence of the peptide 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 the probability that one or more proteins encoded by HLA class I or II alleles of the subject's cells present an epitope sequence of a given peptide sequence; wherein the machine learning HLA-peptide presentation prediction model comprises: at least a plurality of predictor variables determined based 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 including amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing the relationship between the amino acid position information received as input and the 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 peptide sequence is not immunogenic to the subject; and (c) administering a composition containing the drug to the subject.

[0277] In one aspect, this article provides a method for screening drugs containing peptide sequences for immunogenicity in subjects, the method comprising: (a) using a computer processor to input amino acid information of the peptide sequence of the peptide 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 the probability of one or more proteins encoded by HLA class I or II alleles of the subject's cells presenting an epitope sequence of a given peptide sequence; wherein the machine learning HLA-peptide presentation prediction model comprises: at least a plurality of predictor variables determined based 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 including amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing the relationship between the amino acid position information received as input and the presentation probability generated as output based on the amino acid position information and the predictor variables; (b) based on the set of presentation predictions, determining or predicting that at least one of the peptide sequences of the peptide sequence is immunogenic to the subject.

[0278] In one embodiment, the method further includes deciding not to administer the drug to the subject.

[0279] In one embodiment, the drug comprises an antibody or a binding fragment thereof.

[0280] In one embodiment, the peptide sequence of the polypeptide sequence comprises each continuous peptide sequence of the polypeptide sequence having a length of 8, 9, 10, 11, or 12 amino acids, and wherein the protein encoded by the HLA class I or II allele of the subject cell is the protein encoded by the HLA class I allele of the subject cell.

[0281] In one embodiment, the peptide sequence of the polypeptide sequence comprises each continuous peptide sequence of the polypeptide sequence having a length of 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or 25 amino acids, and wherein the protein encoded by the HLA class I or II allele of the subject cell is the protein encoded by the class II MHC allele of the subject cell.

[0282] In one aspect, this article provides a method for treating a subject with an autoimmune disease or condition, comprising: (a) identifying or predicting an epitope of an expressed protein presented by HLA class I or II cells of the subject, wherein the identified or predicted epitope and a complex of HLA class I or II are 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 derived from the subject; and (d) administering the regulatory T cells expressing the TCR to the subject.

[0283] In one implementation, the autoimmune disease or condition is diabetes.

[0284] In one embodiment, the cells are pancreatic islet cells.

[0285] In one aspect, this article provides a method for treating a subject with an autoimmune disease or condition, comprising administering to the subject regulatory T cells 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 the subject's HLA class I or II, and (ii) HLA class I or II, wherein the complex is targeted by the subject's CD8 or CD4 T cells.

[0286] Other aspects and advantages of this disclosure will become apparent to those skilled in the art based on the following detailed description, which only shows and describes illustrative embodiments of the present disclosure. It should be recognized that this disclosure is capable of having other different embodiments, and that several details thereof can be modified in various obvious ways without departing from the scope of this disclosure. Therefore, the drawings and description are to be regarded in an illustrative rather than restrictive manner.

[0287] MAPTAC TM It 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 the construct was isolated, peptides that bind to HLA class II were measured to obtain sequences of peptide populations with different stabilizations using LC-MS / MS.

[0288] In one aspect, this article provides a method for treating cancer in a subject, the method comprising: identifying peptide sequences, wherein the peptide sequences are associated with cancer, wherein the identification includes comparing DNA, RNA, or protein sequences from the subject's cancer cells with DNA, RNA, or protein sequences from the subject's normal cells; and using a computer processor to input the amino acid information of the identified peptide sequences into a machine learning HLA-peptide presentation prediction model to generate a set of presentation predictions for the identified peptide sequences, each presentation prediction representing an HLA-related finding from the subject's cells. The probability of a given sequence of an identified peptide sequence presented by one or more proteins encoded by class II alleles; wherein the machine learning HLA-peptide presentation prediction model comprises: at least a plurality of predictor variables determined based on training data, wherein the training data includes: sequence information of peptide sequences 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 the relationship between amino acid position information received as input and presentation probabilities generated as output based on said amino acid position information and predictor variables; and selection of a subset of identified peptide sequences based on the set of presentation predictions for preparation of personalized cancer treatment; and administration of a composition comprising one or more of said peptides to the subject, wherein the prediction model has a positive predictive value of at least 0.1 at a decision rate of at least 0.1%, 0.1%-50%, or at most 50%.

[0289] In some implementations, the machine learning HLA-peptide presentation prediction model includes sequence information of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry after reverse-phase offline fractionation.

[0290] In some implementations, the prediction model exhibits an improvement of 1.1x to 100x compared to NetMHCIIpan. In some implementations, the prediction model exhibits improvements of 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 compared to NetMHCIIpan. 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 times or more of improvement.

[0291] Incorporation

[0292] All publications, patents, and patent applications mentioned in this specification are incorporated herein by reference to the extent that each individual publication, patent, or patent application is specifically and individually cited and incorporated herein by reference. If any publication, patent, or patent application incorporated herein by reference contradicts the disclosure contained herein, this specification is intended to supersede and / or give precedence to any such contradictory material. Attached Figure Description

[0293] The novel features of the invention are set forth in the appended claims. A better understanding of the features and advantages of the invention will be gained by referring to the following detailed description of illustrative embodiments utilizing the principles of the invention, along with the accompanying drawings (also referred to herein as “Figures”):

[0294] Figure 1A This is a diagram showing peptides docked on MHC class I proteins.

[0295] Figure 1B An exemplary illustration depicts peptides docked on MHC class II proteins.

[0296] Figure 2 An exemplary experimental method for generating data on single-allele HLA class II binding peptides is described. HLA class II peptides are introduced into any cell, including cells that do not express HLA class II, to express specific HLA class II alleles within the cells. Genetically engineered HLA-expressing cell populations are harvested, lysed, labeled (e.g., biotinylated) their HLA-peptide complexes, and immunopurified (e.g., using biotin-streptavidin interactions). HLA-associated peptides specific to a single HLA can be eluted from their labeled (e.g., biotinylated) complexes and evaluated (e.g., sequenced using high-resolution LC-MS / MS).

[0297] Figure 3 Exemplary sequence identifiers for the HLA class II-DRB1*11:01 associated peptide are depicted in the following: NeonBAP, Expi293 cell line; Neon BAP, A375 cell line; IEDB, affinity <50 nM; and all-HLA class II antibody, homozygous LCL. Figure 3 The results showed that the examples of MS-derived motifs matched known patterns and exhibited consistency across transfected cell lines.

[0298] Figure 4 This is an exemplary depiction of the performance of HLA Class II combined predictors. Figure 4This is a bar chart 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 1:19 (hit:decoy) ratio. For the NEON binding predictor neonmhc2, a separate model was built for each MHC II allele shown. The height of the bar represents the positive predictive value (PPV), defined as the fraction of bounds predicted in the validation set that are actually hit peptides. Alleles were ranked by model performance when predicting that allele.

[0299] Figure 5 An exemplary effect of the score peak intensity (SPI) threshold on combined predictor validation is depicted. Figure 5 The performance of the HLA class II binding predictor is shown when trained / validated on groups of peptides with different score peak intensity (SPI) cutoff values. For each allele-specific model trained, the model performance is shown in three settings: training and evaluating the dataset with observed MS hit peptides with ≥70 SPI, training with peptides with ≥50 SPI and validating with peptides with ≥70 SPI, and training and validating with peptides with ≥50 SPI.

[0300] Figure 6 An exemplary bar chart is depicted, showing representative data from the number of peptides observed in allele profiling analysis performed by LC-MS / MS using a score peak intensity (SPI) cutoff value greater than or equal to 70. Each bar represents the total number of peptides from the observed alleles. Data for 35 HLA-DR alleles were collected. The data for the 35 collected HLA-DR alleles represent >95% HLA-DR population coverage (US allele frequency).

[0301] Figure 7A The PPV of the model is shown for the HLA class II alleles shown, when applied to the test partition of the data. The bait peptides used are scrambled sequences of positive (hit) peptide sequences, with a hit-to-bait ratio of 1:19. PPV is determined by identifying the top 5% of peptides in the test partition and determining the score of those peptides that are positive for binding to the protein encoded by the corresponding HLA class II allele.

[0302] Figure 7B Exemplary prediction performance as a function of training set size (a curve obtained by manually downsampling the training set) is depicted. Figure 7B The results show that, typically, for the 35 HLA-DR alleles collected, the PPV value increases as the training set size increases.

[0303] Figure 8 Exemplary illustrations are depicted demonstrating that processing relevant variables can further improve predictions. Distinguishing peptides observed in MS allows for the differentiation of random sequences selected from the protein-coding exon set. On the training data partitions, logistic regression may be suitable for predicting HLA class II presentation using binding strength (NetMHCIIpan or Neon predictor) and processing features (RNA-Seq expression and derived gene-level partial terms). In separate evaluation partitions, exon locations overlapping with MHC II peptides observed in MS (“hit”) are scored alongside random exon locations not observed in MS (1:499 ratio). The top 0.2% (1 / 500) can be termed positive, and a positive predictive value is used to assess this threshold.

[0304] 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 based on the mean and standard deviation of the eigenvalues ​​in the training set. The first convolutional layer has a 9-amino acid kernel and 50 filters (also called channels) with a Modified Linear Unit (ReLU) activation function. This is followed by batch normalization, then spatial dropout at a dropout rate of 20%. This is followed by another convolutional layer with a 3-amino acid kernel and 20 filters, with a ReLU activation function, then again batch normalization and spatial dropout at a dropout rate of 20%. Global max pooling is then applied to extract the most activated neurons 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 as a binding / non-binding prediction. L2 regularization was 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. Other models used varied the number of convolutional layers and the kernel size of each layer.

[0305] Figure 10 An exemplary computer control system is depicted that is programmed or otherwise configured to implement the methods provided herein.

[0306] Figure 11A Depicting MAPTAC TM An exemplary overview of the experimental workflow.

[0307] Figure 11B The peptide counts for each allele in an exemplary cross-duplicate merging are depicted.

[0308] Figure 11C Depicting via MAPTACTM Exemplary peptide length distributions of the analyzed HLA class I and HLA class II alleles.

[0309] Figure 11D Describing MAPTAC TM Exemplary cysteine ​​frequencies per residue observed in IEBDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01 and DRB1*11:01), human proteome and multi-allele MS data from previous publications.

[0310] Figure 12A The Caucasian frequencies of HLA-DR, -DP, and -DQ alleles present in >1% of individuals were depicted, along with counts of peptides from indicated sources that were measured as strong conjugates (<50 nM).

[0311] Figure 12B An exemplary length distribution of IEDB peptides with relevant HLA class II affinity measurements is depicted.

[0312] Figure 12C Exemplary Western blotting was depicted in (1) Expi293, (2) HeLa, and (3) A375 cell lines transfected with two HLA class I alleles 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 membranes were blotted with an anti-biotin ligase epitope tag to show biotin-accepting peptide (BAP) and anti-β-tubulin as loading controls. Lanes correspond to those in MAPTAC. TM The following sub-subjects were collected during the program: lane 1, input; lane 2, biotinylated input; lane 3, input after drop-down.

[0313] Figure 12D Describing MAPTAC TM Exemplary amino acid frequencies per residue observed in IEBDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01 and DRB1*11:01), human proteome and multi-allele MS data from previous publications.

[0314] Figure 12E The Caucasian frequencies of the HLA-DR, -DP, and -DQ alleles, present in >1% of individuals, are depicted, along with counts of peptides from indicated sources measured as strong binders (<50 nM). The figure includes representations relative to... Figure 12AAdditional data. This additional data is obtained from: tools.iedb.org / main / datasets / .

[0315] Figure 12F Describing MAPTAC TM (Reduced and Alkylated), MAPTAC TM Exemplary amino acid frequencies per residue observed (without treatment) and in IEBDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01 and DRB1*11:01), human proteome and multi-allele MS data from previous publications.

[0316] Figure 13 Each MAPTAC is described TM An exemplary representation of the core binding sequence identifiers of the MHC II alleles in IEBDB. The sequence identifiers are graphical representations where the height of each amino acid is proportional to its frequency of occurrence in the peptide that binds to the MHC protein encoded by the allele. Positions with the lowest entropy are represented by color, where the color corresponds to the amino acid property. The peptides are from the dataset shown and aligned according to a CNN-based predictor (method). The identifiers represent all peptides, including those that do not closely match the overall motif (e.g., no peptides are isolated in "junk" clusters).

[0317] Figure 14A Exemplary sequence identifications of HLA-A*02:01 binding peptides (ligands) analyzed using different HLA-ligand profiling techniques in two different cell lines (A375 and expi293), including binding assays, stability assays, soluble HLA (sHLA) mass spectrometry, monoallelic mass spectrometry, and MAPTAC, are described. TM .

[0318] Figure 14B MAPTAC depicts anchors with heuristic definitions of 0, 1, 2, 3, and 4. TM Exemplary fractionation of peptides.

[0319] Figure 14C Describing MAPTAC TM Exemplary distributions of observed peptides (20 peptides per allele, each with SPI>70 and nested groups of size>=2) and NetMHCIIpan-predicted binding affinity of length-matched decoys sampled from the proteome.

[0320] Figure 15AAn exemplary architecture of a convolutional neural network (CNN) is depicted, which is trained to distinguish between monoallelic MHC peptides and decoys with out-of-order lengths. A schematic diagram illustrates two convolutional layers with different filter sizes using amino acid feature embeddings, and global max pooling as the input to the final logical output node.

[0321] Figure 15B The following are exemplary results showing the Kendall Tau statistics on the correlation between measured IEDB affinity and binding predictions from neonhc2 or NetMHCIIpan. The peptides evaluated included only those published to IEDB within one year of NetMHCIIpan's release.

[0322] Figure 16 This is an exemplary description of the performance of neonmhc2 as a function of the training dataset size.

[0323] Figure 17A MapATAC is depicted incorporating pan-DR and pan-II class MHC MS datasets. TM Exemplary cluster assignments for peptides (20 per allele). The dataset is deconvolutioned using GibbsCluster. Each box represents a MAPTAC. TM Peptide. The color of the box indicates which cluster it is assigned to, and the gray bar indicates which allele the peptide actually comes from. The total number of clusters in the Gibbs cluster scheme (right) is selected using an interaction information (MI) metric. The MI score also determines how the samples are ordered; samples with high MI schemes appear at the top.

[0324] Figure 17B Exemplary core binding sequence identifiers are depicted in multi-allelic MS data deconvolved via GibbsCluster. Each peptide group corresponds to MAPTAC. TM Clusters that best match the additives.

[0325] Figure 17C Depicting the use of MAPTAC TM Data or deconvolutional multi-allelic data prediction preserves MAPTAC TM Representative performance of peptide models. For each allele, the larger of the two data sources (usually MAPTAC) is considered. TM Downsampling is performed so that the predictor is based on the same number of training examples. NetMHCIIpan performance is shown as an additional comparison.

[0326] Figure 17D Exemplary core binding sequence markers derived from multi-allelic MS data from the sources shown are depicted.

[0327] Figure 18A An exemplary plot depicts the fraction of peptide expression (transcriptions per million (TPM)) of the source gene for peptides and random proteome decoys observed in MS (data redrawn from Schuster et al., 2017).

[0328] Figure 18B This depicts the exemplary number of observed versus expected numbers of class II peptides for each gene, as determined by a joint analysis of colorectal cancer, melanoma, and ovarian cancer datasets. (E. et al., 2018 and Schuster et al., 2017). Expected counts were derived by multiplying gene length by expression level. Expected and observed counts for relevant samples were summed. Genes known to be present in plasma were labeled according to their concentration (illustrated).

[0329] Figure 18C An exemplary distribution of enrichment scores (the ratio of observed to expected observations, such as...) is depicted for genes associated with autophagy. Figure 18B ).

[0330] Figure 18D An exemplary distribution of enrichment scores based on the location of each source gene is depicted. Source gene locations are determined using Uniprot(uniprot_sprot.dat).

[0331] Figure 18E Exemplary data are depicted, representing a comparison of the expected and observed frequencies of the total number of peptides with MHC-II binding affinity isolated based on their cellular localization properties.

