Method and systems for prediction of HLA class ii-specific epitopes and characterization of CD4+ t cells
A machine learning-based method using LC-MS/MS data and quality metrics enhances HLA class II epitope prediction, improving the positive predictive value and enabling accurate identification of cancer-specific peptides for personalized cancer therapies.
Patent Information
- Application Number
- JP2024122210
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2019-08-23
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Current methods for predicting and identifying HLA class II-specific epitopes are inaccurate and inefficient, leading to suboptimal CD4+ T cell responses in cancer treatment, as they do not robustly quantify gene expression, enzymatic cleavage, and pathway biases, and rely on outdated prediction standards like NetMHCIIpan, which lack comprehensive allele-specific binding data.
A machine learning-based approach using LC-MS/MS monoallelic data and quality metrics to improve HLA class II epitope prediction, incorporating allele-specific binding rules and biological variables, and validating through T cell induction assays to enhance the positive predictive value (PPV) of peptide presentation models.
The method significantly improves the positive predictive value (PPV) of HLA class II epitope prediction to at least 0.07 or 0.1, enabling more accurate identification of cancer-specific peptides for personalized cancer therapies, thereby enhancing CD4+ T cell responses.
Smart Images

Figure 2024164839000026 
Figure 2024164839000027 
Figure 2024164839000028
Abstract
Description
[Technical field]
[0001] cross reference This application claims the benefit of U.S. Provisional Application No. 62 / 783,914, filed December 21, 2018; U.S. Provisional Application No. 62 / 826,827, filed March 29, 2019; U.S. Provisional Application No. 62 / 855,379, filed May 31, 2019; and U.S. Provisional Application No. 62 / 891,101, filed August 23, 2019, each of which is incorporated by reference in its entirety. [Background technology]
[0002] background The major histocompatibility complex (MHC) is a gene complex that encodes human leukocyte antigen (HLA) genes. HLA genes are expressed as protein heterodimers that are displayed on the surface of human cells to circulating T cells. HLA genes are highly polymorphic, allowing for fine-tuning of the adaptive immune system by HLA genes. The adaptive immune response relies, in part, on the ability of T cells to identify and eliminate cells that display disease-associated peptide antigens bound to human leukocyte antigen (HLA) heterodimers.
[0003] In humans, endogenous and exogenous proteins are processed into peptides by the proteasome and by cytosolic 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 class II), and the groups of peptides that bind to them and elicit an immune response are called HLA epitopes. HLA epitopes are important components that allow the immune system to detect danger signals such as pathogen infection and transformation of self. CD4+ T cells recognize class II MHC (HLA-DR, HLA-DQ, and HLA-DP) epitopes displayed on antigen-presenting cells (APCs) such as dendritic cells and macrophages. Endogenous processing and presentation of HLA class II-ligands is a complex process that requires various chaperones and a subset of enzymes, not all of which are well characterized. Presentation of HLA class II peptides activates helper T cells, which then promote B cell differentiation and antibody production as well as CTL responses. Activated helper T cells also secrete cytokines and chemokines that activate other T cells and induce their differentiation.
[0004] Understanding the peptide binding preference of each HLA class II heterodimer is the key to successfully predict which cancer- or tumor-specific antigen is likely to elicit cancer- or tumor-specific T cell response.There is a need for a method to identify and isolate specific HLA class II-associated peptides (e.g., neo-antigen peptides).Such methodologies and isolated molecules are useful, for example, for the development of therapeutics, including but not limited to immune-based therapeutics. Summary of the Invention [Means for solving the problem]
[0005] overview The methods and compositions described herein are used in a wide range of applications.For example, the methods and compositions described herein can be used to identify immunogenic antigen peptides, and can be used for the development of drugs such as personalized medicines, and for the isolation and characterization of antigen-specific T cells.
[0006] CD4+ T cell responses can have anti-tumor activity. When class II prediction is not used, CD4+ T cell responses can be shown to be highly probable (e.g., 60% of SLP epitopes in the NeoVax study (49% in NT-001, see Ott et al., Nature, 2017 Jul 13; 547 (7662): 217-221) and 49% in the Biont study (see Ott et al., Nature, 2017 Jul 13; 547 (7662): 217-221). (48% of mRNA epitopes in a study by ech; see Sahin et al., Nature, 2017 Jul 13; 547 (7662): 222-226). These epitopes are generally It may not be clear whether the HLA class II-binding epitopes are presented natively (by the tumor or by phagocytic DCs). It may be desirable to translate high CD4+ T response rates into therapeutic efficacy by improving the identification of truly presented HLA class II-binding epitopes.
[0007] The role of gene expression, enzymatic cleavage, and pathway / localization bias may not be robustly quantified. It may be unclear whether the more relevant pathway is autophagy (HLA class II presentation by tumor cells) or phagocytosis (HLA class II presentation of tumor epitopes by APCs), but it can be presumed that the majority of existing MS data is derived from autophagy. NetMHCIIpan may be the current predictive standard, but this may not be considered accurate. Of the three HLA class II loci (DR, DP, and DQ), data may only exist for certain common alleles of HLA-DR.
[0008] There may be various data generation methods for learning the rules of HLA Class II presentation, including the standard in the field and the proposed method. The standard in the field may include affinity measurements, which may be the basis of the NetMHCIIpan predictor, resulting in low throughput and the need for radioactive reagents, and the standard in the field may lack the role of processing. The proposed method may include mass spectrometry, where data from cell lines / tissues / tumors may help determine the processing rules for autophagy, and single-allelic MS may allow the determination of allele-specific binding rules (multi-allelic MS data is presumed to be too complex for efficient learning (Bassani-Sternberg. MCP. 2018)).
[0009] There can be different ways to validate a new HLA class II predictor: validation against submitted MS data, which can be the default setting; retrospective vaccine trials (e.g., NT-001) where immune monitoring data can assess vaccine peptide loading on APCs rather than tumor presentation, and data can be spread thinly across many different alleles; biochemical affinity measurements that can be configured to obtain measurements on discordantly predicted peptides (only for 2-3 alleles); ex vivo validation with Neon-preferred epitopes and NetMHCIIpan-preferred epitopes. T cell induction can be configured to test the rate at which T cell responses are induced in vivo.
[0010] For validation by T cell induction, the default approach may include evaluating discordantly predicted TCGA neoORFs, induction material may include healthy donor APCs and T cells, induction and readout may be by SLP (approximately 15mer peptides). Random peptides may generate high response rates, and SLP may not address processing well. A possible solution may include induction by mRNA.
[0011] The method disclosed herein may include generating LC-MS / MS single allele data for training allele-specific machine learning methods for epitope prediction. Such methods may include: improving LC-MS / MS data quality using a set of quality metrics to rigorously remove false positives that improve the performance of the prediction model; identifying allele-specific HLA class II binding cores from the HLA-ligandome LC-MS / MS dataset; improving HLA class II-ligand and epitope prediction using machine learning algorithms; and / or identifying biological variables that affect HLA class II-ligand presentation and improve HLA class II epitope prediction, such as gene expression, cleavage, gene bias, cellular localization, and secondary structure.