[0332] Figure 18F This study presents exemplary representative data illustrating the relative consistency of observed peptides with respect to two distinct gene expression profiles. For each sample, gene-level peptide counts were modeled as a linear combination of bulk tumor gene expression and professional APC (macrophage) gene expression profiles. The ratio of coefficients determines the relative consistency of each expression profile with the peptide repertoire. Error bars correspond to 95% confidence intervals calculated via bootstrapping resampling.

[0333] Figure 19A Exemplary representative data on HLA-DRB1 expression levels from five example studies are presented. Each spot represents the average expression in cells for an individual cell type in an individual patient.

[0334] Figure 19BExemplary representative data on tumor- and matrix-derived HLA-DRB1 expression from RNA-Seq inputs from TCGA patients are depicted. 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 attributable to the tumor was estimated as min(1,2f), where f is the fraction of RNA-Seq reads in the mutated CIITA, CD74, or CTSS.

[0335] Figure 19C Exemplary representative data from additional single-cell RNA-Seq studies, including biopsies before and after checkpoint blockade immunotherapy, are depicted.

[0336] Figure 20 Exemplary representative experimental data are depicted to evaluate the predictive overall performance of natural donor tissues.

[0337] Figure 21A Exemplary representative data is depicted, showing the integrated presentation model predicting cellular HLA class II ligandomes. It represents the PPV of the pan-DR dataset at a hit:decoy ratio of 1:499 (also in...). Figure 30B and Figure 32E The analysis was conducted in [the relevant section]. The predictor uses a combination of prediction (NetMHCIIpan or neonmhc2) and optionally uses gene expression and gene preference (based on [the specific data]). Figure 32A The binding score was calculated for each candidate peptide, and it 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.

[0338] Figure 21B Exemplary representative data are depicted, demonstrating the predictive performance of tumor-derived peptides identified using SILAC, said peptides being presented by dendritic cells (analyzed from cell lysates), wherein the use of SILAC... Figure 21A The same hit rate: decoy ratio and performance metrics, with and without processing features.

[0339] Figure 21C Exemplary expression and gene preference scores of heavily labeled peptides (red spots, plotted based on K562 expression) observed in UV treatment experiments are depicted compared to lightly labeled peptides (gray spots, plotted based on DC expression).

[0340] Figure 21D An example illustration depicts the overlap of relabeled peptide source genes based on lysate and UV treatment experiments. Gene names are colored according to functional category.

[0341] Figure 22A An exemplary flowchart representing the assay protocol disclosed herein is depicted to verify HLA class II driven CD4+ T cell and T cell responses.

[0342] Figure 22B An exemplary HLA protein dimer construct design for peptide exchange assays is depicted (top figure), and an exemplary assay workflow is illustrated (bottom figure).

[0343] Figure 23 An exemplary illustration depicts the design of an exemplary vector for MHC-II expression for screening novel binding peptides, and a representation of the expressed protein product.

[0344] Figure 24 An exemplary flowchart depicts the transfection, purification, and cleavage of the occupier peptide from the β chain.

[0345] Figure 25A An exemplary illustration is depicted showing a vector encoding a CLIP peptide associated with increased secretion of the expressed MHC-II peptide.

[0346] Figure 25B Exemplary illustrations depicting shorter and longer forms of nucleic acids encoding CLIP0 and CLIP1, respectively.

[0347] Figure 25C Exemplary representative results are depicted for Coomassie gel analysis of α and β chains with or without longer clips.

[0348] Figure 26A An exemplary illustration of TR-FRET measurement is depicted.

[0349] Figure 26B Exemplary representative polarization data for HLA class II peptide binding assays using specific peptides and fluorescence resonance energy transfer (FRET) are depicted.

[0350] Figure 26C Exemplary representative polarization data for HLA class II peptide binding assays using specific peptides and fluorescence resonance energy transfer (FRET) are depicted.

[0351] Figure 26D Exemplary substitution percentages of MHC-construction-binding peptides, calculated from fluorescence enhancement, are depicted.

[0352] Figure 26E Exemplary substitution percentages of MHC-construction-binding peptides, calculated from fluorescence enhancement, are depicted.

[0353] Figure 26FExemplary peptide exchange assays obtained using differential scanning fluorometry (DSF) are depicted. The illustration shows an exemplary mechanism for detecting the dissociation of peptides from MHC class II peptides via heating, which also dissociates MHC class II heterodimers, resulting in fluorophore binding and high fluorescence. An exemplary schematic diagram of a occupier peptide translocated by an epitope peptide is also depicted. Exemplary melting curves plotted against temperature are also depicted.

[0354] Figure 26G An exemplary soluble HLA-DM construct and its use for MHC class II peptide exchange are depicted. The depicted construct comprises a CMV promoter, a coding sequence for the HLA-DM β chain, a coding sequence for the HLA-DM α chain downstream of the secretion sequence (leader region), and a BAP sequence at the 3' end of the β chain coding sequence; a His tag is located at the 3' end of the α chain coding sequence. The two chains are separated by an intermediate ribosomal skipping sequence. The construct was expressed in Expi-CHO cells, and the protein secreted into the culture medium was purified.

[0355] Figure 26H Exemplary size exclusion chromatography data for peptide exchange using HLA-sDM are shown.

[0356] Figure 27A An exemplary diagram depicts the construction of an exemplary DRB tetramer library.

[0357] Figure 27B An exemplary diagram depicts the construction of an exemplary class II tetramer group library.

[0358] Figure 27C An exemplary illustration depicts a summary of the DRB tetramer library coverage of the DRB1 alleles used for peptide exchange.

[0359] Figure 27D An exemplary coverage of human MHC class II allele generation is depicted.

[0360] Figure 27E Exemplary results of tetramer staining from samples induced by influenza epitopes (memory response) or HIV epitopes (original response) are shown.

[0361] Figure 28A An exemplary illustration depicts a method for assessing HLA class II restriction of peptides by fluorescence polarization assay, which enables screening methods to rapidly identify allelic restriction of epitope peptides. Figure 28A The assay principle described in the text allows for precise measurement of affinity and peptide exchange.

[0362] Figure 28BAn exemplary summary of several assay conditions explored in fluorescence polarization assays using DRB1*01:01 is depicted (top figure). Soluble MHC class II alleles and full-length MHC class II alleles with transmembrane domains in detergent micelles are also depicted (bottom figure), both constructed with placeholder peptides having cleavable linkers for assay purposes.

[0363] Figure 28C Describing the use of the previous research Figure 28B The figure below shows an exemplary illustration of the determination of full-length and soluble alleles. In short, both full-length and soluble alleles are expressed in the cell. The membrane-bound full-length allele form is collected by permeabilizing the membrane, while the secreted form is collected from the cell supernatant. Harvested class II HLA allele proteins are obtained by crossing nickel (Ni...) 2+ Purification was performed using a column.

[0364] Figure 28D Exemplary data illustrating that the purification method does not affect peptide potency are presented. The left side shows full-length HLA-DR1 and NiO purified using L243. 2+ The mean IC50 value of purified full-length HLA-DR1 in experiments.

[0365] Figure 28E Exemplary data illustrating that the choice between the soluble form (sDR1) and the full-length form (fDR1) does not affect peptide potency are presented. The left side shows the mean IC50 values ​​from experiments using either the sDR1 or fDR1 form. FP, fluorescence polarization.

[0366] Figure 28F An exemplary graphical view depicting the exemplary evaluation of peptides predicted by neonmhc2 and NetMHCIIpan in binding assays and the identification of inconsistent peptides.

[0367] Figure 28G Exemplary fluorescence polarization binding screening data for evaluating peptides predicted by neonmhc2 are depicted; shown as a heatmap and the percentage of probe binding inhibition indicated for each concentration of the peptide used. Green depicts good binding in proportion to color intensity. Yellow depicts moderate binding, and red depicts poor binding, also indicated by the corresponding percentage of inhibition value.

[0368] Figure 28H A summary of the evaluation of peptides predicted by neonmhc2 in exemplary binding assays is provided.

[0369] Figure 29 The average MAPTAC value for each HLA allele was depicted. TM Exemplary average peptide count from experimental replicates (50 million cells).

[0370] Figures 30A-30C Depicting HLA class II MAPTAC TM Exemplary combination of core analysis of allele + / - HLA-DM and multi-allelic deconvolution fidelity. Figure 30A Depicting according to MAPTAC TM Exemplary sequence identifiers for representative HLA-DR, -DQ, and -DP alleles were generated 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 color-coded based on their chemical properties; all others are shown in gray. Peptides were aligned using the GibbsCluster tool (supplementary method), and identifiers representing all peptides were used, including those that did not closely match the overall motif (e.g., no peptides were isolated in "junk" clusters). Figure 30B The MAPTAC incorporation into the pan-DR MS dataset is depicted. TM An exemplary description of cluster assignment for peptides (20 per allele). The dataset is deconvolved using GibbsCluster. Each colored box represents a MAPTAC. TM Peptide. The color of the box indicates which cluster it is assigned to, and the gray bar indicates which allele the peptide comes from. Figure 30C An exemplary diagram is depicted, showing for Figure 30B The alleles shown indicate the share of peptides with 0, 1, 2, 3, or 4 expected residues at anchor sites. Anchor sites are defined as the four sites with the lowest entropy, and "expected" residues are defined as those residues with a frequency of ≥10% at these sites.

[0371] Figures 31A-31F An exemplary architecture and benchmarking of neonmhc2 combined with prediction algorithms are described. Figure 31A An exemplary architecture of a convolutional neural network (CNN) is depicted, trained to distinguish between monoallelic HLA class II peptides and out-of-order length-matched decoys. The schematic illustrates the use of an amino acid feature embedding layer, two convolutional layers of width 6, the presence of jump-to-end connections, and a combination of average pooling and max pooling operations as inputs to the final logical output node. Figure 31B The map depicts MAPTAC not used for training or hyperparameter optimization. TMExemplary positive predictive value (PPV) for NetMHCIIpan and neonhc2 evaluated on the data partition. For each allele, n peptides observed by MS are scored together with 19n length-matched baits sampled from the same set of source genes, and the top n ranked peptides (e.g., top 5%) for each predictor are judged as positive. According to this evaluation scheme, PPV is the same as recall, because the number of false positives and false negatives must be equal. Figure 31C Exemplary PPVs for NetMHCIIpan and neonhc2 on the TGEM dataset are depicted. For each allele, the top n peptides are judged as positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. Figure 31D Exemplary in vitro T cell induction results for neoantigen peptides are depicted. Peptide selection was based on a high neomhc2 score and a weak NetMHCIIpan score for HLA-DRB1*11:01. Figure 31E The model trained on monoallelic MAPTAC data is compared with the deconvolutioned multiallelic data evaluated on monoallelic data. Values ​​are shown in neonmhc2, where the training dataset is downsampled to match the size of the deconvolutioned training set. Figure 31F The PPV on the TGEM dataset is shown for NetMHCIIpan-v3.1, a deconvolution-trained predictor, and neonmhc2 (with and without downsampling). For each allele, the top n peptides are judged as positive, where n is the number of immunogenic epitopes confirmed in the evaluated set.

[0372] Figures 32A-32E Exemplary gene representation and protein processing in HLA class II tumor peptides are depicted. Figure 32A This paper presents exemplary results comparing the observed and expected numbers of HLA class II peptides for each gene identified through a joint analysis of colorectal cancer, melanoma, and ovarian cancer datasets. The expected count was derived by multiplying the gene length by the expression level. The expected and observed counts for the relevant samples were summed. Genes known to be present in plasma were labeled based on their concentration. Figure 32B Exemplary results depicting the expected frequency versus observed frequency of peptides located in each cell are presented. Figure 32C Enrichment scores of genes regulated by the proteasome (the ratio of observed to expected observations, in part, such as...) were plotted. Figure 32B Examples of the distribution. The gene set includes those genes with known ubiquitination sites and those genes whose abundance increases after application of proteasome inhibitors. Figure 32DThe diagram illustrates three exemplary working models of how HLA class II peptides are processed, according to which i) cathepsins and other enzymes cleave proteins into peptide fragments that are subsequently bound by HLA, ii) proteins or unfolded polypeptides bind to HLA and are subsequently cleaved to peptide length, and iii) proteins are partially digested before binding and further pruned after binding. Each model corresponds to a different prediction method. Figure 32E The absolute increase in PPV observed for a logistic regression model that includes treatment for relevant variables and neonmhc2 binding prediction, compared to a model using binding prediction alone, is depicted. Eleven samples were evaluated using HLA-DR antibody analysis. Figure 30B The same samples were analyzed in the same way; each point corresponds to one sample. An asterisk indicates a significant improvement according to a two-tailed paired t-test (*: p < 0.01, **: p < 0.001, ***: p < 0.0001). The same analysis shows... Figure 40B Instead of using NetMHCIIpan, it uses NetMHCIIpan as the basic predictor. The methods used for bait selection and PPV calculation are similar to... Figure 31B The same method is used.

[0373] Figures 33A-33G Exemplary results for identifying and predicting tumor antigens presented by dendritic cells are described. Figure 33A An exemplary illustration depicts an experimental workflow for identifying HLA-II ligands presented by dendritic cells (DCs) derived from cancer cells (K562). Cancer cells were grown in SILAC medium until complete incorporation, lysed, or irradiated, and then seeded with monocyte-derived dendritic cells. The presented peptides were isolated using a pan-DR antibody and sequenced by LC-MS / MS. Figure 33B Depicting the use of representatives and Figure 21A Exemplary data on the predictive performance of the same hit:bait ratio and performance metrics for tumor-derived peptides presented by dendritic cells are shown. The performance of NetMHCIIpan and neonmhc2-based models with and without treatment features is illustrated. Figure 33C An exemplary gene expression distribution of the source gene of the heavily labeled peptide (red curve, plotted based on K562 expression) observed in UV treatment experiments is depicted compared to the source gene of the lightly labeled peptide (grey curve, plotted based on DC expression). Figure 33DAn exemplary illustration shows the predicted PPV of presented tumor antigens using NetMHCIIpan and neonhc2-based models with and without treatment features, at a hit:decoy ratio of 1:499. Data points from left to right represent the following samples: Donor 1, HOCl-treated cells: NetMHCIIpan continuous expression; NetMHCIIpan continuous expression + gene preference; NetMHCIIpan continuous expression + gene preference + DQ overlap, full treatment mode; Donor 1, UV-treated: neonhc2; neonhc2 + threshold expression; neonhc2 + continuous expression; neonhc2 + continuous expression + gene preference; neonhc2 + continuous expression + gene preference + DQ overlap. Figure 33E The significance of various gene localizations and functional categories in predicting heavy (K562-derived) and light (DC-derived) peptides was depicted. P-values ​​were calculated based on logistic regression controlling neonmhc2 binding scores and source gene expression. Bar colors indicate the sign associated with the regression coefficients. Figure 33F An exemplary illustration of the results shows the overlap of tumor cell-derived peptide genes (colored by functional category) in experiments treated with UV and HOCl. Figure 33G Exemplary data are depicted, showing the PPV used to predict tumor antigens presented in a second donor by fitting a logistic model to a relabeled peptide observed in the first donor. Models were fitted using neonmhc2 binding alone; binding and expression; or a binary variable of binding, expression, and an indicator of whether the peptide originates from a mitochondrial gene.

[0374] Figures 34A-34B Depicting and Figure 29 Related MAPTAC TM An exemplary representation of the data. Figure 34A MAPTAC was described by encoding affinity marker HLA-A*02:01-BAP. TM An exemplary HLA cell surface analysis performed using FACS on the Expi293 cell line transfected with the construct. Figure 34B MAPTAC was described by using HLA-DRB1*11:01-BAP with encoded affinity markers. TMExemplary HLA cell surface analysis performed using FACS on the construct-transfected Expi293 cell line (see figure below). HLA cell surface expression in transfected Expi293 cells (orange) is compared to stained untransfected Expi293 cells (blue), unstained untransfected Expi293 cells (red), stained PBMCs (dark green), and unstained PBMCs (light green). W6 / 32 (pan-HLA class I) was used for all HLA class I staining, while REA332 (pan-HLA class II) was used for HLA class II staining.

[0375] Figure 35 Depicting and Figure 30A Related MAPTAC TM An exemplary comparison with the IEDB identifier. The measured affinity of the peptide observed by MS and the affinity predicted by NetMHCIIpan did not show a good NetMHCIIpan score, but were well supported by MS (score peak intensity >70 and nesting group size ≥1).

[0376] Figures 36A-36C Depicting and Figures 30A-30C Related HLA-DR1 MAPTAC TM An exemplary analysis of data fidelity. Figure 36A The graph depicts HLA-DR1MAPTAC compared to 50,000 length-matched bait peptides (blue) randomly sampled from the proteome for common alleles. TM Exemplary NetMHCIIpan 3.1 scores for peptides (green) (length 12-23). Figure 36B Exemplary measured affinity and NetMHCIIpan predicted affinity of peptides observed by exemplary MS are depicted. These peptides did not exhibit good NetMHCIIpan scores but were well supported by MS (score peak intensity >70 and nesting group size ≥1). Figure 36C Depicting, as via MAPTAC TM Exemplary HLA class II sequence identifiers of the HLA-DRB1 allele, as determined in different cell types.

[0377] Figures 37A-37C Depicting and Figures 30A-30C Related MAPTAC TM Additional exemplary analysis of motifs. Figure 37A MAPTAC describes experiments using HLA-DM co-transfection (expi293 cell line) and without. TM Derived sequence identifiers. Figure 37B Depicting according to MAPTAC TMSequence identifiers of several HLA class I alleles from 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 differs from previous studies using multi-allele deconvolution; the identifier of B*52:01 has not been previously disclosed. Figure 37C Depicting MAPAC TM An exemplary alignment of the observed peptides with the CD74 gene sequence.

[0378] Figure 38Ai-38D Depicting and Figures 31A-31D Related exemplary neonmhc2 performance statistics and T cell flow staining. Figure 38Ai An exemplary performance of neonmhc2 as a function of the training dataset size is depicted. (Compared to...) Figure 31B The same method and the same evaluation peptide were used to evaluate PPV; however, the training data was randomly downsampled to simulate a smaller training dataset. Figure 38Aii Exemplary sequence identifications of peptide clusters derived from the multi-allelic HLA-DR ligand set using GibbsCluster (default setting; "Allow junk clusters") are depicted. Figure 38B An exemplary representative flow cytometry plot depicts 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 expressing IFN-γ when determined in the absence of neoantigens (peptide-free). The two flow cytometry plots on the left represent neoantigens that induce T cell responses in CD4+ T cells (PEASLYGALSKGSGG) and neoantigens that do not induce T cell responses (PATYILILKEFCLVG). Figure 38C Exemplary delta values ​​from wells determined using a single neomch2 neoantigen peptide are depicted. A peptide with a positive response (highlighted as a delta response greater than 3%) is considered an induced hit. Figure 38D The example sequence identification of peptide clusters derived from the multi-allelic HLA-DR ligand set is shown using GibbsCluster (default setting; "junk" clusters allowed).

[0379] Figures 39A-39C Depicting and Figures 32A-32E Additional exemplary origin cell analysis of relevant HLA class II. Figure 39A The study describes how the presence of peptide-derived genes in human plasma, through pan-DR antibodies (from...) Figure 30BExemplary percentile rankings of the neonmhc2 score for HLA class II peptides observed in four PBMC samples (RG1248, RG1104, RG1095, and HDSC) are presented. For each peptide, the best (lowest) percentile ranking was used among the alleles present in the donor. Scores of proteomic baits of random length matching are shown for comparison. Box plots label the 5th, 25th, 50th, 75th, and 95th percentiles. Figure 39B Describes the use and Figure 32A The same method was used to perform exemplary counts of observed peptides versus expected peptides for each gene in HLA class I. The data correspond to the same tumor types (colorectal cancer, ovarian cancer, and melanoma). Genes present in human plasma are highlighted in blue and their size is determined according to their concentration. Figure 39C Exemplary relative consistency of peptide observations across two distinct gene expression profiles is depicted. For each sample, gene-level peptide counts are modeled as a linear combination of bulk tumor gene expression and professional APC gene expression profiles. The ratio of coefficients determines the relative consistency of each expression profile with the peptide library. Error bars correspond to 95% confidence intervals calculated via bootstrap resampling.

[0380] Figures 40A-40B Depicting and Figures 32A-32E Additional exemplary analysis of the related processing motifs. Figure 40A The study described the use of donor PBMCs, monocyte-derived dendritic cells, colorectal cancer, melanoma, ovarian cancer, and the expi293 cell line (used in most MAPTACs). TM Exemplary amino acid frequencies near N-terminal and C-terminal peptide cleavage sites, observed in the data generation process, relative to the average proteome frequency (applicable to upstream positions U3-U1 and downstream positions D1-D3) or relative to the average peptide frequency (applicable to internal positions N1-C1). Figure 40B The description covers aspects other than using NetMHCIIpan as the basic predictor. Figure 32E The same analysis was performed on 8 samples analyzed by HLA-DR antibody. Figure 31B The absolute increase in PPV observed in the logistic regression model that treated relevant variables in addition to NetMHCIIpan predictions (compared to the model using only NetMHCIIpan) was analyzed for the same samples in the analysis. Asterisks indicate significant improvements according to the paired two-tailed t-test (*: p < 0.01, **: p < 0.001, ***: p < 0.0001).

[0381] Figure 41 An exemplary naming system for referring to upstream, intrapeptide, and downstream locations of peptides is described.

[0382] Figure 42AThe illustrations depict an exemplary workflow for analyzing endogenously processed and HLA-1 and HLA class II presented peptides via nLC-MS / MS.

[0383] Figure 42B The illustrations depict exemplary experimental results from nLC-MS / MS analysis of trypsin peptides with and without FAIMS. Representative overlap of HLA-1 and HLA class II peptides detected by nLC-MS / MS analysis with and without FAIMS is also depicted at the analytical scales shown.

[0384] Figure 43A Exemplary HLA class I acidic and basic reverse-phase fractionation peptide detection with and without FAIMS are described.

[0385] Figure 43B Exemplary experimental results are depicted, showing the detection of unique HLA class I binding peptides plotted by retention time.

[0386] Figure 44A Exemplary HLA class II acidic and basic reverse-phase fractionation peptide detection with and without FAIMS are described.

[0387] Figure 44B Exemplary experimental results are depicted, showing the detection of unique HLA class II binding peptides plotted by retention time.

[0388] Figure 45 An exemplary illustration (left) depicting the cross-size of HLA class I binding peptides detected using the method shown, and a Venn diagram (right) illustrating an exemplary standard workflow and an optimized workflow for LC-MS / MS detection of HLA class I binding peptides.

[0389] Figure 46 An exemplary illustration (left) depicting the cross-size of HLA class II binding peptides detected using the method shown, and a Venn diagram (right) illustrating an exemplary standard workflow and an optimized workflow for LC-MS / MS detection of HLA class II binding peptides. Detailed Implementation

[0390] All terms should be understood in the manner that they would be understood by someone 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 pertains.

[0391] The chapter titles used in this article are for organizational purposes only and should not be construed as limiting the topics described.

[0392] Although the various features of this disclosure may be described in the context of a single embodiment, these features may also be provided individually or in any suitable combination. Conversely, although this disclosure may be described in the context of a single embodiment for clarity, this disclosure may also be implemented in a single embodiment.

[0393] This disclosure is based on the following important finding: the use of a novel computer-based machine learning HLA-peptide presentation prediction model—which allows for the use of HLA class II specific peptides to improve immunotherapy and can predict with high confidence the presentation of antigens, particularly cancer antigens, by specific HLA class II α and β chain pairs.

[0394] In one aspect, this disclosure provides a method for predicting peptides that can accurately pair or bind to specific HLA class II α and β heterodimers. This high-fidelity binding of the peptide to HLA class II proteins (including α and β heterodimers) ensures the presentation of the specific peptide to T lymphocytes, thereby triggering a specific immune response and avoiding any cross-reactivity or immune confounding. Several recent studies have shown that CD4+ T cells can also recognize HLA class II presented ligands and contribute to tumor control. Ideally, cancer vaccines and other immunotherapies would utilize guiding CD4+ T cell responses, but current efforts have completely abandoned HLA class II antigen prediction due to the insufficient accuracy of current predictive tools.

[0395] In one aspect, this disclosure provides a method for predicting peptides that can accurately bind to specific HLA class II proteins, such that when the peptide is therapeutically administered to a subject expressing a specific homologous HLA class II protein, the peptide can activate a more sustained and robust immune response by leveraging the ability of HLA class II proteins to activate CD4+ T cells and stimulate immune memory. In some embodiments, the method provided herein shows improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein shows at least about 1.1-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein shows at least about 2-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein shows at least about 3-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein shows at least about 4-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 5-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 6-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 7-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 8-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 9-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 10-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 15-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least about 20-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least approximately 30-fold improvement in predicting specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least approximately 40-fold improvement in predicting specific HLA class II proteins compared to currently available predictors.In some embodiments, the method provided herein exhibits at least approximately 50-fold improvement in the prediction of specific HLA class II proteins compared to currently available predictors. In some embodiments, the method provided herein exhibits at least approximately 60-fold improvement in the prediction of specific HLA class II proteins compared to currently available predictors.

[0396] On one hand, this article proposes immunotherapies tailored or personalized for specific subjects. 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 for the identification of a 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 with high fidelity which peptides can bind to specific HLA class II α and β chain heterodimers can ensure the generation of a specific immune response tailored specifically for that subject.

[0397] In this application, unless otherwise specified, the use of the singular form includes the plural form. It must be noted that, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” as used herein include plural indicators. In this application, unless otherwise specified, the use of “or” means “and / or.” Furthermore, the use of the term “comprising” and other forms such as “including,” “containing,” and “having” is not restrictive. The terms “one or more” or “at least one,” such as one or more or at least one member of a group of members, are self-explanatory, and by further exemplification, this term particularly includes reference to any one of the stated members, or any two or more of the stated members, for example, any 3, 4, 5, 6, or 7 of the stated members, up to all of the stated members.

[0398] The references to “some implementation schemes,” “implementation scheme,” “one implementation scheme,” or “other implementation schemes” in this specification mean that the features, structures, or characteristics described in connection with that implementation scheme are included in at least some of the implementation schemes disclosed herein, but not necessarily in all of them.

[0399] As used in this specification and claims, the terms “comprising” (and any form of inclusion), “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 unlisted elements or method steps. It is conceivable that any embodiment discussed in this specification can be implemented using any method or composition of this disclosure, and vice versa. Furthermore, the compositions of this disclosure can be used to implement the methods of this disclosure.

[0400] When referring to measurable values ​​such as parameters, quantities, or durations, the terms “about” or “approximately” as used herein are intended to cover variations of + / -20% or less, + / -10% or less, + / -5% or less, or + / -1% or less of the specified value, provided such variations are appropriate in this disclosure. It should be understood that the values ​​referred to by the modifiers “about” or “approximately” are themselves specifically disclosed.

[0401] The term "immune response" includes T-cell-mediated and / or B-cell-mediated immune responses regulated by T-cell co-stimulation. Exemplary immune responses include T-cell responses, such as cytokine production and cytotoxicity. Additionally, the term "immune response" includes immune responses indirectly influenced by T-cell activation, such as antibody production (humoral response) and activation of cytokine-responsive cells, such as macrophages.

[0402] "Receptor" should be understood as referring to a biomolecule or group of molecules capable of binding a ligand. Receptors can be used to transmit information in cells, cell formation, or organisms. A receptor contains 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. The receptor has a structure complementary to that of the ligand and can complex with the ligand as a binding partner. Signaling information can be transmitted through conformational changes following the binding of the receptor to the ligand on the cell surface. According to this disclosure, a receptor may refer to MHC class I and II proteins capable of forming receptor / ligand complexes with ligands, such as peptides or peptide fragments of appropriate length. Class I and II MHC peptides encoded by HLA class I and 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 HLA class II proteins, or HLA class I and HLA class II proteins, or HLA class I and HLA class II molecules, or such common variants thereof, as are well known to those skilled in the art in the context of this discussion.

[0403] A "ligand" is a molecule capable of forming a complex with a receptor. According to this disclosure, a ligand should be understood to mean, for example, a peptide or peptide fragment having 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 MHC class I or MHC class II proteins (i.e., HLA class I and HLA class II proteins) and forming a complex.

[0404] An 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 (CTLs)) are not recognized as intact proteins, but as small peptides associated with HLA class I or II proteins on the cell surface. In a naturally occurring immune response, antigens recognized by association with HLA class II molecules on antigen-presenting cells (APCs) are acquired from outside the cell, internalized, and processed into small peptides that associate 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 typically intracellularly produced peptides, 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 II molecule are characterized as having a common binding motif, and binding motifs targeting a large number of different HLA class I and II molecules have been identified. Synthetic peptides corresponding to the amino acid sequence of a given antigen and containing a binding motif targeting a given HLA class I or II molecule can also be synthesized. These peptides can then be added to an appropriate APC, and the APC can be used to stimulate T helper cell or CTL responses in vitro or in vivo. The binding motifs, methods of synthesizing the peptides, and methods of stimulating T helper cell or CTL responses are all known and readily available to those skilled in the art.

[0405] In this specification, the term "peptide" is used interchangeably with "mutant peptide" and "neoantigen peptide." Similarly, in this specification, the term "polypeptide" is used interchangeably with "mutant polypeptide" and "neoantigen polypeptide." A "neoantigen" or "neoepitope" refers to a class of tumor antigens or tumor epitopes resulting from tumor-specific mutations in expressed proteins. This disclosure further includes peptides containing tumor-specific mutations, peptides containing known tumor-specific mutations, and mutant polypeptides or fragments thereof identified by the methods of this disclosure. These peptides and polypeptides are referred to herein as "neoantigen peptides" or "neoantigen polypeptides." These polypeptides or peptides may have various lengths, may be in a neutral (uncharged) form or a salt form, and may be free of modifications such as glycosylation, side-chain oxidation, phosphorylation, or any post-translational modifications, or contain such modifications, provided that the modifications do not impair the biological activity of the polypeptides described herein. In some embodiments, the neoantigen peptide of this disclosure may include: 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, a length of 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 neoantigen peptide or neoantigen polypeptide contains a novel epitope.

[0406] The term "epitope" includes any protein determinant capable of specifically binding to antibodies, antibody peptides, and / or antibody-like molecules (including, but not limited to, T-cell receptors) as defined herein. Epitope determinants typically consist of chemically active surface groups of a molecule, such as amino acid or sugar side chains, and usually possess specific three-dimensional structural features and specific charge characteristics.

[0407] "T cell epitopes" are peptide sequences that can be bound by class I or II MHC molecules in the form of peptide-presenting MHC molecules or MHC complexes, and then recognized and bound by cytotoxic T lymphocytes or T helper cells in this form, respectively.

[0408] 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 intact antibodies, including single-chain intact antibodies and their antigen-binding (Fab) fragments. Antigen-binding antibody fragments include, but are not limited to, Fab, Fab', and F(ab')2, Fd (composed of VH and CH1), single-chain variable fragments (scFv), single-chain antibodies, disulfide-linked variable fragments (dsFv), and fragments containing VL or VH domains. Antibodies can be derived from any animal source. Antigen-binding antibody fragments, including single-chain antibodies, may contain, alone or in combination with all or some of the following variable regions: hinge region, CH1, CH2, and CH3 domains. Any combination of variable regions and hinge regions, CH1, CH2, and CH3 domains is also included. Antibodies can be, for example, monoclonal antibodies, polyclonal antibodies, chimeric antibodies, humanized antibodies, and human monoclonal and polyclonal antibodies that specifically bind to HLA-associated peptides or HLA-HLA-binding peptide (HLA-peptide) complexes. 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 peptides, for example, those cleaved from a membrane by proteolysis. This includes techniques in which (1) one or more antibodies capable of specifically binding to soluble proteins are immobilized on a fixed or movable substrate (e.g., plastic pores 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 containing the antibodies and the bound soluble proteins is separated from the solution, and optionally, for example, the antibodies and soluble proteins are dissociated by changing the pH and / or ionic strength and / or ionic composition of the solution in which the antibodies are bathed. Alternatively, immunoprecipitation can be used, in which antibodies and soluble proteins are combined to form large molecular aggregates. These macromolecular aggregates can then be separated from the solution using size exclusion techniques or by centrifugation.