[0012] Provided herein is a method comprising: (a) processing amino acid information of a plurality of candidate peptide sequences using a machine learning HLA peptide presentation prediction model to generate a plurality of presentation predictions, each candidate peptide sequence of the plurality of candidate peptide sequences being encoded by a genome or exome of a subject, the plurality of presentation predictions including an HLA presentation prediction for each of the plurality of candidate peptide sequences, each HLA presentation prediction indicating a likelihood that a given candidate peptide sequence of the plurality of candidate peptide sequences may be presented by one or more proteins encoded by class II HLA alleles of a cell of the subject, the machine learning HLA peptide presentation prediction model being trained using training data including sequence information of sequences of training peptides identified by mass spectrometry to be presented by HLA proteins expressed in the training cells; and (b) identifying, based on at least the plurality of presentation predictions, peptide sequences among the plurality of peptide sequences that are presented by at least one of the one or more proteins encoded by class II HLA alleles of the cell of the subject, wherein a positive predictive value (PPV) of the machine learning HLA peptide presentation prediction model is at least 0.07 according to a presentation PPV determination method.
[0013] Provided herein is a method comprising: (a) processing amino acid information of a plurality of peptide sequences encoded by a genome or exome of a subject using a machine learning HLA peptide binding prediction model to generate a plurality of binding predictions, the plurality of binding predictions including an HLA binding prediction for each of a plurality of candidate peptide sequences, each binding prediction indicating a likelihood that one or more proteins encoded by class II HLA alleles of a cell of the subject bind to a given candidate peptide sequence of the plurality of candidate peptide sequences, the machine learning HLA peptide binding prediction model being trained using training data including sequence information of sequences of peptides identified to bind to HLA class II proteins or HLA class II protein analogs; and (b) identifying peptide sequences among the plurality of peptide sequences that have a probability greater than a threshold binding prediction probability value for binding to at least one of the one or more proteins encoded by class II HLA alleles of a cell of the subject based on at least the plurality of binding predictions, wherein a positive predictive value (PPV) of the machine learning HLA peptide binding prediction model is at least 0.1 according to a binding PPV determination method.
[0014] In some embodiments, the machine learning HLA peptide presentation prediction model is trained using training data that includes sequence information for sequences of training peptides identified by mass spectrometry as being presented by HLA proteins expressed in training cells.
[0015] In some embodiments, the method includes ranking at least two peptides identified as being presented by at least one of the one or more proteins encoded by class II HLA alleles of the subject's cell based on the presentation prediction.
[0016] In some embodiments, the method comprises selecting one or more peptides out of the two or more ranked peptides.
[0017] In some embodiments, the method includes selecting one or more peptides from a plurality identified as being presented by at least one of the one or more proteins encoded by class II HLA alleles of the subject's cells.
[0018] In some embodiments, the method includes selecting one or more peptides of the two or more peptides ranked based on the presentation prediction.
[0019] In some embodiments, the amino acid information of the plurality of test peptide sequences is processed to generate a plurality of test presentation predictions, where each test presentation prediction indicates a likelihood that a given test peptide sequence of the plurality of test peptide sequences may be presented by one or more proteins encoded by class II HLA alleles of a cell of the subject, the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.07, the plurality of test peptide sequences comprises at least 500 test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, and (ii) at least 499 decoy peptide sequences contained within a protein encoded by the genome of the organism, the organism and the subject are of the same species, the plurality of test peptide sequences comprises at least one hit peptide sequence and at least 499 decoy peptide sequences in a ratio of 1:499, and the machine learning HLA peptide presentation prediction model predicts that a top percentage of the plurality of test peptide sequences are presented by HLA proteins expressed in the cell.
[0020] In some embodiments, the amino acid information of the plurality of test peptide sequences is processed to generate a plurality of test binding predictions, where each test binding prediction indicates a likelihood that one or more proteins encoded by class II HLA alleles of the cell of the subject binds to a given test peptide sequence of 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, where the plurality of test peptide sequences includes at least 20 test peptide sequences including (i) at least one hit peptide sequence identified by mass spectrometry to be presented by an HLA protein expressed in the cell, and (ii) at least 19 decoy peptide sequences contained within a protein that includes at least one peptide sequence identified by mass spectrometry to be presented by an HLA protein expressed in the cell, e.g., a single HLA protein expressed in the cell (e.g., a single allele cell), where the plurality of test peptide sequences includes at least one hit peptide sequence and at least 19 decoy peptide sequences in a 1:19 ratio, and the machine learning HLA peptide presentation prediction model predicts that a top percentage of the plurality of test peptide sequences bind to an HLA protein expressed in the cell.
[0021] In some embodiments, there is no amino acid sequence overlap between at least one hit peptide sequence and the decoy peptide sequence.
[0022] In some embodiments, the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.08, 0.09, 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.2, 0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29, 0.3, 0.31, 0.32, 0.33, 0.34, 0.35, 0.36, 0.37, 0.38, 0.39, 0.4, 0.41, 0.42, 0.43, 0.44, 0.45, 0.46, 0.47, 0.48, 0.49 ,0.5,0.51,0.52,0.53,0.54,0.55,0.56,0.57,0.58,0.59,0.6,0.61,0.62,0.63,0.64,0.65,0.66,0.67,0.68,0.69,0.7,0.71,0.72,0.73,0.74,0. 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.
[0023] In some embodiments, at least one hit peptide sequence is 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.
[0024] In some embodiments, the at least 499 decoy peptide sequences are 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, 540 0, 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, 1 2000, 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, 5 5000, 57500, 60000, 62500, 65000, 67500, 70000, 72500, 75000, 77500, 80000, 82500, 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000,The hit:decoy peptide sequence may be 800,000, 900,000 or 1,000,000. One skilled in the art can appreciate that varying the hit:decoy ratio will alter the PPV.
[0025] In some embodiments, the at least 500 test peptide sequences are at least 600, 70 0, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 320 0, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 57 00, 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, 1500 0, 16000, 17000, 18000, 19000, 20000, 21000, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000, 3 6000, 37000, 38000, 39000, 40000, 41000, 42000, 43000, 44000, 45000, 46000, 47000, 48000, 49000, 50000, 52500, 55000, 57500, 60000, 62500, 65000 , 67500, 70000, 72500, 75000, 77500, 80000, 82500, 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 test peptide sequences.
[0026] In some embodiments, the top 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%, 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%, 5.0 ... %, 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%.
[0027] In some embodiments, at least one hit peptide sequence is 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.
[0028] In some embodiments, the at least 19 decoy peptide sequences are at least 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500,600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 310 0, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 560 0, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 81 00, 8200, 8300, 8400, 8500, 8600, 8700, 8800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 10000, 11000, 12000, 13000, 14000, 15 000, 16000, 17000, 18000, 19000, 20000, 21000, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41000, 42000, 43000, 44000, 45000, 46000, 47000, 48000, 49000, 50000, 52500, 55000, 57500, 60000, 62500, 65000 , 67500, 70000, 72500, 75000, 77500, 80000, 82500, 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 decoy peptide sequences.
[0029] 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, 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, 44 00, 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, 8 200, 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, 260 00, 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,The test peptide sequences include 70000, 72500, 75000, 77500, 80000, 82500, 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 test peptide sequences.
[0030] In some embodiments, 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%.
[0031] In some embodiments, the PPV is greater than each PPV in the second column of Table 11 for the protein encoded by the corresponding HLA allele in Table 11. In some embodiments, the PPV is at least equal to each PPV in the third column of Table 11 for the protein encoded by the corresponding HLA allele in Table 11.
[0032] In some embodiments, the PPV is equal to or greater than the respective PPV in column 2 of Table 12 for proteins encoded by HLA class II alleles.