[0409] The term "immunopurification (IP)" (or immunoaffinity purification or immunoprecipitation) is a well-known method in the art and is widely used to isolate desired antigens from samples. Typically, this method involves contacting a sample containing the desired antigen with an affinity matrix containing antibodies against that antigen covalently attached to a solid phase. The antigen in the sample binds 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 altering the chemical composition of the solution in contact with the matrix. Immunopurification can be performed on a column containing the affinity matrix, in which case the solution is the eluent. Alternatively, immunopurification can be a batch process, in which case the affinity matrix is ​​held as a suspension in solution. A crucial step in this process is the removal of the antigen from the matrix. This is typically achieved by increasing the ionic strength of the solution in contact with the affinity matrix, for example, by adding an inorganic salt. Changes in pH can also effectively dissociate the immunochemical bonds between the antigen and the affinity matrix.

[0410] "Pharmaceutical" is any small molecule compound, antibody, nucleic acid molecule, or polypeptide or fragment thereof.

[0411] "Change" or "transformation" means an increase or decrease. Changes can be as small as 1%, 2%, 3%, 4%, 5%, 10%, 20%, 30% or 40%, 50%, 60%, or even as large as 70%, 75%, 80%, 90% or 100%.

[0412] "Biological sample" is any tissue, cell, fluid, or other substance derived from a living organism. As used herein, the term "sample" includes biological samples, such as any tissue, cell, fluid, or other substance derived from a living organism. "Specific binding" refers to a compound (e.g., peptide) that recognizes and binds to molecules (e.g., polypeptides) but substantially does not recognize and bind to other molecules in a sample (e.g., a biological sample).

[0413] "Capture reagent" refers to a reagent that specifically binds to a molecule (e.g., a nucleic acid molecule or a polypeptide) to select or separate that molecule (e.g., a nucleic acid molecule or a polypeptide).

[0414] As used herein, the terms “determine,” “assess,” “determine,” “measure,” “detect,” and their grammatical equivalents refer to quantitative and qualitative determination; therefore, the term “determine” is used interchangeably with “determine,” “measure,” etc., in this document. When quantitative determination is desired, the phrase “determine the amount of the analyte, etc.” is used. When qualitative and / or quantitative determination is desired, the phrase “determine the level of the analyte” or “detect” the analyte is used.

[0415] 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%, 80%, 90%, 95%, or even 99% of the biological activity of the reference protein or nucleic acid described herein.

[0416] The terms “isolated,” “purified,” “biologically pure,” and their grammatical equivalents refer to substances released to varying degrees from their natural components. “Isolation” indicates the degree of separation from the original source or environment. “Purity” indicates a degree of separation beyond isolation. A “purified” or “biologically pure” protein is sufficiently free of other substances such that any impurities do not materially affect the protein’s biological properties or cause other adverse consequences. That is, the nucleic acids or peptides disclosed herein are purified if they are substantially free of cellular material, viral material, or culture medium when produced by recombinant DNA technology, 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 the nucleic acid or protein produces essentially one band in the electrophoresis gel. For proteins that can be modified, such as phosphorylated or glycosylated, different modifications can produce different isolated proteins, which can be purified separately.

[0417] "Isolated" polypeptides (e.g., peptides from HLA-peptide complexes) or polypeptide complexes (e.g., HLA-peptide complexes) are polypeptides or polypeptide complexes of this disclosure that have been isolated from their naturally associated components. Typically, a polypeptide or polypeptide complex is considered isolated when it contains at least 60% by weight of proteins and naturally occurring organic molecules naturally associated with it. The article may be at least 75%, at least 90%, or at least 99% by weight of the polypeptides or polypeptide complexes of this disclosure. The isolated polypeptides or polypeptide complexes of this 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 the polypeptide or one or more components of the polypeptide complex. Purity can be measured by any suitable method, such as column chromatography, polyacrylamide gel electrophoresis, or analysis by HPLC. In some cases, the MHC class II protein encoded by the HLA allele (i.e., MHC class II peptide) may be interchangeably referred to as HLA class II protein (or HLA class II peptide) in this document.

[0418] 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 type of vector 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 operatively linked thereto is referred to herein as an "expression vector." Typically, useful expression vectors are usually in the form of a "plasmid," which refers to a circular double-stranded DNA molecule that does not bind to a chromosome in vector form and typically contains 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, bacteriophages, or viral vectors, and such vectors can integrate into the host genome or replicate autonomously within the cell. Vectors can be DNA or RNA vectors. Other forms of expression vectors known to those skilled in the art that perform equivalent functions may also be used, such as self-replicating extrachromosomal vectors or vectors capable of integrating into the host genome. Exemplary vectors are those capable of autonomously replicating and / or expressing nucleic acids linked thereto.

[0419] The term "spacer" or "linker" used for fusion proteins refers to a peptide that links a protein containing the fusion protein. Typically, spacers have no specific biological activity other than linking or maintaining a 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 influence certain properties of the molecule, such as folding, net charge, or hydrophobicity. Suitable linkers used in embodiments of this disclosure are well known to those skilled in the art and include, but are not limited to, straight-chain or branched-chain carbon linkers, heterocyclic carbon linkers, or peptide linkers. Linkers are used to separate two antigenic peptides at a distance sufficient to ensure that each antigenic peptide folds correctly in some embodiments. Exemplary peptide linker sequences employ a flexible, extended conformation and do not exhibit a tendency to develop ordered secondary structures. Typical amino acids in flexible protein regions include Gly, Asn, and Ser. In fact, any arrangement of amino acid sequences containing Gly, Asn, and Ser is expected to meet the above criteria for linker sequences. Other near-neutral amino acids, such as Thr and Ala, may also be used in linker 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, U.S. Patent 4,935,233 and U.S. Patent 4,751,180.

[0420] The term "tumor formation" refers to any disease that results in an inappropriately high level of cell division, an inappropriately low level of apoptosis, or both, or both. Glioblastoma is a non-limiting example of tumor formation or cancer. The terms "cancer," "tumor," or "hyperproliferative disorder" refer to the presence of cells exhibiting typical characteristics of cancerous cells, such as uncontrolled proliferation, unlimited proliferation, metastatic potential, rapid growth and proliferation rate, and certain characteristic morphological features. Cancer cells are usually in the form of tumors, but such cells can exist alone in an animal or as non-tumorigenic cancer cells, such as leukemia cells. Cancers include, but are not limited to, B-cell cancers, such as multiple myeloma, Waldenstrom's macroglobulinemia, heavy chain diseases (e.g., alpha chain disease, gamma chain disease, and μ chain disease), benign monoclonal gammopathy and immune cell amyloidosis, melanoma, breast cancer, lung cancer, bronchial cancer, colorectal cancer, prostate cancer (e.g., metastatic, hormone-resistant prostate cancer), pancreatic cancer, stomach cancer, ovarian cancer, bladder cancer, cancers of the brain or central nervous system, cancers of the peripheral nervous system, esophageal cancer, cervical cancer, uterine or endometrial cancer, cancers of the oral cavity or pharynx, liver cancer, kidney cancer, testicular cancer, bile duct cancer, small bowel or appendix cancer, salivary gland cancer, thyroid cancer, adrenal cancer, osteosarcoma, chondrosarcoma, and cancers of blood tissues. Other non-limiting examples of cancer types suitable for the methods covered in this disclosure include human sarcomas and carcinomas, such as fibrosarcoma, myxosarcoma, liposarcoma, chondrosarcoma, osteosarcoma, chordoma, angiosarcoma, endothelial sarcoma, lymphangiosarcoma, lymphangioendothelial sarcoma, synovoma, mesothelioma, and Ewing's tumor. Tumors, leiomyosarcomas, rhabdomyosarcomas, colon cancer, colorectal cancer, pancreatic cancer, breast cancer, ovarian cancer, squamous cell carcinoma, basal cell carcinoma, adenocarcinoma, sweat gland carcinoma, sebaceous gland carcinoma, papillary carcinoma, papillary gland carcinoma, cystic adenocarcinoma, medullary carcinoma, bronchial carcinoma, renal cell carcinoma, hepatocellular carcinoma, bile duct carcinoma, 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 gland. Somatic tumors, hemangioblastomas, acoustic neuromas, oligodendrogliomas, meningiomas, melanomas, neuroblastomas, retinoblastomas; leukemias, such as acute lymphoblastic leukemia and acute myeloid leukemia (myeloblastic, promyelocytic, granulocytic, 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, Waldenström macroglobulinemia, and heavy chain disease.In some embodiments, the cancer is epithelial carcinoma, 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 still other embodiments, the epithelial carcinoma is non-small cell lung cancer, non-papillary renal cell carcinoma, cervical cancer, ovarian cancer (e.g., serous ovarian cancer), or breast cancer. Epithelial carcinoma can be characterized in various other ways, including but not limited to serous, endometrioid, mucinous, clear cell, Brenner type, or undifferentiated. In some embodiments, this 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 proliferative diseases.

[0421] The term "vaccine" should be understood to refer to a composition intended to generate immunity for the prevention and / or treatment of a disease (e.g., tumor formation / tumor / infectious agent / autoimmune disease). Therefore, a vaccine is a medicine containing an antigen and intended for use in humans or animals to produce specific defenses and protective substances through inoculation. A "vaccine composition" may contain pharmaceutically acceptable excipients, carriers, or diluents. Aspects of this disclosure relate to the use 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 the corresponding nucleic acids encoding them). In some implementations, 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 ten, at least eleven, at least twelve, at least thirteen, at least fourteen, at least fifteen, at least sixteen, at least seventeen, at least eighteen, at least nineteen, at least twenty, at least twenty-one, at least twenty-two, at least twenty-three, at least twenty-four, at least twenty-five, at least twenty-six, at least twenty-seven, at least twenty-eight, at least twenty-nine, at least thirty, or more antigenic peptides. In some implementations, 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 Up to 8, 3 to 7, 3 to 6, 3 to 5, 4 to 100, 4 to 75, 4 to 50, 4 to 25, 4 to 20, 4 to 19, 4 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 implementations, 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 neoantigen peptide. In some cases, the antigenic peptide contains one or more novel epitopes.

[0422] The term "pharmaceutically acceptable" means approved or permitted by a federal or state regulatory agency, or listed in the United States Pharmacopeia or other recognized pharmacopoeia for use in animals, including humans. "Pharmaceutically acceptable excipients, carriers, or diluents" means excipients, carriers, or diluents that can be administered to a subject with the drug without destroying its pharmacological activity and are non-toxic when administered at a dose sufficient to deliver a therapeutic amount of the drug. As described herein, a "pharmaceutically acceptable salt" of a combined disease-specific antigen can be an acidic or basic salt that is generally considered in the art suitable for contact with human or animal tissues without excessive toxicity, irritation, allergic reactions, 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-aminosulfonic 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, alkanonic 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). Generally, pharmaceutically acceptable acid or basic salts can be synthesized from parent compounds containing a basic or acidic moiety by any conventional chemical method. In short, such salts can be prepared by reacting the free acidic or basic form of these compounds with a stoichiometric amount of a suitable base or acid in a suitable solvent.

[0423] Nucleic acid molecules that can be used in the methods of this disclosure include any nucleic acid molecule encoding a polypeptide or fragment thereof of the disclosed method. Such nucleic acid molecules need not be 100% identical to an endogenous nucleic acid sequence, but generally exhibit substantial identity. Polynucleotides having substantial identity with an endogenous sequence are generally capable of hybridizing with at least one strand of a double-stranded nucleic acid molecule. “Hybridization” refers to the pairing of nucleic acid molecules between complementary polynucleotide sequences or portions thereof under various stringent conditions to form a double-stranded molecule. (See, for example, Wahl, GM and SLBerger (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. Low-tightness hybridization can be obtained in the absence of organic solvents such as formamide, while high-tightness hybridization can be obtained in the presence of at least about 35% or at least about 50% formamide. Tight temperature conditions typically include temperatures of at least about 30°C, at least about 37°C, or at least about 42°C. Variations in other parameters, such as hybridization time, detergent concentrations such as sodium dodecyl sulfate (SDS), and inclusion or exclusion of vector DNA, are well known to those skilled in the art. Various levels of tightness can be achieved by combining these various conditions as needed. In one exemplary embodiment, hybridization can occur at 30°C in 750 mM NaCl, 75 mM trisodium citrate, and 1% SDS. In another exemplary embodiment, hybridization can be performed at 37°C in 500 mM NaCl, 50 mM 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 regarding these conditions will be apparent to those skilled in the art. For most applications, the stringency of the post-hybridization wash step may also vary. Wash stringency conditions can be defined by salt concentration and temperature. As described above, wash stringency can be increased by decreasing 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. The stringent temperature conditions for the wash step can include temperatures of at least about 25°C, at least about 42°C, or at least about 68°C.In an exemplary embodiment, the washing step may be performed at 25°C in 30 mM NaCl, 3 mM trisodium citrate, and 0.1% SDS. In other exemplary embodiments, the washing step may be performed at 42°C in 15 mM NaCl, 1.5 mM trisodium citrate, and 0.1% SDS. In yet another exemplary embodiment, the washing step may be performed at 68°C in 15 mM NaCl, 1.5 mM trisodium citrate, and 0.1% SDS. Other variations of these conditions will be apparent to those skilled in the art. Hybridization techniques are well known to those skilled in the art and have been described, for example, by 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.

[0424] "Substantially identical" means a polypeptide or nucleic acid molecule that shows at least 50% identity with a reference amino acid sequence (e.g., any of the amino acid sequences described herein) or nucleic acid sequence (e.g., any of the nucleic acid sequences described herein). Such sequences may be at least 60%, 80%, or 85%, 90%, 95%, 96%, 97%, 98%, or even 99% or higher at the amino acid or nucleic acid level of the sequence used for comparison. 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 program). Such software matches identical or similar sequences by assigning degrees of homology to various substitutions, deletions, and / or other modifications. Conservative substitutions typically 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 procedure can be used, where the probability score between e-3 and em° represents closely related sequences. "Reference" is the comparison standard.

[0425] The terms “subject” or “patient” refer to an animal used as a subject of treatment, observation, or experimentation. By way of example only, subjects include, but are not limited to, mammals, including, but not limited to, human or non-human mammals such as non-human primates, rats, cattle, horses, dogs, sheep, or cats.

[0426] The terms “treatment,” “management,” etc., refer to reducing, preventing, or improving a condition and / or related symptoms (e.g., tumor formation or tumor, infectious agent, or autoimmune disease). “Treatment” can refer to the administration of treatment to a subject after the onset or suspected onset of a disease (e.g., cancer, 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 and / or treatment-related side effects or recurrences associated with the disease. The term “treatment” also encompasses the concept of “management,” which refers to reducing the severity of a disease or condition in a patient, for example, prolonging the life or survival of a patient with the disease, or delaying its recurrence, for example, prolonging the remission period in a patient already suffering from the disease. It should be understood that, although not excluded, treating a disease or condition does not necessarily require the complete elimination of the condition, illness, or related symptoms.

[0427] As used herein, the terms “prevention,” “avoidance,” and their grammatical equivalents refer to the avoidance or delay of the onset of symptoms associated with a disease or condition in subjects who have not yet developed such symptoms at the time of initiation of administration of a drug or compound.

[0428] The term "therapeutic effect" refers to a certain degree of relief of one or more symptoms or associated pathology of a condition (e.g., tumor formation, tumor, infection with an infectious agent, or autoimmune disease). As used herein, "therapeutic effective amount" refers to a dose of medicine that, after a single or multiple administration to cells or a subject, effectively prolongs the survival of a patient with such a condition, reduces one or more signs or symptoms of the condition, prevents or delays the occurrence of the condition beyond what would be expected without such treatment. "Therapeutic effective amount" is intended to define the amount required to achieve a therapeutic effect. A physician or veterinarian with ordinary skills in the art can readily determine and prescribe a "therapeutic effective amount" (e.g., ED50) of the desired pharmaceutical composition. For example, a physician or veterinarian may begin the use of the compound of the invention in the pharmaceutical composition at a level below that required to achieve the desired therapeutic effect and gradually increase the dose until the desired effect is achieved. Disease, condition, and symptom are used interchangeably herein.