[0033] In some embodiments, the PPV is greater than each of the PPVs in column 2 of Table 16 for proteins encoded by HLA class II alleles.
[0034] In some embodiments, the subject is a single subject.
[0035] In some embodiments, the subject is a mammal.
[0036] In some embodiments, the subject is a human.
[0037] In some embodiments, the training cells are cells that express a single protein encoded by a class II HLA allele of the subject's cells.
[0038] In some embodiments, the training cells are monoallelic HLA cells, or cells expressing an HLA allele with an affinity tag.
[0039] In some embodiments, the cells of the subject comprise cancer cells.
[0040] In some embodiments, the method is for identifying a peptide sequence.
[0041] In some embodiments, the method is for selecting a peptide sequence.
[0042] In some embodiments, the method is for preparing a cancer treatment.
[0043] In some embodiments, the method is for preparing a target specific cancer therapy.
[0044] In some embodiments, the method is for preparing a cancer cell specific cancer therapy.
[0045] In some embodiments, each peptide sequence of the plurality of peptide sequences is associated with cancer.
[0046] In some embodiments, at least one peptide sequence of the plurality of peptide sequences is overexpressed by cancer cells of the subject.
[0047] In some embodiments, each peptide sequence of the plurality of peptide sequences is overexpressed by cancer cells of the subject.
[0048] In some embodiments, at least one peptide sequence of the plurality of peptide sequences is a cancer cell-specific peptide.
[0049] In some embodiments, each peptide sequence of the plurality of peptide sequences is a cancer cell-specific peptide.
[0050] In some embodiments, each peptide sequence of the plurality of peptide sequences is expressed by a cancer cell of the subject.
[0051] In some embodiments, at least one peptide sequence of the plurality of peptide sequences is not encoded by non-cancer cells of the subject.
[0052] In some embodiments, each peptide sequence of the plurality of peptide sequences is not encoded by non-cancer cells of the subject.
[0053] In some embodiments, at least one peptide sequence of the plurality of peptide sequences is not expressed in non-cancer cells of the subject.
[0054] In some embodiments, each peptide sequence of the plurality of peptide sequences is not expressed in non-cancer cells of the subject.
[0055] In some embodiments, the method includes obtaining a plurality of peptide sequences of the subject.
[0056] In some embodiments, the methods include obtaining a plurality of polynucleotide sequences of the subject.
[0057] In some embodiments, the methods include obtaining a plurality of polynucleotide sequences from a subject that encode a plurality of peptide sequences encoded by the genome or exome of the subject, or by a pathogen or virus in the subject.
[0058] In some embodiments, the method comprises obtaining, by a computer processor, a plurality of polynucleotide sequences of the subject that encode a plurality of peptide sequences encoded by the genome or exome of the subject.
[0059] 2. The method of claim 1, wherein the plurality of polynucleotide sequences is a sequence of a subject.
[0060] In some embodiments, the methods include obtaining the plurality of polynucleotide sequences of the subject by whole genome sequencing or whole exome sequencing.
[0061] In some embodiments, the processing step comprises processing by a computer processor.
[0062] In some embodiments, the processing step includes generating a plurality of predictor variables based on amino acid information of at least the plurality of peptide sequences.
[0063] In some embodiments, a machine learning HLA-peptide presentation predictive model is used to process a plurality of predictor variables.
[0064] In some embodiments, the one or more proteins encoded by class II HLA alleles in the cells of the subject are one or more proteins encoded by class II HLA alleles expressed by the subject.
[0065] In some embodiments, the one or more proteins encoded by class II HLA alleles of the subject's cells are one or more proteins encoded by class II HLA alleles expressed by cancer cells of the subject.
[0066] In some embodiments, the one or more proteins encoded by class II HLA alleles of the subject's cells is a single protein encoded by class II HLA alleles of the subject's cells.
[0067] In some embodiments, the one or more proteins encoded by class II HLA alleles of the subject's cells are two, three, four, five, or six or more proteins encoded by class II HLA alleles of the subject's cells.
[0068] In some embodiments, the one or more proteins encoded by class II HLA alleles of the subject's cells is each protein encoded by class II HLA alleles of the subject's cells.
[0069] In some embodiments, the method further comprises administering to the subject a composition comprising one or more of the selected subset of peptide sequences.
[0070] In some embodiments, the step of identifying a plurality of peptide sequences comprises comparing a DNA, RNA, or protein sequence from the subject's cancer cells to a DNA, RNA, or protein sequence from the subject's normal cells, wherein each of the plurality of peptides comprises at least one mutation that is present in the subject's cancer cells and not present in the subject's normal cells.
[0071] In some embodiments, the machine learning HLA-peptide presentation prediction model includes a plurality of predictor variables identified based on at least training data, where the training data includes training peptide sequence information including amino acid position information, the training peptide sequence information including the plurality of predictor variables associated with HLA proteins expressed in the cell; and a function expressing a relationship between the amino acid position information and the likelihood of presentation generated as an output based on the amino acid position information and the plurality of predictor variables.
[0072] In some embodiments, the identifying step comprises identifying peptide sequences among the plurality of peptide sequences that have a probability of being presented by at least one of the one or more proteins encoded by class II HLA alleles in the subject's cell that is greater than a threshold predicted probability of presentation value.
[0073] In some embodiments, one or more of 0.2% of a plurality of test peptide sequences predicted to be presented by a machine learning HLA peptide presentation prediction model have a higher probability of being presented by at least one of one or more proteins encoded by class II HLA alleles of a subject's cell than a threshold presentation prediction probability value.
[0074] In some embodiments, 0.2% of a plurality of test peptide sequences predicted to be presented by the machine learning HLA peptide presentation prediction model each have a probability of being presented by at least one of the one or more proteins encoded by class II HLA alleles of the subject's cell that is greater than a threshold presentation prediction probability value.
[0075] In some embodiments, the number of positives is constrained to be equal to the number of hits.
[0076] In some embodiments, the mass spectrometry is single-allelic mass spectrometry.
[0077] In some embodiments, peptides are presented by HLA proteins expressed in cells through autophagy.
[0078] In some embodiments, peptides are presented by HLA proteins expressed in cells through phagocytosis.
[0079] In some embodiments, the plurality of predictor variables comprises an expression level predictor of a source protein comprising a peptide.
[0080] In some embodiments, the plurality of predictor variables comprises a stability predictor of a source protein that includes a peptide.
[0081] In some embodiments, the plurality of predictor variables comprises a degradation rate predictor of a source protein that includes a peptide.
[0082] In some embodiments, the plurality of predictor variables comprises a proteolytic cleavability predictor for a source protein that includes the peptide.
[0083] In some embodiments, the plurality of predictor variables comprises a cellular or tissue localization predictor of a source protein, including a peptide.
[0084] In some embodiments, the plurality of predictor variables includes predictors of the intracellular processing manner of a source protein comprising a peptide, the processing manner of the source protein including predictors of whether the source protein is subject to autophagy, phagocytosis, and intracellular trafficking, among others.
[0085] In some embodiments, the quality of the training data is improved using multiple quality metrics.
[0086] In some embodiments, the multiple quality metrics include general contaminating peptide removal, high scored peak intensity, high score, and high mass accuracy.
[0087] In some embodiments, the scored peak intensity is at least 50%.
[0088] In some embodiments, the scored peak intensity is at least 60%.
[0089] In some embodiments, the score is at least 7.
[0090] In some embodiments, the mass accuracy is up to 5 ppm.
[0091] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single immunoprecipitated HLA protein expressed in a cell.