[0429] Those skilled 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 an 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); and herpes simplex virus glycoprotein D (gD) tags (e.g., Paborsky et al., Protein Engineering, 3:547, 1990); FLAG tags (e.g., Hopp et al., BioTechnology, 6:1204, 1988; US Patents 4,703,004 and 4,851,341); KT3 epitope tags (e.g., Martine et al., Science, 255:192, 1992); tubulin epitope tags (e.g., Skinner, Biol. Chem., 266:15173, 1991); T7 gene 10 protein peptide tags (e.g., Lutz-Freyemuth et al., P... roc. Natl. Acad. Sci. USA, 87:6393, 1990); streptavidin tags (StrepTag™ or StrepTagII™; see, for example, Schmidt et al., J. Mol. Biol., 255(5):753-766, 1996 or U.S. Patent 5,506,121; also available commercially from Sigma-Genosys); or VSV-G epitope tags derived from vesicular stomatitis virus glycoproteins; or V5 tags derived from small epitopes (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 to 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 that can bind to IgG. In some implementations, the matrix of the IgG Sepharose 6Fast 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 components are known and conceived by those skilled in the art, and are conceived here. Any peptide tag can be used, as long as it can be expressed as an element of an HLA-peptide complex with an affinity label.

[0430] 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 to a specific ligand. The fewer ligands a protein can bind to, 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 balance between the bound and unbound states of a protein-ligand system.

[0431] The term "affinity receptor-tagged HLA-peptide complex" refers to a complex containing an HLA class I or II associated peptide or a portion thereof that specifically binds to a monoallelic recombinant HLA class I or II peptide containing an affinity receptor peptide.

[0432] When used to describe the interaction between affinity molecules and affinity receptor tags or epitopes with HLA peptides, the term "specific binding" or "specific binding" means that the interaction depends on the presence of a specific structure on the protein (e.g., an antigenic determinant or epitope); in other words, the affinity molecule recognizes and binds to a specific affinity receptor peptide structure, rather than binding to the protein in general.

[0433] 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 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 binding energy. Affinity can be determined experimentally, for example, using commercially available Biacore SPR units via surface plasmon resonance (SPR). Affinity can also be expressed as the inhibition concentration 50 (IC50), i.e., the concentration at which 50% of the peptide is replaced. Similarly, lnIC50 refers to the natural logarithm of IC50. off It refers to the dissociation rate constant, for example, the rate constant of the dissociation of an affinity molecule from an affinity receptor-labeled HLA-peptide complex.

[0434] In some embodiments, the affinity receptor-tagged HLA-peptide complex contains a biotin receptor peptide (BAP) and is immunopurified from a complex cell mixture using streptavidin / NeutrAvidin beads. Biotin-avidin / streptavidin binding is the strongest known non-covalent interaction in nature. This property is used as a biological tool for a wide range of applications, such as the immunopurification of proteins covalently linked to biotin. In one exemplary embodiment, the nucleic acid sequence encoding an HLA allele uses the biotin receptor peptide (BAP) as an affinity receptor tag for immunopurification. The BAP can be specifically biotinylated at a single lysine residue within the tag, either in vivo or in vitro (e.g., U.S. Patents 5,723,584; 5,874,239; and 5,932,433; and British Patent GB2370039). BAPs are typically 15 amino acids long and contain one lysine as the biotin receptor residue. In some embodiments, the BAP is positioned at or near the N- or C-terminus of a monoallelic HLA peptide. In some embodiments, BAP is positioned between the heavy chain domain and the β2-microglobulin domain of an HLA class I peptide. In some embodiments, BAP is positioned between the β-chain domain and the α-chain domain of an HLA class II peptide. In some embodiments, BAP is positioned in a loop region between the α1, α2, and α3 domains of the HLA class I heavy chain, or in loop regions between the α1 and α2 and β1 and β2 domains of the α and β chains of HLA class II, respectively. Figure 2 The paper describes exemplary constructs designed for HLA class I and II expression, which use BAP for biotinylation and immunopurification.

[0435] As used herein, the term "biotin" refers to the compound biotin itself and its analogues, derivatives, and variants. Therefore, the term "biotin" includes biotin (cis-hexahydro-2-oxo-1H-thieno[3,4]imidazol-4-pentanoic acid) and any of its derivatives and analogues, including biotin-like compounds. Such compounds include, for example, amino or thiol derivatives of biotin-eN-lysine, biotinylate, 2-iminobiotin, and biotinyl-E-aminohexanoic acid-N-hydroxysuccinimide ester, sulfosuccinimide-iminobiotin, biotin bromoacetylhydrazine, p-diazobenzoyl biotinylate, 3-(N-maleimide propionyl) biotinylate, desulfurized biotin, etc. The term "biotin" also includes biotin variants that can specifically bind to one or more of Rhizavidin, avidin, streptavidin, tamavidin moiety, or other avidin-like peptides.

[0436] As used herein, “PPV determination method” may refer to a method for presenting a PPV determination. For example, “PPV determination method” may refer to a method comprising the following steps: (a) using an HLA peptide presentation prediction model, such as a machine learning HLA peptide presentation prediction model, to process the amino acid information of a plurality of test peptide sequences to generate a plurality of test presentation predictions, each test presentation prediction indicating the probability 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 cells) can present 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 presenting an HLA protein expressed in the cell, and (ii) at least 499 contained in an organism (e.g., belonging to the subject’s family) (a) Decoy peptide sequences within proteins encoded by the genome of an organism of the same species, wherein the ratio of the number of hit peptide sequences to the number of decoy 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 499 decoy peptide sequences is 1:499; (b) identifying or determining that the top percentage of the plurality of test peptide sequences, such as the top 0.2% of the plurality of test peptide sequences, is presented by a class II HLA allele of the cell; (c) calculating the PPV of the HLA peptide presentation prediction model, wherein the PPV is the fraction of test peptide sequences identified or determined to be presented by a class II HLA allele of the cell, which are peptides observed to be presented by a class II HLA allele of the cell by mass spectrometry. In some embodiments, the decoy peptides have the same length, i.e., containing the same number of amino acids as the hit peptides. In some embodiments, the decoy peptides may contain one more or one less amino acid than the hit peptides. In some embodiments, the decoy peptides are peptides that are endogenous peptides. In some embodiments, the decoy peptide is a synthetic peptide. In some embodiments, the decoy peptide is an endogenous peptide identified by mass spectrometry as binding to a first MHC class I or II protein, wherein the first MHC class I or II protein is different from a second MHC class I or II protein binding to the hit peptide. In some embodiments, the decoy peptide can be a scrambled peptide; for example, the decoy peptide can contain an amino acid sequence in which the amino acid positions are rearranged relative to the positions of the hit peptides along 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 decoy peptide sequences is approximately 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. 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 499 bait peptide sequences include 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, 6 700, 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, 3700038000, 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, 92 500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000, or 1,000000 bait peptide sequences. In some embodiments, the at least 500 test peptide sequences include 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, 66 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, 3500036000, 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 1,000000 test peptide sequences. In some implementations, identifying or determining the top percentages of the plurality of test peptide sequences as those presented by class II HLA alleles in cells includes placing 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%, 1.9% of the total number of alleles. 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% were identified or determined to be HLA class II alleles presented by the cell. In some embodiments, the cell is a monoallele cell.

[0437] As used herein, “PPV determination method” may refer to a method incorporating PPV determination. For example, “PPV determination method” may refer to a method comprising the following steps: (a) using an HLA peptide binding prediction model, such as a machine learning HLA peptide binding prediction model, to process the amino acid information of a plurality of test peptide sequences to generate a plurality of binding presentation predictions, each binding presentation prediction indicating the probability 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 presenting an HLA protein expressed in the cell, and (ii) at least 19 decoy peptide sequences contained within a protein, the protein The test peptide sequence comprises at least one peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the 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 that the top percentage of the plurality of test peptide sequences, such as the top 5% of the plurality of test peptide sequences, is bound to the HLA protein; and (c) calculating the PPV of the HLA peptide binding prediction model, wherein the PPV is the fraction of test peptide sequences identified or determined to bind to class II HLA alleles in the cell, which are peptides presented by class II HLA alleles in the cell as observed by mass spectrometry. In some embodiments, the ratio of the number of hit peptide sequences to the number of bait peptide sequences is approximately 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, 9899 or 100 hit peptide sequences. In some embodiments, the at least 19 bait 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, 1 400, 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, 7750080,000, 82,500, 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. In some embodiments, the at least 20 test peptide sequences include 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, and 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, 2800029000, 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 82,500, 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 test peptide sequences. In some embodiments, identifying or determining the top percentage of the plurality of test peptide sequences as being presented by class II HLA alleles in the cell includes identifying or determining 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 class II HLA alleles in the cell. In some embodiments, the cell is a monoallele cell.

[0438] Human Leukocyte Antigen (HLA) System

[0439] The immune system can be classified 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; most potential pathogens are rapidly neutralized by this system before causing, for example, an obvious infection. The adaptive immune system responds to molecular structures (called antigens) that invade an organism. Unlike the innate immune system, the adaptive immune system is highly specific to pathogens. Adaptive immunity can also provide lasting protection; for example, a person who has recovered from measles is now immune to measles for life. There are two types of adaptive immune responses: humoral immune responses and cell-mediated immune responses. In humoral immune responses, antibodies secreted into the body fluids by B cells bind to pathogen-derived antigens, leading to the elimination of the pathogen through various mechanisms, such as complement-mediated lysis. In cell-mediated immune responses, T cells capable of destroying other cells are activated. For example, if disease-related proteins are present in the cells, they are proteased and broken down into peptides within the cells. Specific cellular proteins then attach themselves to the antigens or peptides formed in this way and transport them to the cell surface, where they are presented to molecular defense mechanisms in the body's T cells. Cytotoxic T cells recognize these antigens and kill cells carrying those antigens.

[0440] The terms "major histocompatibility complex (MHC)," "MHC molecule," or "MHC protein" refer to proteins capable of binding to peptides generated by the proteolytic cleavage of protein antigens and representing potential T cell epitopes, transporting them to the cell surface, and presenting said peptides to specific cells, such as cytotoxic T lymphocytes or T helper cells. Human MHC is also known as the HLA complex. Therefore, the terms "human leukocyte antigen (HLA) system," "HLA molecule," or "HLA protein" refer to gene complexes encoding human MHC proteins. The term MHC is referred to as the "H-2" complex in rodent species. Those skilled in the art will recognize that the terms "major histocompatibility complex (MHC)," "MHC molecule," "MHC protein," and "human leukocyte antigen (HLA) system," "HLA molecule," and "HLA protein" are used interchangeably herein.

[0441] HLA proteins are classified into two types, known as HLA class I and HLA class II. The proteins of the two HLA types are structurally very similar; however, they have very different functions. HLA class I proteins are present on the surface of almost all cells in the body, including most tumor cells. HLA class I proteins are loaded with antigens, which typically 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 present on antigen-presenting cells (APCs), including but not limited to dendritic cells, B cells, and macrophages. They primarily present helper T cells with antigens from external sources, such as peptides processed extracellularly. 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.

[0442] 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 called β2-microglobulin (or β2m), (12 kD). The α chain has three extracellular domains—α1, α2, and α3—and a transmembrane region, where the α1 and α2 regions are 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 I molecules bind to peptides with suitable binding motifs and present them to cytotoxic T lymphocytes. The HLA class I heavy chain can be a protein product of the HLA-A allele, also known as the HLA-A monomer, or a protein product of the HLA-B allele (again, the HLA-B monomer), or a protein product of the HLA-C allele (the 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 connects to the transmembrane region, anchoring HLA class IA molecules to the cell membrane. The presented peptide is held at the bottom of the peptide-binding groove in the central region of the α1 / α2 heterodimer (a molecule composed of two distinct subunits). HLA classes IA, IB, and IC are highly polymorphic. The HLA class 1-A genes (called HLA-A genes), HLA class 1-B genes (called HLA-B genes), and HLA class 1-C genes (called HLA-C genes) 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 presented antigens. This group is further divided into those encoded within HLA loci, such as HLA-E, HLA-F, and HLA-G, and those not encoded within them, such as stress ligands like ULBP, Rae1, and H60. The antigen / ligand composition of many of these molecules remains unknown, but they can interact with CD8+ T cells, NKT cells, and NK cells.

[0443] In some embodiments, this disclosure utilizes non-classical HLA IE class 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, leading to T cell activation. HLA-E is also known to bind to the CD94 / NKG2 receptor expressed on NK cells and CD8+ T cells. CD94 can pair with several different isotypes of NKG2 to form receptors with the potential to inhibit (NKG2A, NKG2B) or promote (NKG2C) cell activation. HLA-E can bind peptides derived from amino acid residues 3-11 of the leader sequences of most HLA-A, -B, -C, and -G molecules, but cannot bind its own leader peptide. HLA-E has also been shown to present peptides derived from endogenous proteins similar to the HLA-A, -B, and -C alleles. Under physiological conditions, binding of CD94 / NKG2A to HLA-E loaded with a peptide derived from an HLA class I leader sequence typically induces repressive signaling. Cytomegalovirus (CMV) utilizes a mechanism to evade NK cell immune surveillance by expressing the UL40 glycoprotein (a mimicking the HLA-A leader sequence). However, it has also been reported that CD8+ T cells can recognize HLA-E loaded with the UL40 peptide derived from the CMV Toledo strain and play a role in CMV defense. Numerous studies have revealed several important functions of HLA-E in infectious diseases and cancer.

[0444] Before being presented on the cell surface, peptide antigens attach themselves to HLA-like molecules via competitive affinity binding within the endoplasmic reticulum. Here, the affinity of a single peptide antigen is directly related to its amino acid sequence and the presence of a specific binding motif at a defined position within that amino acid sequence. If the sequence of such a peptide is known, it can be used, for example, with peptide vaccines to manipulate the immune system to fight diseased cells.

[0445] MHC molecules are highly polymorphic, meaning that many MHC variants exist. Each variant is encoded by a variant of the gene that encodes the protein, 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 chain having a transmembrane domain: α2 and β2, anchoring HLA class II molecules to the cell membrane. The peptide-binding groove is formed by heterodimers of α1 and β1. The most widely studied HLA-DR molecule has DRA and DRB, corresponding to the α and β domains, respectively. DRB is diverse, while DRA is nearly identical. Therefore, the binding specificity of the DRB allele indicates the binding specificity of the corresponding HLA-DR. Each MHC protein has its own binding specificity, meaning that a group of peptides that binds to an MHC molecule may be different from the group that binds to another MHC molecule. Classical molecules present peptides to CD4+ lymphocytes. Non-classical molecules with intracellular functions, which are appendages and not exposed on the cell membrane, but rather in the inner membrane of the lysosome, typically load antigenic peptides onto classical HLA class II molecules.

[0446] In the HLA class II system, phagocytes such as macrophages and immature dendritic cells take up proteins via phagocytosis into phagosomes—although B cells exhibit more general endocytosis into endosomes—the phagosomes fuse with lysosomes, whose acidic enzymes cleave the taken-up proteins into a variety of different peptides. Autophagy is another source of HLA class II peptides. Through the physicochemical dynamics of molecular interactions with host-carried HLA class II variants (encoded in the host genome), specific peptides exhibit immunodominance and are loaded onto HLA class II molecules. They are transported to the cell surface and externalized there. The most studied subclasses of HLA class II genes are: HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA-DQB1, HLA-DRA, and HLA-DRB1.

[0447] HLA class II peptide presentation 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 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 that of HLA class II peptides (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 protruding from the binding groove on either the N-terminus or C-terminus (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 the P1, P4, P6 / 7 and P9 positions of the binding site (Rossjohn et al., 2015).