[0092] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single exogenous HLA protein expressed in a cell.
[0093] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single recombinant HLA protein expressed in a cell.
[0094] In some embodiments, the plurality of predictor variables comprises a peptide-HLA affinity predictor variable.
[0095] In some embodiments, peptides presented by HLA proteins include peptides identified by searching an enzyme-specific, unmodified peptide database.
[0096] In some embodiments, peptides presented by HLA proteins include peptides identified by searching peptide databases using a reverse database searching strategy.
[0097] In some embodiments, the HLA protein comprises an HLA-DR protein, an HLA-DQ protein, or an HLA-DP protein.
[0098] In some embodiments, the HLA protein is 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.
[0099] In some embodiments, HLA-DR is DRA * Pairs with 01:01.
[0100] In some embodiments, the HLA protein is DPA * 01:03 / DPB * 04:01, DRB1 * 01:01, DRB1 * 01:02, DRB1 * 03:01, DRB1 * 04:01, DRB1 * 04:02, DRB1 * 04:04, DRB1 * 04:05, DRB1 * 07:01, DRB1 * 08:01, DRB1 * 08:02, DRB1 * 08:03, DRB1 * 09:01, DRB1 * 11:01, DRB1 * 11:02, DRB1 * 11:04, DRB1 * 12:01, DRB1 * 13:01, DRB1 * 13:02, DRB1 * 13:03, DRB1 * 14:01, DRB1 *15:01, DRB1 * 15:02, DRB1 * 15:03, DRB1 * 16:02, DRB3 * 01:01, DRB3 * 02:01, DRB3 * 02:02, DRB3 * 03:01, DRB4 * 01:01, DRB4 * 01:03 and DRB5 * 01:01.
[0101] In some embodiments, the HLA-DR protein is * Contains 01:01 in the dimer.
[0102] In some embodiments, the HLA protein is DPB1 * 01:01, DPB1 * 02:01, DPB1 * 02:02, DPB1 * 03:01, DPB1 * 04:01, DPB1 * 04:02, DPB1 * 05:01, DPB1 * 06:01, DPB1 * 11:01, DPB1 * 13:01, DPB1 * 17:01.
[0103] In some embodiments, the HLA-DP protein is DPA1 * Forms a pair including 01:03.
[0104] In some embodiments, the HLA protein is 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.
[0105] In some embodiments, peptides presented by HLA proteins include peptides identified by comparing the MS / MS spectrum of an HLA-peptide with the MS / MS spectrum of one or more peptides or proteins in a peptide or protein database.
[0106] In some embodiments, the mutation is selected from the group consisting of a point mutation, a splice site mutation, a frameshift mutation, a readthrough mutation, and a gene fusion mutation.
[0107] In some embodiments, peptides presented by HLA proteins are between 15 and 40 amino acids in length.
[0108] In some embodiments, peptides presented by an HLA protein include peptides identified by identifying a peptide presented by an HLA protein by comparing the MS / MS spectrum of the HLA-peptide with the MS / MS spectrum of one or more peptides or proteins in a peptide or protein database.
[0109] In some embodiments, the personalized cancer treatment further comprises an adjuvant.
[0110] In some embodiments, the personalized cancer treatment further comprises an immune checkpoint inhibitor.
[0111] In some embodiments, the training data includes structured data, time series data, unstructured data, relational data, or any combination thereof.
[0112] In some embodiments, the unstructured data includes image data.
[0113] In some embodiments, the relational data includes data from customer systems, enterprise systems, operational systems, websites, web-accessible application program interfaces (APIs), or any combination thereof.
[0114] In some embodiments, the training data is uploaded to a cloud-based database.
[0115] In some embodiments, the training is performed using a convolutional neural network.
[0116] In some embodiments, the convolutional neural network includes at least two convolutional layers.
[0117] In some embodiments, the convolutional neural network includes at least one batch normalization step.
[0118] In some embodiments, the convolutional neural network includes at least one spatial dropout step.
[0119] In some embodiments, the convolutional neural network includes at least one global max pooling step.
[0120] In some embodiments, the convolutional neural network includes at least one dense layer.
[0121] In some embodiments, the step of identifying a peptide sequence comprises identifying a peptide sequence with a mutation expressed in cancer cells of the subject.
[0122] In some embodiments, the step of identifying a peptide sequence comprises identifying a peptide sequence that is not expressed in normal cells of the subject.
[0123] In some embodiments, the step of identifying a peptide sequence comprises identifying a viral peptide sequence.
[0124] In some embodiments, the step of identifying peptide sequences comprises identifying overexpressed peptide sequences.
[0125] A method of identifying HLA class II specific peptides for immunotherapy of a subject, comprising the steps of: obtaining, by a computer processor, candidate peptides comprising an epitope, and a plurality of peptide sequences, each comprising the epitope; and processing, by the computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences to an immune cell, each presentation prediction indicating a likelihood that a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by an HLA class II allele, the machine learning HLA-peptide presentation prediction model being based on the sequence of peptides presented by HLA proteins expressed in a cell and identified by mass spectrometry. Provided herein is a method comprising the steps of: training the model using training data comprising sequence information of a sequence; selecting a protein predicted by the machine learning HLA-peptide presentation prediction model to bind to a candidate peptide from one or more proteins encoded by HLA class II alleles of a subject's cell, the protein having a probability greater than a threshold presentation prediction probability value for presentation of the candidate peptide to an immune cell; contacting the candidate peptide with the selected protein, thus allowing the candidate peptide to compete with a placeholder peptide associated with the selected protein; and identifying the candidate peptide as a peptide for immunotherapy specific to the selected protein based on whether the candidate peptide replaces the placeholder.
[0126] In some embodiments, the obtaining step comprises identifying a candidate peptide, and identifying the candidate peptide comprises 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.
[0127] In some embodiments, the processing step includes identifying a plurality of predictor variables based on amino acid information of at least a plurality of peptide sequences, and processing the plurality of predictor variables using a machine learning HLA-peptide presentation prediction model.
[0128] In some embodiments, the machine learning HLA-peptide presentation prediction model includes a plurality of predictor variables identified based on at least training data, where the training data includes training peptide sequence information including amino acid position information, the training peptide sequence information being associated with HLA proteins expressed in a cell; and a function expressing a relationship between the amino acid position information and a presentation probability generated as an output based on the amino acid position information and the plurality of predictor variables.
[0129] In some embodiments, the number of positives is constrained to be equal to the number of hits.
[0130] In some embodiments, the mass spectrometry is single-allelic mass spectrometry.
[0131] In some embodiments, the plurality of predictor variables comprises any one or more of a source protein expression level predictor, a stability predictor, a degradation rate predictor, a cleavability predictor, a cell or tissue localization predictor, and an intracellular processing mode predictor, including autophagy, phagocytosis, and intracellular trafficking, including a peptide.
[0132] In some embodiments, the quality of the training data is improved using multiple quality metrics.
[0133] In some embodiments, the multiple quality metrics include general contaminating peptide removal, high scored peak intensity, high score, and high mass accuracy.
[0134] In some embodiments, the scored peak intensity is at least 50%.
[0135] In some embodiments, the scored peak intensity is at least 60%.
[0136] In some embodiments, the placeholder peptide is a CLIP peptide.
[0137] In some embodiments, the placeholder peptide is a CMV peptide.
[0138] In some embodiments, the three methods further comprise the step of determining the IC50 of replacement of the placeholder peptide with the target peptide.