[0448] HLA alleles are expressed in a codominant manner, meaning that 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), thus allowing the expression of six different types of HLA class II. Within the HLA class II loci, 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 over 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 among different individuals within a population. Genes encoding HLA proteins have many possible variations, enabling each person's immune system to respond to a wide range of foreign invaders. Some HLA genes have hundreds of identified forms (alleles), each assigned a specific number. In some implementations, the HLA class I alleles are HLA-A*02:01, HLA-B*14:02, HLA-A*23:01, and HLA-E*01:01 (non-classical). In some implementations, 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.

[0449] Subject-specific HLA alleles or HLA genotypes of a subject 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 June 11, 2015 as WO2015085147), which is incorporated herein by reference in its entirety. In short, the method includes determining a polymorphic gene type, which may include generating an alignment of reads extracted from a sequencing dataset with a gene reference set containing 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 with 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 define the gene type of the polymorphic gene; and providing the outputs of the first and second allelic variants.

[0450] In some embodiments, the MHC class II peptide:antigen peptide binding and presentation prediction method described herein has the ability to predict binders from a large library of MHC class II peptides encoded by individual HLA alleles. In some embodiments, the MAPTAC technology is trained with a large database of mass spectrometry-validated HLA-matching peptides. In some embodiments, the large database of mass spectrometry-validated HLA-matching peptides contains more than 1.2 x 10^6 such HLA-matching peptides. In some embodiments, the large database of mass spectrometry-validated HLA-matching peptides covers more than 150 HLA alleles, including MHC class I and II allele subtypes. In some embodiments, the database covers at least 95% of the US population for HLA-I and HLA-II (DR subtypes).

[0451] As described in this article, there is substantial evidence in animals and humans that mutated epitopes can effectively induce immune responses, and that spontaneous tumor regression or long-term survival is associated with CD8+ T cell responses to mutated epitopes. Furthermore, “immune editing” can track changes in the expression of dominant mutant antigens in mice and humans.

[0452] Sequencing technology has revealed that each tumor contains multiple patient-specific mutations that alter the protein-coding content of genes. These 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, readthroughs of stop codons, or translation of intron regions (new open reading frame mutations; neoORF). These mutant proteins are valuable targets for the host's immune response to tumors because, unlike natural proteins, they are not affected by the immunosuppressive effects of self-tolerance. Therefore, mutant proteins are more likely to be immunogenic and more specific to tumor cells compared to the patient's normal cells. Essentially, short peptides (8-24 amino acids long) containing cancer-associated mutations are candidates for cancer immunotherapy.

[0453] In some embodiments, the algorithm of the driver prediction method can be further used to determine peptide mutations. In some embodiments, the prediction method can be used to determine driver mutation status, and / or RNA expression status, and / or intrapeptide cleavage prediction.

[0454] The term "T cell" includes CD4+ T cells and CD8+ T cells. It also includes helper T cells type 1 and helper T cells type 2. In this article, T cells are generally classified into two main categories based on function and the cell surface antigens (cluster differentiation antigens or CDs) that also contribute to the binding of T cell receptors to antigens: helper T (TH) cells and cytotoxic T lymphocytes (CTLs).

[0455] Mature helper T (TH) cells express the surface protein CD4 and are called CD4+ T cells. After T cell development, mature naive T cells leave the thymus and begin to spread throughout the body, including the lymph nodes. Naive 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 which antigens the T cell can respond to. CD4+ T cells have a TCR with affinity for MHC class II proteins, and CD4 is involved in determining MHC affinity during thymic maturation. MHC class II proteins are typically only found on the surface of specialized antigen-presenting cells (APCs). Specialized antigen-presenting cells (APCs) are primarily dendritic cells, macrophages, and B cells, although dendritic cells are the only group of cells constitutively (always) expressing MHC class II proteins. Some APCs also bind natural (or unprocessed) antigens to their surface, such as follicular dendritic cells; however, unprocessed antigens do not interact with T cells or participate in their activation. Peptide antigens that bind to HLA class I proteins are generally shorter than those that bind to HLA class II proteins.

[0456] Cytotoxic T lymphocytes (CTLs), also known as cytotoxic T cells, cytolytic T cells, CD8+ T cells, or killer T cells, are lymphocytes that induce apoptosis in targeted cells. CTLs interact with antigens (Ag) processed on the surface of target cells via their TCRs, forming antigen-specific conjugates that lead to apoptosis of the target cells. The apoptotic bodies are then eliminated by macrophages. The term "CTL response" refers to a primary immune response mediated by CTL cells. Cytotoxic T lymphocytes possess both a T cell receptor (TCR) and a CD8 molecule on their surface. The T cell receptor recognizes and binds to peptides complexed with HLA class I molecules. Each cytotoxic T lymphocyte expresses a unique T cell receptor that binds to a specific MHC / peptide complex. Most cytotoxic T cells express a T cell receptor (TCR) that recognizes a specific antigen. For the TCR to bind to HLA class I molecules, it must be accompanied by a glycoprotein called CD8, which binds to a constant portion of the HLA class I molecule. Therefore, these T cells are called CD8+ T cells. The affinity between CD8 and MHC molecules allows T cells and target cells to bind tightly 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.

[0457] The T-cell receptor (TCR) is a cell surface receptor involved in the activation of T cells in response to antigen presentation. The TCR typically consists of two chains, α and β, which assemble to form a heterodimer and associate with the CD3 transduction subunit to form the T-cell receptor complex present on the cell surface. Each α and β chain of the TCR comprises 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 the α and β chains are generated through V(D)J recombination, resulting in a high degree of antigen specificity across the T cell population. However, compared to immunoglobulins that recognize intact antigens, T cells are activated by processed peptide fragments associated with MHC molecules, introducing an additional dimension to T cell antigen recognition known as MHC restriction. MHC differences between donor and recipient T-cell receptor recognition 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 allogeneic recognition and thus preventing GVHD. However, TCR disruption generally leads to the elimination of CD3 signaling components and alters the manner of further T cell expansion.

[0458] The term "HLA peptidomimetry" refers to a group of peptides that specifically interact with a particular HLA class and can contain thousands of different sequences. The HLA peptidomimetry includes a diverse range of peptides derived from both normal and abnormal proteins expressed in cells. Therefore, the HLA peptidomimetry can be studied to identify cancer-specific peptides for the development of tumor immunotherapies and as a source of information about protein synthesis and degradation protocols within cancer cells. In some embodiments, the HLA peptidomimetry is a group of soluble HLA peptides (sHLA). In some embodiments, the HLA peptidomimetry is a group of membrane-bound HLAs (mHLA).

[0459] "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 β 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.

[0460] Monoallelic HLA cell lines

[0461] Monoalleic 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 polynucleotide encoding a single HLA allele (e.g., a vector). Figure 2Suitable 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 II deficient cell lines in which a single HLA class I and / or a pair of class II alleles is exogenously expressed. As an exemplary embodiment, the HLA class I deficient B cell line is B721.221. However, it will be clear to those skilled in the art that other cell populations that are HLA class I and / or HLA class II deficient 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 professional antigen-presenting cells, such as macrophages, B cells, and dendritic cells. The cells may be B cells or dendritic cells. In some embodiments, the cells are tumor cells or cells derived from tumor cell lines. In some embodiments, the cells are 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 Individual 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, this gene encodes a member of the immunoproteasome. Immunoplasmosomes are known to be involved in the processing of HLA class-binding peptides and include LMP2 (β1i), MECL-1 (β2i), and LMP7 (β5i) subunits. Immunoplasmosomes 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 disruptors, 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.

[0462] Virus-based systems (e.g., adenovirus systems, adeno-associated virus (AAV) vectors, poxviruses, or lentiviruses) can be used to encode and express the single HLA alleles of this disclosure. Plasmids that can be used for delivery with adenovirus, adenovirus, and lentivirus have been previously described (see, for example, 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 this disclosure, integration into the host genome of a cell can be achieved using retroviral gene transfer methods, typically resulting in long-term expression of the inserted transgene. In one exemplary embodiment, the retrovirus is a lentivirus. Furthermore, high transduction efficiency has been observed in many different cell types and target tissues. The tropism of a retrovirus can be altered by incorporating foreign envelope proteins, expanding the potential target population of target cells. Retroviruses can also be engineered to allow conditional expression of the inserted transgene, such 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 (therefore, both lentiviruses and retroviral vectors can be used in the practice of this disclosure). Furthermore, lentiviral vectors are capable of transducing or infecting non-dividing cells and typically produce high viral titers.

[0463] The choice of retroviral gene transfer system can depend on the target tissue. Retroviral vectors consist of cis-acting long terminal repeats (LTRs) with a packaging capacity of 6-10 kb of foreign sequences. Minimal cis-acting LTRs are sufficient to replicate and package the vector, which is then used to integrate the desired nucleic acid into target cells for permanent expression. Widely used retroviral vectors that can be used in the practice of this disclosure include vectors based on murine leukemia virus (MuLV), gibberish leukemia virus (GaLV), simmon immunodeficiency virus (SIV), human immunodeficiency virus (HIV), and combinations thereof (see, for example, 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 equine infectious anemia virus (EIAV)-based lentiviral vectors, are useful in the practice of this disclosure (see, e.g., Balagaan, (2006) J Gene Med; 8:275-285, published online on 21 November 2005 in Wiley InterScience DOI:10.1002 / jgm.845). These vectors may have a cytomegalovirus (CMV) promoter that drives the expression of the target gene. Therefore, this disclosure relates to one or more vectors that can be used to implement the contents of this disclosure: viral vectors, including retroviral vectors and lentiviral vectors.

[0464] Any HLA allele can be expressed in a cell population. In one 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. The sequences of HLA class I and II alleles are available 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.

[0465] In some embodiments, an HLA allele is selected to correspond to a target genotype. In some embodiments, the HLA allele is a mutated HLA allele, which may be an allele not naturally present in diseased patients or a naturally present 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 present at low frequencies. Therefore, in some embodiments, the methods provided herein can identify HLA alleles even if they exist at a frequency of less than 1% in populations such as Caucasian populations.

[0466] In some embodiments, the nucleic acid sequence encoding the HLA allele further includes an affinity receptor tag that can be used for immunopurification of 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 tag, a c-myc tag, a herpes simplex virus glycoprotein D (gD) tag, a FLAG tag, a KT3 epitope tag, a tubulin 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 bluetongue virus tag (B-tag), a calmodulin-binding peptide (CBP) tag, a chloramphenicol acetyltransferase (CAT) tag, or a choline tag. This list includes tags for various gene sequences, including: Chitosan-binding domain (CBD), Cellulose-binding domain (CBP), Dihydrofolate reductase (DHFR), Galactose-binding protein (GBP), 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, and 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 vesicular stomatitis virus glycoprotein, or V5 tag derived from small epitopes (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 to the corresponding antibody, thereby allowing identification or affinity purification of the labeled protein. Non-limiting examples of epitope tags are protein A or protein G, which can bind to IgG. In some embodiments, affinity receptor tags include biotin receptor peptide (BAP) or human influenza hemagglutinin (HA) peptide sequences.Many other tag components are known and conceived by those skilled in the art, and are conceived here. Any peptide tag can be used, as long as it can be expressed as an element of an HLA-peptide complex with an affinity label.

[0467] The methods described in this article include isolating HLA-peptide complexes from cells affinity-pull-transfected or transduced with HLA constructs. Figure 3 In some embodiments, the complex can be separated using standard immunoprecipitation techniques known in the art and commercially available antibodies. Cells may be lysed first. HLA class I-peptide complexes can be separated using HLA class I specific antibodies such as W6 / 32 antibodies, while HLA class II-peptide complexes can be separated using HLA class II specific antibodies such as M5 / 114.15.2 monoclonal antibodies. In some embodiments, a single (or a pair) HLA allele is expressed as a peptide-tagged fusion protein, and the HLA-peptide complex is separated using a binding molecule that recognizes the peptide tag.

[0468] The method further includes 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, such as liquid chromatography-mass spectrometry (LC-MS or LC-MS / MS, or HPLC-MS or HPLC-MS / MS), is employed. These sequencing methods are well known to those skilled in the art and have been reviewed in Medzihradszky KF and Chalkley RJ. Mass Spectrom Rev. Jan-Feb 2015; 34(1):43-63.

[0469] 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 both endogenous HLA class I 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-sorted by FACS for cell surface expression of HLA class I or II, or both HLA class I and II. For example, FACS can be used to sort cell populations based on cell surface expression of HLA class I alleles, HLA class II alleles, or combinations thereof.

[0470] Methods for preparing personalized cancer vaccines

[0471] Once a cancer-specific mutation is identified—one that exists in the DNA of cancer cells but not in normal cells of the same human subject, and that causes a change in one or more amino acids in the protein encoded by that DNA—it can become a target of the host's immune response. The innate immune response can target the mutated protein, leading to the destruction of cancer cells expressing that protein. Because of the natural tolerance response and immunodeficiency environment present in cancerous tissue, immunotherapy is a clinical pathway that attempts to enhance this immune response beyond the body's tolerance and immunosuppression. Therefore, proteins or peptides containing the aforementioned mutations are suitable candidates for immunotherapy.

[0472] Mutated proteins are taken up, cleaved, and displayed as antigens on the cell surface by specialized phagocytes acting as antigen-presenting cells (APCs), for activation by T cells in antigen-presenting complexes containing major histocompatibility complex (MHC) proteins. Human MHC proteins are referred to as human leukocyte antigens, HLA. MHC proteins can be MHC class I or class II proteins. Some functional differences are attributed to the presentation of peptides by class I or class II MHC proteins (HLA class I and HLA class II proteins). A significant difference 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 in the task of cell-by-cell clearance of pathogenic cells (e.g., infected or tumor cells). CD4+ T cells have a more durable effect after activation, most importantly in the generation of immune memory. A subset of CD4 cells is recruited differently depending on the type of immune threat, and multiple subsets with overlapping or different functions can be recruited together. This helps to balance the immune response associated with pathogen threats. In these respects, HLA class II peptide-mediated antigen presentation enables sustained and customized immune responses. On the other hand, the binding of HLA class II to peptides can be promiscuous, and therefore nonspecific peptide binding and presentation to the immune system can lead to aberrant immune responses, such as autoimmunity.

[0473] In one aspect, this disclosure provides a method for predicting peptides that can accurately pair or bind to specific HLA class II α and β heterodimers, such that the high-fidelity binding of the peptide to HLA class II proteins (including α and β heterodimers) ensures the presentation of the specific peptide to T lymphocytes, thereby triggering a specific immune response and avoiding any cross-reactivity or immune confounding.

[0474] In one aspect, this disclosure provides a method for predicting peptides that can accurately bind to specific HLA class II proteins, such that when the peptide is therapeutically administered to a subject expressing a specific homologous HLA class II protein, the peptide can activate a more sustained and robust immune response by leveraging the ability of HLA class II proteins to activate CD4+ T cells and stimulate immune memory. In some embodiments, the given peptide predicted to bind with high specificity to HLA class II proteins is a peptide containing a mutation, wherein the mutation is prevalent in the subject's cancer or tumor cells; while the same HLA class II protein predicted to bind to the mutated peptide either (a) does not bind or (b) has a significantly lower affinity for binding to the corresponding non-mutated wild-type peptide than for binding to the subject's mutated peptide. Preferential binding of HLA to mutated peptides is advantageous in the development of immunotherapeutic agents because cells expressing wild-type peptides will be protected from immune attack by T cells that are reactive to HLA-presented peptides. In some embodiments, the predicted peptide that specifically binds to HLA class II proteins is a peptide with post-translational modifications. Exemplary post-translational modifications include, but are not limited to, phosphorylation, ubiquitination, dephosphorylation, glycosylation, methylation, or acetylation. In some embodiments, the predicted peptide undergoes post-translational modification prior to its use in immunotherapy.

[0475] In some implementations, the immunotherapeutic methods and strategies disclosed herein can also be applied to suppress unwanted immune activation, such as in autoimmune responses. Specifically, peptides identified as potential binders to specific HLA subtypes can be tailored to bind to specific HLA molecules and induce tolerance rather than triggering an immunogenic response.

[0476] On one hand, this article proposes immunotherapies tailored or personalized for specific subjects. 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 for the identification of a 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 with high fidelity which peptides can bind to specific HLA class II α and β chain heterodimers can ensure the generation of a specific immune response tailored specifically for that subject.