[0139] In some embodiments, the IC50 of replacement of the placeholder peptide with the target peptide is less than 500 nM.
[0140] In some embodiments, at least one protein from the one or more proteins encoded by HLA class II alleles of the subject's cells is an HLA class II tetramer or multimer.
[0141] In some embodiments, the target peptide is further identified by mass spectrometry.
[0142] In some embodiments, at least one protein encoded by an HLA class II allele of the subject's cells is a recombinant protein.
[0143] In some embodiments, at least one protein encoded by an HLA class II allele of the subject's cells is expressed in a eukaryotic cell.
[0144] In some embodiments, peptides are presented by HLA proteins expressed in cells through autophagy.
[0145] In some embodiments, peptides are presented by HLA proteins expressed in cells through phagocytosis.
[0146] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single immunoprecipitated HLA protein expressed in a cell.
[0147] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single exogenous HLA protein expressed in a cell.
[0148] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single recombinant HLA protein expressed in a cell.
[0149] In some embodiments, the plurality of predictor variables comprises a peptide-HLA affinity predictor variable.
[0150] In some embodiments, peptides presented by HLA proteins include peptides identified by searching an enzyme-specific, unmodified peptide database.
[0151] In some embodiments, peptides presented by HLA proteins include peptides identified by searching peptide databases using a reverse database searching strategy.
[0152] In some embodiments, the HLA protein comprises an HLA-DR protein, an HLA-DQ protein, or an HLA-DP protein.
[0153] In some embodiments, the immunotherapy is cancer immunotherapy.
[0154] In some embodiments, the epitope is a cancer-specific epitope.
[0155] In some embodiments, at least one protein encoded by an HLA class II allele comprises at least an alpha 1 and a beta 1 subunit of an HLA protein that exists in a dimeric form.
[0156] In some embodiments, it is known what the peptide is.
[0157] In some embodiments, it is not known what kind of peptide the peptide is.
[0158] In some embodiments, what a peptide is is determined by mass spectrometry.
[0159] In some embodiments, the peptide exchange assay comprises detection of a peptide fluorescent probe or tag.
[0160] In some embodiments, the placeholder peptide is a CLIP peptide. In some embodiments, the placeholder peptide has the amino acid sequence PVSKMRMATPLLMQA.
[0161] In some embodiments, the polynucleic acid construct comprises an expression vector further comprising one or more of a promoter, a secretion signal, a dimerization factor, a ribosome skipping sequence, one or more tags for purification and / or detection.
[0162] In some embodiments, the placeholder peptide sequence is encoded by a nucleic acid sequence in a vector.
[0163] In some embodiments, the sequence encoding the cleavable domain is placed between the sequence encoding the placeholder peptide and the sequence encoding the HLA beta 1 peptide.
[0164] Provided herein is a method for assaying immunogenicity of an MHC class II binding peptide, comprising: selecting a protein encoded by an HLA class II allele that is predicted to bind to the MHC class II binding peptide by a machine learning HLA-peptide presentation prediction model configured to generate a presentation prediction for a given peptide sequence, the presentation prediction indicating a likelihood that the given peptide sequence may be presented by one or more proteins encoded by the HLA class II alleles, the protein having a probability greater than a presentation prediction probability value threshold for presentation of the MHC class II binding peptide; contacting the peptide with the selected protein, thus allowing the peptide to compete with a placeholder peptide associated with the selected protein and displacing the placeholder peptide, 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 assaying for one or more activation parameters of the CD4+ T cells selected from the group consisting of induction of cytokines, induction of chemokines, and expression of cell surface markers.
[0165] In some embodiments, the HLA class II allele is a tetramer or multimer.
[0166] In some embodiments, the cytokine is IL-2.
[0167] 1. A method for inducing CD4+ T cell activation in a subject for cancer immunotherapy, comprising the steps of: identifying a peptide sequence associated with cancer and comprising a cancer mutation, wherein the identification of the peptide sequence comprises comparing a DNA, RNA, or protein sequence from cancer cells of the subject with a DNA, RNA, or protein sequence from normal cells of the subject; and selecting a protein encoded by an HLA class II allele that is normally expressed by cells of the subject and predicted to bind to the peptide by a machine learning HLA-peptide presentation prediction model, wherein the prediction model has a positive predictive value of at least 0.1 with a recall of at least 0.1%, between 0.1% and 50%, or up to 50%, and the protein is selected from the peptides identified. Provided herein is a method comprising the steps of: determining whether an identified peptide has a probability of presentation of the identified peptide sequence greater than a threshold predicted probability value for presentation of the identified peptide sequence; contacting the identified peptide with a protein encoded by a selected HLA class II allele to verify whether the identified peptide competes with and replaces a placeholder peptide associated with the protein encoded by the selected HLA class II allele with an IC50 value of less than 500 nM; optionally purifying the identified peptide; and administering to a subject an effective amount of a polypeptide comprising the sequence of the identified peptide or a polynucleotide encoding the polypeptide.
[0168] Provided herein is a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: obtaining, by a computer processor, a plurality of peptide sequences of the polypeptide sequence; processing, by the computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate presentation predictions for each of the plurality of peptide sequences, wherein each presentation prediction indicates a likelihood that an epitope sequence of a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information associated with HLA proteins expressed in the cell; determining or predicting that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject based on the plurality of presentation predictions; and administering to the subject a composition comprising the drug.
[0169] Provided herein is a method for producing HLA class II tetramers or multimers by conjugating four individual HLA protein alpha1 and beta1 heterodimers, comprising expressing in eukaryotic cells a vector comprising nucleic acid sequences encoding the alpha and beta chains of HLA proteins, a secretion signal, a biotinylation motif and at least one tag for identification or purification, thus secreting each HLA protein alpha1 and beta1 heterodimer in a dimerized state, where the heterodimer is associated with a placeholder peptide, purifying the secreted heterodimers from cell culture medium, verifying peptide binding activity using a peptide exchange assay, adding streptavidin, thereby conjugating the heterodimers to tetramers, and purifying the tetramers to obtain a yield of more than 1 mg / L. Multimers, such as pentamers, hexamers or octamers, can also be produced and are equally intended herein.
[0170] In some embodiments, the vector comprises a CMV promoter.
[0171] In some embodiments, the vector comprises a sequence encoding a placeholder peptide linked to the beta 1 chain via a cleavable site.
[0172] In some embodiments, the peptide exchange assay involves pre-cleavage of the placeholder peptide from the beta chain.
[0173] In some embodiments, the cleavable site is a thrombin cleavage site.
[0174] In some embodiments, the peptide exchange assay is a FRET assay.
[0175] In some embodiments, purification is by any one of column chromatography, ion exchange chromatography, molecular sieve chromatography, affinity chromatography, or LC-MS.
[0176] Provided herein is an HLA class II tetramer or multimer comprising either an HLA-DR heterodimer, or an HLA-DP heterodimer, or an HLA-DQ heterodimer, each heterodimer comprising an alpha chain and a beta chain, and wherein the heterodimers are purified and are present at a concentration of greater than 1 mg / L.
[0177] In some embodiments, the HLA class II tetramer is selected from Tables 8A-8C.
[0178] In some embodiments, the HLA class II tetramer comprises a heterodimeric pair selected from the group consisting of an HLA-DR protein, an HLA-DP protein, and an HLA-DQ protein.
[0179] In some embodiments, the HLA protein is HLA-DPB1* 01:01 / HLA-DPA1 * 01:03、HLA-DPB1 * 02:01 / HLA-DPA1 * 01:03、HLA-DPB1 * 03:01 / HLA-DPA1 * 01:03、HLA-DPB1 * 04:01 / HLA-DPA1 * 01:03、HLA-DPB1 * 04:02 / HLA-DPA1 * 01:03、HLA-DPB1 * 06:01 / HLA-DPA1 * 01:03、HLA-DQB1 * 02:01 / HLA-DQA1 * 05:01、HLA-DQB1 * 02:02 / HLA-DQA1 * 02:01、HLA-DQB1 * 06:02 / HLA-DQA1 * 01:02、HLA-DQB1 * 06:04 / HLA-DQA1 * 01:02、HLA-DRB1 * 01:01、HLA-DRB1 * 01:02、HLA-DRB1 * 03:01、HLA-DRB1 * 03:02、HLA-DRB1 * 04:01、HLA-DRB1 * 04:02、HLA-DRB1 * 04:03、HLA-DRB1 * 04:04、HLA-DRB1 * 04:05、HLA-DRB1 * 04:07、HLA-DRB1 * 07:01、HLA-DRB1 * 08:01、HLA-DRB1 * 08:02、HLA-DRB1 * 08:03、HLA-DRB1 * 08:04、HLA-DRB1 * 09:01、HLA-DRB1 * 10:01、HLA-DRB1* 11:01, HLA-DRB1 * 11:02, HLA-DRB1 * 11:04, HLA-DRB1 * 12:01, HLA-DRB1 * 12:02, HLA-DRB1 * 13:01, HLA-DRB1 * 13:02, HLA-DRB1 * 13:03, HLA-DRB1 * 14:01, HLA-DRB1 * 15:01, HLA-DRB1 * 15:02, HLA-DRB1 * 15:03, HLA-DRB1 * 16:01, HLA-DRB3 * 01:01, HLA-DRB3 * 02:02, HLA-DRB3 * 03:01, HLA-DRB4 * 01:01 and HLA-DRB5 * 01:01.
[0180] In some embodiments, the heterodimer pair is expressed in a eukaryotic cell.
[0181] In some embodiments, the heterodimer pair is encoded by a vector.
[0182] Provided herein are vectors that contain nucleic acid sequences encoding the alpha and beta chains of the HLA proteins described herein, a secretion signal, a biotinylation motif and at least one tag for identification or purification, thereby allowing the secretion of each HLA protein alpha1 and beta1 heterodimer in a dimeric state, the secreted heterodimers optionally being associated with a placeholder peptide.
[0183] Provided herein are cells comprising the vectors described herein.
[0184] In some embodiments, the HLA class II heterodimer is secreted from eukaryotic cells into cell culture medium and further purified by any one of column chromatography, ion exchange chromatography, molecular sieve chromatography, affinity chromatography, or LC-MS.
[0185] Provided herein is a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: obtaining, by a computer processor, a plurality of peptide sequences of the polypeptide sequence; processing, by the computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate presentation predictions for each of the plurality of peptide sequences, wherein each presentation prediction indicates a likelihood that an epitope sequence of a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; and determining or predicting that at least one of the plurality of peptide sequences of the polypeptide sequence is immunogenic to the subject based on the plurality of presentation predictions.
[0186] Provided herein is a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction indicating a probability that an epitope sequence of the given peptide sequence will be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model comprising a plurality of predictor variables identified based on at least training data, the training data comprising: sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, the training peptide sequence information being associated with the HLA proteins expressed in the cell; and a function expressing a 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 that each of the peptide sequences of the polypeptide sequence is not immunogenic to the subject based on the set of presentation predictions; and administering a composition comprising the drug to the subject.
[0187] Provided herein is a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction indicating a probability that an epitope sequence of the given peptide sequence will be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model comprising a plurality of predictor variables identified based on at least training data, the training data comprising: sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; training peptide sequence information comprising amino acid position information, the training peptide sequence information being associated with the HLA proteins expressed in the cell; and a function expressing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables; and determining or predicting that at least one of the peptide sequences of the polypeptide sequence is immunogenic to the subject based on the set of presentation predictions.
[0188] Provided herein is a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: obtaining, by a computer processor, a plurality of peptide sequences of the polypeptide sequence; processing, by the computer processor, amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate presentation predictions for each of the plurality of peptide sequences, wherein each presentation prediction indicates a likelihood that an epitope sequence of a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information associated with HLA proteins expressed in the cell; determining or predicting that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject based on the plurality of presentation predictions; and administering to the subject a composition comprising the drug.
[0189] In some embodiments, the method further comprises determining not to administer the drug to the subject.
[0190] In some embodiments, the drug comprises an antibody or a binding fragment thereof.
[0191] In some embodiments, the peptide sequence of the polypeptide sequence has a length of 8, 9, 10, 11, or 12 amino acids, and the protein encoded by a class I or class II MHC allele of the subject's cell is a protein encoded by a class I MHC allele of the subject's cell.
[0192] 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 the protein encoded by a class I or class II MHC allele of the subject's cell is a protein encoded by a class II MHC allele of the subject's cell.
[0193] Provided herein are methods of treating a subject having an autoimmune disease or condition, the methods comprising: (a) identifying or predicting an epitope of an expressed protein presented by class I or class II MHC on a cell of the subject, wherein a complex comprising the identified or predicted epitope and class I or class II MHC is targeted to CD8 T cells 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.
[0194] In some embodiments, the autoimmune disease or condition is diabetes.
[0195] In some embodiments, the cells are pancreatic islet cells.
[0196] Provided herein is a method of treating a subject having an autoimmune disease or condition, the method comprising administering to the subject a regulatory T cell that expresses a T cell receptor (TCR) that binds to a complex comprising (i) an epitope of an expressed protein identified or predicted to be presented by class I or class II MHC on cells of the subject, and (ii) class I or class II MHC, the complex being targeted to CD8 T cells or CD4 T cells of the subject.
[0197] Provided herein is a computer system for identifying peptide sequences for personalized cancer treatment of a subject, the computer system including: a database configured to store a plurality of peptide sequences of the subject; and one or more computer processors operably coupled to the database, the one or more computer processors 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, wherein each presentation prediction indicates a likelihood that a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data including sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; and select a subset of the plurality of peptide sequences for personalized cancer treatment of the subject based on at least the plurality of presentation predictions.
[0198] 1. A computer system for identifying HLA class II specific peptides for immunotherapy of a subject, comprising: a database configured to store candidate peptides comprising an epitope and a plurality of peptide sequences each comprising the epitope; and one or more computer processors operatively coupled to said database, processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences to an immune cell, each presentation prediction indicating a likelihood that a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by an HLA class II allele, the machine learning HLA-peptide presentation prediction model comprising a training model comprising sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry. Provided herein is a computer system comprising one or more computer processors individually and collectively programmed to: train the model using training data; select a protein predicted by the machine learning HLA-peptide presentation prediction model to bind to a candidate peptide from one or more proteins encoded by HLA class II alleles of a subject's cell, the protein having a probability greater than a threshold presentation predicted probability value for presentation of the candidate peptide to an immune cell; and identify the candidate peptide as a peptide for immunotherapy specific to the selected protein based on whether the candidate peptide replaces the placeholder peptide when the candidate peptide is allowed to contact the selected protein and thus compete with the placeholder peptide associated with the selected protein.