[0477] The genes encoding HLA heterodimers are highly polymorphic, with over 4,000 HLA class II allelic variants identified across the human population. For each of the HLA class II loci, individuals can inherit different alleles from their maternal and paternal HLA haplotypes, and each HLA class II heterodimer consists of an α-chain and a β-chain. Due to the large number of α-chain and β-chain pairings, particularly for the HLA-DP and HLA-DQ alleles, the population of possible HLA heterodimers is highly complex. HLA class II heterodimers are translated in the endoplasmic reticulum (ER) and assemble with an invariant chain (Ii) derived from the protein CD74 to form a stable complex. Ii stabilizes the class II complex by allowing proper protein folding and enables the HLA class II heterodimer to be exported to endosome / lysosomal compartments. Within these HLA class II loading compartments, Ii is proteasically cleaved by cathepsins into a placebo peptide called CLIP. CLIP is then exchanged for a higher-affinity peptide via 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 reaches the trans-Golgi apparatus and finally the cell surface for display by CD4+ T cells.

[0478] It is estimated that each HLA heterodimer binds to thousands of peptides with allele-specific binding preferences. In fact, it is estimated that each HLA allele binds to and presents approximately 1,000–10,000 unique peptides to T cells. Given this diversity in HLA binding, accurately predicting whether a peptide is likely to bind to a specific HLA allele is extremely challenging. Little is known about the allele-specific peptide binding characteristics of HLA class II molecules due to the heterogeneity of α and β chain pairing, the complexity of the data limiting the ability to confidently identify core binding epitopes, and the lack of allele-specific antibodies at the immunoprecipitation level required for high-resolution biochemical analysis. Furthermore, analyzing peptide epitopes derived from a given HLA allele introduces uncertainty when multiple HLA alleles are presented on the cell surface.

[0479] The prediction of candidate neoantigens primarily targets HLA class I epitopes (given the availability of experimental data for class I prediction algorithms compared to class II), but CD4+ T cell responses are 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 crucial role in cancer therapy. Although HLA class II prediction algorithms exist, they are inaccurate because the open peptide-binding groove on HLA class II heterodimers allows for the binding of longer peptides (typically 15–40 amino acids), increasing the heterogeneity and complexity of epitope presentation. Therefore, further work is needed to better understand the characteristics 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 the isolation of HLA class II-peptide complexes. To overcome these challenges, a monoallelic HLA profiling workflow was developed, relying on LC-MS / MS to characterize allele-specific HLA class II-ligandomes 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 circumstances, such as any co-owned patents or applications. Exemplary materials and methods are described herein; however, any methods and materials similar to or equivalent to those described herein may be used in practice to test this disclosure. Therefore, the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0480] This article discloses a method for preparing a personalized cancer vaccine. The method may include identifying mutated peptide sequences expressed in a subject's cancer cells; using a computer processor to input the amino acid position information of the identified peptide sequences into a machine learning HLA-peptide presentation prediction model to generate a set of presentation predictions for the identified peptide sequences, each presentation prediction representing the probability that one or more proteins encoded by class II MHC alleles of the subject's cancer cells will present a given sequence of the identified peptide sequence; and selecting a subset of peptide sequences identified based on this set of presentation predictions for use in preparing the personalized cancer vaccine.

[0481] In some embodiments, one or more results obtained from the methods described herein can provide one or more quantitative values ​​indicative of one or more of the following: the probability of diagnostic accuracy, the probability of a condition being present in the subject, the probability of a subject developing a condition, the probability of success of a particular treatment, or any combination thereof. In some embodiments, the methods described herein can predict the risk or probability of a condition occurring. In some embodiments, the methods described herein can be an early diagnostic indication of a condition. In some embodiments, the methods described herein can confirm a 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 treatment on a subject's condition.

[0482] Methods for identifying MHC-II peptides

[0483] In one aspect, this document provides a method for identifying one or more peptides presented by MHC-II proteins for immune activation. In some embodiments, the one or more peptides comprise epitopes. In some embodiments, the method involves computationally predicting the likelihood of a particular epitope being presented by an MHC-II protein. In some embodiments, the method involves computationally predicting the specificity of the epitope for MHC-II presentation. In some embodiments, the computational prediction method involves assessing peptide-MHC interactions. In some embodiments, the computational prediction method involves predicting the allelic specificity of the peptide for antigen presentation.

[0484] In some implementations, computational prediction methods involve the integration of bioinformatics information, such as nucleotide sequences, structural motifs of biomolecules, protein-protein interaction characteristics, and functional efficiencies such as immunogenicity. In other implementations, computational prediction methods involve machine learning. Based on machine learning methods, such as simple pattern motifs, support vector machines (SVMs), hidden Markov models (HMMs), neural network (NN) models, quantitative structure-activity relationship (QSAR) analysis, structure-based methods, and biophysical methods, numerous immunoinformatics methods for predicting peptide-MHC interactions have been developed for MHC classes I and II. These methods can be categorized into two types: intra-allelic (allele-specific) and trans-allelic (pan-specific) methods. Intra-allelic methods are trained on a limited set of experimental peptide binding data for a specific MHC molecule and applied to predict peptides binding to that molecule. Due to the extreme polymorphism of MHC molecules, the existence of thousands of allele variants, and the lack of sufficient experimental binding data, it is impossible to build a prediction model for each allele. Therefore, cross-allele and universal methods have been developed using peptide binding data across many alleles or extended across species, 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). Similar methods for MHC-I are also available, such as NetMHCpan and KISS.

[0485] In some implementations, the peptide sequence may not be expressed in the subject's normal cells. In some implementations, each and every cell of the subject may not be a cancer cell. The cancer cells can arise from various cancers, including but not limited to thyroid cancer, adrenocortical cancer, anal cancer, aplastic anemia, bile duct cancer, bladder cancer, bone cancer, bone metastases, central nervous system (CNS) cancer, peripheral nervous system (PNS) cancer, breast cancer, Castleman's disease, cervical cancer, childhood non-Hodgkin's lymphoma, lymphoma, colorectal cancer, endometrial cancer, esophageal cancer, Ewing tumor family (e.g., Ewing 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 lymphoblastic leukemia, acute myeloid leukemia. Leukemia, childhood leukemia, chronic lymphocytic leukemia, chronic myeloid leukemia, liver cancer, lung cancer, lung carcinoid tumor, non-Hodgkin's lymphoma, male breast cancer, malignant mesothelioma, multiple myeloma, spinal dysplasia syndrome, myeloproliferative disorders, nasal cavity and paranasal carcinoma, nasopharyngeal carcinoma, neuroblastoma, oral and oropharyngeal carcinoma, osteosarcoma, ovarian cancer, pancreatic cancer, penile cancer, pituitary adenoma, prostate cancer, retinoblastoma, rhabdomyosarcoma, salivary gland cancer, sarcoma (adult soft tissue cancer), melanoma skin cancer, non-melanoma skin cancer, gastric cancer, testicular cancer, thymic cancer, uterine cancer (e.g., uterine sarcoma), vaginal cancer, vulvar cancer, or Waldenström macroglobulinemia.

[0486] The identification may include comparing DNA, RNA, or protein sequences from the subject's cancer cells with DNA, RNA, or protein sequences from the subject's normal cells. The DNA, RNA, or protein sequences from the subject's cancer cells may differ from those from the DNA, RNA, or protein sequences from the subject's normal cells. The identification can detect nucleic acid variants with high sensitivity.

[0487] The machine learning HLA-peptide presentation prediction model may include at least a number of predictive variables identified based 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 containing amino acid position information, wherein the training peptide sequence information is associated with HLA proteins expressed in cells; and a function representing the relationship between the amino acid position information received as input and the presentation probability generated as output based on the amino acid position information and the predictive variables.

[0488] In some implementations, 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, used to accurately simulate or train 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, web-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.

[0489] In some implementations, training data may be stored in a database. The database may be stored in a computer-readable format. A computer processor may be configured to access the data stored in the computer-readable storage. In some implementations, the computer system may be used to analyze the data to obtain results. These results may be stored remotely or internally on a storage medium and transmitted to personnel such as drug specialists. In some implementations, the computer system may be operatively coupled to components for transmitting results. Components for transmission may include wired and wireless components. Examples of wired communication components may include a Universal Serial Bus (USB) connection, a coaxial cable connection, an Ethernet cable such as Cat5 or Cat6 cable, a fiber optic cable, or a telephone line. Examples of wireless communication components may include a Wi-Fi receiver, components for accessing mobile data standards such as 3G or 4G LTE data signals, or a Bluetooth receiver. In some implementations, all this data in the storage medium is collected and archived to build a data warehouse.

[0490] In some implementations, the database includes external databases. External databases can be medical databases, such as, but not limited to, adverse drug reaction databases, AHFS supplemental documents, allergen selection list files, average WAC pricing files, brand probability files, Canada Drug Files v2, comprehensive price history, controlled substance files, drug allergy cross-reference files, drug application documents, drug administration and delivery databases, drug image database v2.0 / drug imprinting database v2.0, drug expiration date files, drug indication databases, drug laboratory conflict databases, Drug Therapeutic Monitoring System (DTMS) v2.2 / DTMS consumer monographs, repeat treatment databases, federal government pricing documents, Healthcare Common Procedure Coding System (HCPCS) databases, ICD-10 mapping files, immune cross-reference files, comprehensive A-to-Z drug fact module, comprehensive patient education, master parameter databases, half-span electronic drug file (MED-File) v2, Medicaid... Rebate files, healthcare plan files, condition selection list files, condition master database, drug order management database (MOMD), monitoring parameter database, patient safety plan files, payment limit-part B (PAF-B) v2.0, precautionary measures database, RxNorm cross-reference files, standard drug identifier database, replacement group files, supplementary name files, uniform classification cross-reference files, or warning label database.

[0491] In some implementations, training data may also be obtained from other data sources. Data sources may include sensors or smart devices such as appliances, smart meters, wearable devices, monitoring systems, data storage, customer systems, billing systems, financial systems, population source data, weather data, social networks, or any other sensor, enterprise system, or data storage. Examples of smart meters or sensors may include meters or sensors located at the customer's location, or meters or sensors located between the customer and the generating or source location. By integrating data from a wide range of sources, the system may be able to perform complex and detailed analyses. In some implementations, data sources may include, but are not limited to, sensors or databases used for other medical platforms.

[0492] HLA typing is routinely performed using antibody-based serological methods or PCR-based methods such as sequence-specific oligonucleotide probe hybridization (SSOP) or sequence-based typing (SBT). The first is hampered by potentially high cross-reactivity and limited resolution, while the second presents difficulties related to PCR efficiency due to the very limited availability of positioning primers resulting from polymorphic locations.

[0493] In some embodiments, sequence information is identified by sequencing methods or by mass spectrometry methods, 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 have been reviewed in Medzihradszky KF and Chalkley RJ. Mass Spectrom Rev. Jan-Feb 2015; 34(1):43-63. In some embodiments, the mass spectrometry is monoallelic mass spectrometry. In some embodiments, the mass spectrometry can be MS analysis, MS / MS analysis, LC-MS / MS analysis, or a combination thereof. In some embodiments, MS analysis can be used to determine the quality of the intact peptide. For example, the determination may include determining the quality of the intact peptide (e.g., MS analysis). In some embodiments, MS / MS analysis can be used to determine the quality of a peptide fragment. For example, the determination may include determining the quality of a peptide fragment, which can be used to determine the amino acid sequence of the peptide or a portion thereof (e.g., MS / MS analysis). In some embodiments, the quality of the peptide fragment can be used to determine the amino acid sequence within the peptide. In some implementations, LC-MS / MS analysis can be used to separate complex peptide mixtures. For example, this determination may include, for instance, separating complex peptide mixtures by liquid chromatography, and determining the mass of intact peptides, the mass of peptide fragments, or combinations thereof (e.g., LC-MS / MS analysis). This data can be used, for example, for peptide sequencing.

[0494] In some embodiments, the training peptide sequence information includes amino acid position information of the training peptide. In some embodiments, the training peptide sequence information includes up to about 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20%, 10%, or less of the sequence information of a peptide presented by an HLA protein expressed in cells and identified by mass spectrometry. In some embodiments, the training peptide sequence information may include at least about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, or more of the sequence information of a peptide presented by an HLA protein expressed in cells and identified by mass spectrometry.

[0495] Any information and data can be paired with the subject who is the source of that information and data. Subjects or medical professionals can retrieve information and data from storage or servers using the subject's identity. Subject identity can include a patient's photograph, name, address, Social Security number, date of birth, telephone number, postal code, or any combination thereof. Subject identity can be encrypted and encoded as 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, MSIPlessey barcode, QR barcode, Datamatrix code, PDF417 code, or Aztec barcode. The visual graphic code can be configured to be displayed on a screen. The barcode can include a QR code that can be captured and read by machine optics. Elements such as the barcode version, format, position, alignment, or timing can be defined to enable barcode 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 they can be scanned from a reasonable distance by an imaging device. QR codes can be in any image file format (e.g., EPS or SVG vector graphics, PNG, TIF, GIF, or JPEG raster graphics formats).

[0496] In some implementations, the function representing the relationship between the amino acid position information received as input and the presentation probability generated as output based on that amino acid position information and a predictor variable includes a linear or nonlinear function. This function can be, for example, a modified linear unit (ReLU) activation function, a Leaky ReLU activation function, or other functions such as saturated hyperbolic tangent, identity, binary step function, logic function, arcTan, softsign, parametric modified linear unit, exponential linear unit, softPlus, bent identity, softExponential, sine curve (Sinusoid), Sinc, Gaussian, or sigmoid function, or any combination thereof.

[0497] In some implementations, linear functions are obtained through linear regression. In some implementations, linear regression is a method of predicting a target variable by fitting the best linear relationship between the dependent and independent variables. A best fit might mean minimizing the sum of all distances between the shape at each point and the actual observed value. Linear regression can include simple linear regression or multiple linear regression. Simple linear regression can predict the dependent variable using a single independent variable. Multiple linear regression can predict the dependent variable by fitting the best linear relationship using more than one independent variable. Nonlinear functions can be obtained through nonlinear regression. Nonlinear regression can be a form of regression analysis where observed data are 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.

[0498] In some embodiments, the likelihood of presentation is represented by a one-dimensional value (e.g., probability). In some embodiments, the probability is configured to measure the likelihood that an event may occur. In some embodiments, the probability ranges from 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, the event includes any type of situation, as non-limiting examples, including whether an HLA-peptide will present certain peptides with specific amino acid position information, and whether a person will become ill based on amino acid position information. In some embodiments, the likelihood may be represented by a multidimensional value. Multidimensional values ​​can be represented by a multidimensional space, a heatmap, or a spreadsheet.

[0499] In one embodiment, a subset of peptide sequences based on a set of presentation predictions is selected and configured 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 the peptide sequences identified based on that 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 the peptide sequences identified based on that set of presentation predictions. The cancer vaccine may be a vaccine for treating existing cancer or for preventing the development of cancer. The vaccine may be prepared from a sample taken from a patient and may be specific to that patient.

[0500] In some embodiments, poxviruses are used in disease vaccines or immunogenic compositions. These include orthopoxvirus, avipox, cowpox, MVA, NYVAC, canarypox, ALVAC, fowlpox, TROVAC, etc. Advantages of this vector may include simple construction, the ability to adapt to large amounts of foreign DNA, and high expression levels. Regarding poxviruses that can be used to implement this disclosure, such as Chordopoxvirinae poxviruses (vertebrate poxviruses), for example, orthopoxvirus and avianpoxvirus, such as cowpoxvirus (e.g., Wyeth strain, WR strain (e.g., Information on recombinants of 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, especially their synthetic or non-naturally occurring recombinants, their uses, and methods for preparing and using such recombinants can be found in scientific and patent literature.

[0501] In some implementations, vaccinia virus is used in disease vaccines or immunogenic compositions to express antigens. Recombinant vaccinia virus can replicate within the cytoplasm of infected host cells, thus the target peptide can induce an immune response.

[0502] In some implementations, ALVAC is used as a carrier 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 method of inoculation against prokaryotic and eukaryotic antigens.