[0199] Provided herein is a computer system for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the computer system comprising: a database configured to store a plurality of peptide sequences of the polypeptide sequence; and one or more computer processors operably coupled to the database, the one or more computer processors 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, wherein each presentation prediction indicates a likelihood that an epitope sequence of a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information associated with HLA proteins expressed in the cell; and determine or predict that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject based on the plurality of presentation predictions, administering a composition comprising the drug to the subject.
[0200] Provided herein is a computer system for screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the computer system comprising: a database configured to store a plurality of peptide sequences of the polypeptide sequence; and one or more computer processors operably coupled to the database, the computer system 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, wherein each presentation prediction indicates a likelihood that an epitope sequence of a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; and determining or predicting that at least one of the plurality of peptide sequences of the polypeptide sequence is immunogenic to the subject based on the plurality of presentation predictions.
[0201] Provided herein is a non-transitory computer readable medium comprising machine executable code that, when executed by one or more computer processors, implements a method for identifying peptide sequences for personalized cancer treatment of a subject, the method comprising: obtaining a plurality of peptide sequences 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, wherein each presentation prediction indicates a likelihood that a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; and selecting a subset of the plurality of peptide sequences for personalized cancer treatment of the subject based on at least the plurality of presentation predictions.
[0202] A method for identifying HLA class II specific peptides for immunotherapy to a subject, when executed by one or more computer processors, includes the steps of obtaining candidate peptides comprising an epitope and a plurality of peptide sequences, each of which comprises the epitope; and processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate presentation predictions for each of the plurality of peptide sequences to an immune cell, each presentation prediction indicating a likelihood that a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by an HLA class II allele, the machine learning HLA-peptide presentation prediction model using training data including sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry. Provided herein is a non-transitory computer readable medium comprising machine-executable code that implements a method comprising: training a machine-learning HLA-peptide presentation prediction model to predict binding to a candidate peptide from one or more proteins encoded by HLA class II alleles of a cell of a subject, the protein having a probability greater than a threshold presentation predicted probability value for presentation of the candidate peptide to an immune cell; contacting the candidate peptide with the selected protein, thereby identifying the candidate peptide as a peptide for immunotherapy specific for the selected protein based on whether the candidate peptide replaces the placeholder peptide when the candidate peptide is in competition with the placeholder peptide.
[0203] Provided herein is a non-transitory computer readable medium comprising machine executable code that, when executed by one or more computer processors, implements a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: obtaining a plurality of peptide sequences of the polypeptide sequence; processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, wherein each presentation prediction indicates a likelihood that an epitope sequence of a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information associated with HLA proteins expressed in the cell; and determining or predicting that each of the plurality of peptide sequences of the polypeptide sequence is not immunogenic to the subject based on the plurality of presentation predictions; administering a composition comprising the drug to the subject.
[0204] Provided herein is a non-transitory computer readable medium comprising machine executable code that, when executed by one or more computer processors, implements a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: obtaining a plurality of peptide sequences of the polypeptide sequence; processing amino acid information of the plurality of peptide sequences using a machine learning HLA-peptide presentation prediction model to generate a presentation prediction for each of the plurality of peptide sequences, wherein each presentation prediction indicates a likelihood that an epitope sequence of a given peptide sequence of the plurality of peptide sequences may be presented by one or more proteins encoded by class I or class II MHC alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model being trained using training data comprising sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; and determining or predicting that at least one of the plurality of peptide sequences of the polypeptide sequence is immunogenic to the subject based on the plurality of presentation predictions.
[0205] the machine learning HLA peptide presentation prediction model is trained using training data including sequence information for sequences of training peptides identified by mass spectrometry to be presented by HLA proteins expressed in the training cells; and processing the amino acid information of the plurality of candidate peptide sequences using a machine learning HLA peptide presentation prediction model to generate a plurality of presentation predictions, each candidate peptide sequence of the plurality being encoded by the genome or exome of the subject, the plurality of presentation predictions including an HLA presentation prediction for each of the plurality of candidate peptide sequences, each presentation prediction indicating a likelihood that a given candidate peptide sequence of the plurality may be presented by one or more proteins encoded by class II HLA alleles of the cell of the subject, the machine learning HLA peptide presentation prediction model being trained using training data including sequence information for sequences of training peptides identified by mass spectrometry to be presented by HLA proteins expressed in the training cells; and identifying, based on at least the plurality of presentation predictions, peptide sequences of the plurality of peptide sequences that have a probability of being presented by at least one of the one or more proteins encoded by class II HLA alleles of the cell of the subject that is greater than a presentation prediction probability value threshold.
[0013] Provided herein is a method, wherein the machine learning HLA peptide presentation prediction model has a positive predictive value (PPV) of at least 0.07 when indicating a likelihood that a given test peptide sequence of the plurality of test peptide sequences may be presented by one or more proteins encoded by HLA alleles, the plurality of test peptide sequences comprises at least 500 test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in a cell, and (ii) at least 499 decoy peptide sequences contained within a protein encoded by the genome of the organism, wherein the organism and the subject are of the same species, the plurality of test peptide sequences comprises at least one hit peptide sequence and at least 499 decoy peptide sequences in a 1:499 ratio, and the machine learning HLA peptide presentation prediction model predicts that 0.2% of the plurality of test peptide sequences are presented by an HLA protein expressed in the cell.
[0206] the machine learning HLA-peptide binding prediction model is trained using training data including sequence information for sequences of peptides identified to bind to HLA class II proteins or HLA class II protein analogs; and processing the amino acid information of a plurality of peptide sequences encoded by the genome or exome of the subject using a machine learning HLA-peptide binding prediction model to generate a plurality of binding predictions, the plurality of binding predictions including an HLA binding prediction for each of a plurality of candidate peptide sequences, each binding prediction indicating a likelihood that one or more proteins encoded by class II HLA alleles of the cell of the subject will bind to a given candidate peptide sequence of the plurality of candidate peptide sequences, the machine learning HLA peptide binding prediction model being trained using training data including sequence information for sequences of peptides identified to bind to HLA class II proteins or HLA class II protein analogs; and identifying, based on at least the plurality of binding predictions, peptide sequences among the plurality of peptide sequences that have a probability greater than a threshold binding prediction probability value for binding to at least one of the one or more proteins encoded by class II HLA alleles of the cell of the subject; Provided herein is a method, in which a machine learning HLA peptide binding prediction model has a positive predictive value (PPV) of at least 0.1 when indicating a likelihood that one or more proteins encoded by HLA alleles bind to a given test peptide sequence of the plurality of test peptide sequences, the plurality of test peptide sequences comprises at least 50 test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry to be presented by an HLA protein expressed in the cell, and (ii) at least 19 decoy peptide sequences contained in a protein comprising the peptide sequence identified by mass spectrometry to be presented by an HLA protein expressed in the cell, the organism and the subject are of the same species, the plurality of test peptide sequences comprises at least one hit peptide sequence and at least 19 decoy peptide sequences in a 1:19 ratio, and the machine learning HLA peptide presentation prediction model predicts that 5% of the plurality of test peptide sequences bind to an HLA protein expressed in the cell.
[0207] In some embodiments, the machine learning HLA peptide presentation prediction model is trained using training data that includes sequence information for sequences of training peptides identified by mass spectrometry to be presented by HLA proteins expressed in training cells.