[0503] In some implementations, modified Ankara vaccinia virus (MVA) is used as a viral vector for antigenic vaccines or immunogenic compositions. MVA can be a member of the Orthopodoviridae family and has been produced through approximately 570 consecutive 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., reduced virulence or infectivity, while still retaining excellent immunogenicity. When tested in various animal models, MVA has proven to be non-toxic, even in immunosuppressed individuals. Furthermore, MVA- -HER2 is a candidate immunotherapy for the treatment of HER-2 positive breast cancer and is currently undergoing clinical trials.

[0504] In some implementations, the Positive Predictive Value (PPV) is used as part of the predictive model. PPV, also known as a measure of accuracy, is the probability that an individual diagnosed with a disease or condition by, for example, a test or model, actually has that disease or condition. It can be calculated by dividing the number of true positive results by the total number of positive results returned (including false positives). PPV = True Positives / (True Positives + False Positives). For example, if in a cohort of 100 patients, the model identifies positive results in 50 patients, of which 25 are true positives, then the PPV would be 25 / 50 = 0.5. The closer the PPV is to 1, the more accurate the diagnostic method, such as the test or model, is. PPV can be used to determine the accuracy of a predictive model. PPV can be used to adjust a predictive model to accommodate false positive results that the model may generate.

[0505] 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 out of the total number of positive results in the sample set. Decision rate = True positives / (True positives + False negatives). For example, if in a group of 100 patients, the model identifies positive results in 50 patients, of which 25 are true positives, and the total number of positive results in the group is 75, then the decision rate is {25 / (25+25)} x 100 = 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 accommodate false positives or false negatives that the model might produce.

[0506] In some embodiments, the predictive 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 higher at a decision rate of 0.1%-10%. In some embodiments, the predictive model may 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 lower at a decision rate of 0.1%-10%. The predictive model may 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 higher at a decision rate below 0.1%. In some embodiments, the predictive model may have a positive predictive value of up to 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1, or lower at a decision rate below 0.1%. The predictive model may 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 higher at a decision rate above 10%. In some embodiments, the predictive model may have a positive predictive value of up to 0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2, 0.1, or lower at a decision rate above 10%.

[0507] In some embodiments, the predictive 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 higher at a decision rate of 0.1% to 10%. In some embodiments, the predictive model has a positive predictive value of 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%, 0.5% to 3%, 0.5% to 1%, 0.5% to 2 ... 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 9%, 5% to 10% 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, or higher 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, or 0.9 or higher at a decision rate of 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, or 10%. In some embodiments, the predictive 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, or 0.9 or higher at a decision rate of at least 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, or 10%. In some embodiments, the predictive 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, or 0.9 or higher at a decision rate of at most 0.5%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, or 10%.

[0508] In some embodiments, the predictive 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 higher at a decision rate of 10% to 20%. In some embodiments, the predictive model has a positive predictive value of 10% to 11%, 10% to 12%, 10% to 13%, 10% to 14%, 10% to 15%, 10% to 16%, 10% to 17%, 10% to 18%, 10% to 19%, 10% to 20%, 11% to 12%, 11% to 13%, 11% to 14%, 11% to 15%, 1 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% Up 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 The predictive 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, or higher at a decision rate of 8%, 16% to 19%, 16% to 20%, 17% to 18%, 17% to 20%, 18% to 20%, or 19% to 20%. In some embodiments, the predictive 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, or 0.9 or higher at a decision rate of 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, or 20%. In some embodiments, the predictive 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, or 0.9 or higher at a decision rate of at least 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, or 20%. In some embodiments, the predictive 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, or 0.9 or higher at a decision rate of at most 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, or 20%.

[0509] In some implementations, the predictive 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, or higher 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 predictive model may have a positive predictive value of at least 0.1 at a decision rate of at least 10%. For example, the predictive model may have a positive predictive value of at least 0.2 at a decision rate of at least 10%. For example, the predictive model may have a positive predictive value of at least 0.3 at a decision rate of at least 10%. For example, the predictive model may have a positive predictive value of at least 0.4 at a decision rate of at least 10%. For example, a predictive model can have a positive predictive value of at least 0.5 with a decision rate of at least 10%. For example, a predictive model can have a positive predictive value of at least 0.6 with a decision rate of at least 10%. For example, a predictive model can have a positive predictive value of at least 0.7 with a decision rate of at least 10%. For example, a predictive model can have a positive predictive value of at least 0.8 with a decision rate of at least 10%. For example, a predictive model can have a positive predictive value of at least 0.9 with a decision rate of at least 10%. For example, a predictive model can have a positive predictive value of at least 0.1 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.2 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.3 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.4 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.5 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.6 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.7 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.8 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.9 with a decision rate of at least 5%. For example, a predictive model can have a positive predictive value of at least 0.1 with a decision rate of at least 20%. For example, a predictive model can have a positive predictive value of at least 0.2 with a decision rate of at least 20%. For example, a predictive model can have a positive predictive value of at least 0.3 with a decision rate of at least 20%. For example, a predictive model can have a positive predictive value of at least 0.4 with a decision rate of at least 20%. For example, a predictive model can have a positive predictive value of at least 0.5 with a decision rate of at least 20%. For example, a predictive model can have a positive predictive value of at least 0.6 with a decision rate of at least 20%.For example, a predictive model can have a positive predictive value of at least 0.7 with a decision rate of at least 20%. For example, a predictive model can have a positive predictive value of at least 0.8 with a decision rate of at least 20%. For example, a predictive model can have a positive predictive value of at least 0.9 with a decision rate of at least 20%.

[0510] In some implementations, the predictive 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, or higher 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 predictive model may have a positive predictive value of at least 0.1 at a decision rate of about 10%. For example, the predictive model may have a positive predictive value of at least 0.2 at a decision rate of about 10%. For example, the predictive model may have a positive predictive value of at least 0.3 at a decision rate of about 10%. For example, the predictive model may have a positive predictive value of at least 0.4 at a decision rate of about 10%. For example, a predictive model may have a positive predictive value of at least 0.5 with a decision rate of approximately 10%. For example, a predictive model may have a positive predictive value of at least 0.6 with a decision rate of approximately 10%. For example, a predictive model may have a positive predictive value of at least 0.7 with a decision rate of approximately 10%. For example, a predictive model may have a positive predictive value of at least 0.8 with a decision rate of approximately 10%. For example, a predictive model may have a positive predictive value of at least 0.9 with a decision rate of approximately 10%. For example, a predictive model may have a positive predictive value of at least 0.1 with a decision rate of approximately 5%. For example, a predictive model may have a positive predictive value of at least 0.2 with a decision rate of approximately 5%. For example, a predictive model may have a positive predictive value of at least 0.3 with a decision rate of approximately 5%. For example, a predictive model may have a positive predictive value of at least 0.4 with a decision rate of approximately 5%. For example, a predictive model may have a positive predictive value of at least 0.5 with a decision rate of approximately 5%. For example, a predictive model can have a positive predictive value of at least 0.6 with a decision rate of approximately 5%. For example, a predictive model can have a positive predictive value of at least 0.7 with a decision rate of approximately 5%. For example, a predictive model can have a positive predictive value of at least 0.8 with a decision rate of approximately 5%. For example, a predictive model can have a positive predictive value of at least 0.9 with a decision rate of approximately 5%. For example, a predictive model can have a positive predictive value of at least 0.1 with a decision rate of approximately 20%. For example, a predictive model can have a positive predictive value of at least 0.2 with a decision rate of approximately 20%. For example, a predictive model can have a positive predictive value of at least 0.3 with a decision rate of approximately 20%. For example, a predictive model can have a positive predictive value of at least 0.4 with a decision rate of approximately 20%. For example, a predictive model can have a positive predictive value of at least 0.5 with a decision rate of approximately 20%. For example, a predictive model can have a positive predictive value of at least 0.6 with a decision rate of approximately 20%. For example, a predictive model can have a positive predictive value of at least 0.7 with a decision rate of approximately 20%.For example, a predictive model can have a positive predictive value of at least 0.8 with a decision rate of approximately 20%. For example, a predictive model can have a positive predictive value of at least 0.9 with a decision rate of approximately 20%.

[0511] In some implementations, the predictive 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, or higher 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 predictive model may have a positive predictive value of at least 0.1 at a decision rate of up to 10%. For example, the predictive model may have a positive predictive value of at least 0.2 at a decision rate of up to 10%. For example, the predictive model may have a positive predictive value of at least 0.3 at a decision rate of up to 10%. For example, the predictive model may have a positive predictive value of at least 0.4 at a decision rate of up to 10%. For example, a predictive model may have a positive predictive value of at least 0.5 with a decision rate of up to 10%. For example, a predictive model may have a positive predictive value of at least 0.6 with a decision rate of up to 10%. For example, a predictive model may have a positive predictive value of at least 0.7 with a decision rate of up to 10%. For example, a predictive model may have a positive predictive value of at least 0.8 with a decision rate of up to 10%. For example, a predictive model may have a positive predictive value of at least 0.9 with a decision rate of up to 10%. For example, a predictive model may have a positive predictive value of at least 0.1 with a decision rate of up to 5%. For example, a predictive model may have a positive predictive value of at least 0.2 with a decision rate of up to 5%. For example, a predictive model may have a positive predictive value of at least 0.3 with a decision rate of up to 5%. For example, a predictive model may have a positive predictive value of at least 0.4 with a decision rate of up to 5%. For example, a predictive model may have a positive predictive value of at least 0.5 with a decision rate of up to 5%. For example, a predictive model may have a positive predictive value of at least 0.6 with a decision rate of up to 5%. For example, a predictive model may have a positive predictive value of at least 0.7 with a decision rate of up to 5%. For example, a predictive model may have a positive predictive value of at least 0.8 with a decision rate of up to 5%. For ex...

Claims

1. A pharmaceutical composition for treating cancer in a human subject in need thereof, comprising: (i) a polypeptide comprising a peptide sequence selected from a machine learning HLA peptide presentation prediction model, (ii) a polynucleotide encoding the polypeptide of (i), (iii) an APC comprising (i) or (ii), or (iv) a T cell comprising a T cell receptor (TCR) specific for a protein encoded by a class II HLA allele of the human subject that forms a complex with the polypeptide; in, The peptide sequence is selected by a method comprising the following steps: (a) using a computer processor to input amino acid information of a set of candidate peptide sequences into a machine learning HLA peptide presentation prediction model to generate a plurality of presentation predictions, wherein each candidate peptide sequence of the set of candidate peptide sequences is encoded by a genome, transcriptome or exome of a subject, or is encoded by a pathogen or virus in the subject, wherein the plurality of presentation predictions comprises an HLA presentation prediction for each candidate peptide sequence of the set of candidate peptide sequences, wherein each presentation prediction of the plurality of presentation predictions indicates a likelihood that a peptide sequence of the set of candidate peptide sequences is presented by one or more proteins encoded by a class II HLA allele of a cell of the human subject, wherein the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.07, wherein the PPV is calculated as the fraction of peptide sequences in the training data identified as presented by the class II HLA alleles of the cell among the peptide sequences predicted to be presented by the class II HLA alleles of the cell, and (b) selecting a subset of peptide sequences from the set of candidate peptide sequences based at least on the plurality of presentation predictions, wherein the subset of peptide sequences includes the selected peptide sequence; Among them, the selected peptide sequence (i) is encoded by the genome, transcriptome, or exome of a human subject, or is encoded by a pathogen or virus in a human subject, and (ii) Presentation of proteins predicted to be encoded by HLA class II alleles in cells of human subjects.

2. The pharmaceutical composition of claim 1, wherein the PPV of the machine learning HLA peptide presentation prediction model is at least 0.1 according to the presentation PPV determination method.

3. The pharmaceutical composition according to claim 1, wherein the machine learning HLA peptide presentation prediction model comprises: (i) a plurality of parameters identified based at least on training data, comprising: (1) The sequence of the training peptide, (2) the identity of the protein encoded by the class II HLA allele related to the sequence of the training peptide, and (3) observing presentation of one or more training peptides by proteins encoded by class II HLA alleles in the trained cells by mass spectrometry; and (ii) a function representing the relationship between the amino acid sequence information received as input and the presentation likelihood generated as output based on the amino acid sequence information and the plurality of parameters.

4. The pharmaceutical composition according to claim 1, wherein the machine learning HLA peptide presentation prediction model has a PPV of at least 0.2 at a judgment rate of 10% according to the presentation PPV determination method.

5. The pharmaceutical composition of claim 1, wherein the machine learning HLA peptide presentation prediction model has a PPV of at least 0.3 according to the presentation PPV determination method.

6. The pharmaceutical composition according to claim 3, wherein: (i) the training cells include training cells that express a single MHC class II complex or a protein encoded by a single allelic variant of a class II HLA locus selected from DR, DP and DQ, wherein the single MHC class II complex or the protein encoded by the single allelic variant of a class II HLA locus is expressed by cells of the subject; or (ii) The training data includes training data obtained by deconvolution.

7. A system for selecting one or more peptide sequences for preparing a pharmaceutical composition, the system comprising a computer processor, the system comprising: (a) an input module configured to receive amino acid sequence information of a set of candidate peptide sequences expressed by cells of a human subject, wherein each candidate peptide sequence of the set of candidate peptide sequences is encoded by a genome, transcriptome or exome of the human subject, or is encoded by a pathogen or virus in the human subject; (b) a processing module operably connected to the input module, the processing module comprising executable code, the executable code comprising a trained machine learning class II HLA-peptide presentation prediction model, wherein the trained machine learning class II HLA-peptide presentation prediction model is configured to generate an output peptide sequence having a plurality of presentation predictions, wherein each of the plurality of presentation predictions indicates a likelihood that a peptide sequence of the set of candidate peptide sequences is presented by one or more proteins encoded by a class II HLA allele of a cell of a human subject, wherein each peptide sequence of the subset of output peptide sequences is used to prepare a therapeutic composition for a human subject based on the presentation probability generated by the output as a peptide sequence, forming a complex with one or more proteins encoded by a class II HLA allele of a cell of the human subject, and Wherein according to the PPV determination method, the trained machine learning class II HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.

07.

8. A method for screening the immunogenicity of a drug comprising a polypeptide sequence in a subject, comprising: (a) obtaining a plurality of peptide sequences of the polypeptide sequence by a computer processor; (b) processing the amino acid information of the plurality of peptide sequences by a computer processor, generating a presentation prediction for each of the plurality of peptide sequences using a machine learning HLA peptide presentation prediction model, 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 can present an epitope sequence of a given peptide sequence of 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 identified by mass spectrometry as presented by HLA proteins expressed in cells; and (c) 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.

9. 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 a polypeptide sequence into a machine-learned HLA-peptide presentation prediction model to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that one or more proteins encoded by a class I or class II MHC allele of a cell of the subject will present an epitope sequence of a given peptide sequence; wherein 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: sequence information of peptide sequences presented by HLA proteins expressed in cells 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 likelihood generated as output based on the amino acid position information and the predictor variables; and (b) determining or predicting that each of the peptide sequences of the polypeptide sequences is not immunogenic to the subject based on the presentation prediction of the panel.

10. A method comprising: (a) using a computer processor to input amino acid information of a set of candidate peptide sequences into a trained machine learning class II HLA peptide presentation prediction model to generate a plurality of presentation predictions, wherein each candidate peptide sequence of the set of candidate peptide sequences is encoded by a genome, transcriptome or exome of a subject, or is encoded by a pathogen or virus in the subject; wherein the plurality of presentation predictions comprises an HLA presentation prediction for each candidate peptide sequence of the set of candidate peptide sequences, wherein each presentation prediction of the plurality of presentation predictions indicates a likelihood that a peptide sequence of the set of candidate peptide sequences is presented by one or more proteins encoded by a class II HLA allele of a cell of the subject; as well as (b) selecting a subset of peptide sequences from the set of candidate peptide sequences based at least on the plurality of presentation predictions to generate a selected subset of peptide sequences, wherein the selected subset of peptide sequences (i) is encoded by the genome, transcriptome, or exome of the subject, or is encoded by a pathogen or virus in the subject, and (ii) is predicted to be presented by a protein encoded by a class II HLA allele of a cell of the subject; Wherein, according to the presentation PPV determination method, the trained machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.2 at a judgment rate of 10%.

Images