[0208] In some embodiments, one or more of 0.2% of a plurality of test peptide sequences predicted to be presented by a machine learning HLA peptide presentation prediction model have a higher probability of being presented by at least one of one or more proteins encoded by class II HLA alleles of a subject's cell than a threshold presentation prediction probability value.
[0209] In some embodiments, 0.2% of a plurality of test peptide sequences predicted to be presented by the machine learning HLA peptide presentation prediction model each have a probability of being presented by at least one of the one or more proteins encoded by class II HLA alleles of the subject's cell that is greater than a threshold presentation prediction probability value.
[0210] In some embodiments, the PPV is greater than the respective PPV in the second column of Table 11 for the protein encoded by the corresponding HLA allele in Table 13. In some embodiments, the PPV is at least equal to the respective PPV in the third column of Table 11 for the protein encoded by the corresponding HLA allele in Table 11.
[0211] In some embodiments, the PPV is greater than each of the PPVs in column 2 of Table 12 for proteins encoded by HLA class II alleles.
[0212] In some embodiments, the PPV is at least equal to each of the PPVs in column 2 of Table 16 for the proteins encoded by the corresponding HLA alleles in Table 16.
[0213] 1. A method for preparing a personalized cancer therapy, comprising the steps of: identifying peptide sequences associated with cancer, the step comprising comparing DNA, RNA or protein sequences from cancer cells of a subject with DNA, RNA or protein sequences from normal cells of the subject; and inputting amino acid position information of the identified peptide sequences into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the identified peptide sequences, each presentation prediction representing a probability that one or more proteins encoded by HLA class II alleles of the subject's cells will present a given sequence among the identified peptide sequences, the machine learning HLA-peptide presentation prediction model generating a set of presentation predictions for the identified peptide sequences, the machine learning HLA-peptide presentation prediction model generating a set of presentation predictions for the identified peptide sequences based on at least the training data. Provided herein is a method comprising the steps of: selecting a subset of peptide sequences for preparing a personalized cancer treatment identified based on a set of presentation predictions, the subset of peptide sequences including measurement variables, the training data including 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, the training peptide sequence information associated with HLA proteins expressed in cells; and a function expressing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables; wherein the prediction model has a positive predictive value of at least 0.1%, 0.1%-50%, or at most 50% recall.
[0214] Provided herein is a method for training a machine learning HLA-peptide presentation prediction model, the method comprising: inputting, using a computer processor, amino acid position information sequences of HLA-peptides isolated from one or more HLA-peptide complexes derived from cells expressing HLA class II alleles into the HLA-peptide presentation prediction model, wherein the machine learning HLA-peptide presentation prediction model comprises a plurality of predictor variables identified based on at least the training data, the plurality of predictor variables including: sequence information of sequences of peptides presented by HLA proteins expressed in the cells and identified by mass spectrometry; training peptide sequence information including amino acid position information of the training peptides, the training peptide sequence information being associated with the HLA proteins expressed in the cells; and a function expressing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and the predictor variables.
[0215] In some embodiments, the positive predictive value of the presented model is at least 0.25 with a recall rate of at least 0.1%, 0.1%-50%, or up to 50%.
[0216] In some embodiments, the positive predictive value of the presented model is at least 0.4 with a recall rate of at least 0.1%, between 0.1% and 50%, or up to 50%.
[0217] In some embodiments, the positive predictive value of the presented model is at least 0.6 with a recall rate of at least 0.1%, 0.1%-50%, or up to 50%.
[0218] In some embodiments, the mass spectrometry is single-allelic mass spectrometry.
[0219] In some embodiments, peptides are presented by HLA proteins expressed in cells through autophagy.
[0220] In some embodiments, peptides are presented by HLA proteins expressed in cells through phagocytosis.
[0221] In some embodiments, the quality of the training data is improved using multiple quality metrics.
[0222] In some embodiments, the multiple quality metrics include general contaminating peptide removal, high scored peak intensity, high score, and high mass accuracy.
[0223] In some embodiments, the scored peak intensity is at least 50%.
[0224] In some embodiments, the scored peak intensity is at least 60%.
[0225] In some embodiments, the score is at least 7.
[0226] In some embodiments, the mass accuracy is up to 5 ppm.
[0227] In some embodiments, the mass accuracy is up to 2 ppm.
[0228] In some embodiments, the skeletal cleavage score is at least 5.
[0229] In some embodiments, the skeletal cleavage score is at least 8.
[0230] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single immunoprecipitated HLA protein expressed in a cell.
[0231] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single exogenous HLA protein expressed in a cell.
[0232] In some embodiments, the peptides presented by an HLA protein expressed in a cell are peptides presented by a single recombinant HLA protein expressed in a cell.
[0233] In some embodiments, the plurality of predictor variables comprises a peptide-HLA affinity predictor variable.
[0234] In some embodiments, the plurality of predictor variables comprises a source protein expression level predictor variable.
[0235] In some embodiments, the plurality of predictors comprises a peptide cleavability predictor.
[0236] In some embodiments, the training peptide sequence information comprises sequences derived from peptides presented by HLA proteins, including peptides identified by searching peptide databases without enzyme specificity or modification. In some embodiments, the peptides presented by HLA proteins comprise peptides identified by searching novel peptide sequencing tools.
[0237] In some embodiments, peptides presented by HLA proteins include peptides identified by searching peptide databases using a reverse database searching strategy.
[0238] In some embodiments, the HLA protein comprises an HLA-DR protein and an HLA-DP protein or an HLA-DQ protein. In some embodiments, the HLA protein comprises an HLA-DR protein selected from the group consisting of an HLA-DR protein and an HLA-DP protein or an HLA-DQ protein. In some embodiments, the HLA protein comprises an HLA-DR protein selected from the group consisting of an HLA-DR protein and an HLA-DP protein or an HLA-DQ protein. * 01:01 / HLA-DPA1 * 01:03, HLA-DPB1 * 02:01 / HLA-DPA1 * 01:03, HLA-DPB1 *03:01 / HLA-DPA1 * 01:03、HLA-DPB1 * 04:01 / HLA-DPA1 * 01:03、HLA-DPB1 * 04:02 / HLA-DPA1 * 01:03、HLA-DPB1 * 06:01 / HLA-DPA1 * 01:03、HLA-DQB1 * 02:01 / HLA-DQA1 * 05:01、HLA-DQB1 * 02:02 / HLA-DQA1 * 02:01、HLA-DQB1 * 06:02 / HLA-DQA1 * 01:02、HLA-DQB1 * 06:04 / HLA-DQA1 * 01:02、HLA-DRB1 * 01:01、HLA-DRB1 * 01:02、HLA-DRB1 * 03:01、HLA-DRB1 * 03:02、HLA-DRB1 * 04:01、HLA-DRB1 * 04:02、HLA-DRB1 * 04:03、HLA-DRB1 * 04:04、HLA-DRB1 * 04:05、HLA-DRB1 * 04:07、HLA-DRB1 * 07:01、HLA-DRB1 * 08:01、HLA-DRB1 * 08:02、HLA-DRB1 * 08:03、HLA-DRB1 * 08:04、HLA-DRB1 * 09:01、HLA-DRB1 * 10:01、HLA-DRB1 * 11:01、HLA-DRB1 * 11:02、HLA-DRB1 * 11:04、HLA-DRB1 * 12:01、HLA-DRB1 *12:02, HLA-DRB1 * 13:01, HLA-DRB1 * 13:02, HLA-DRB1 * 13:03, HLA-DRB1 * 14:01, HLA-DRB1 * 15:01, HLA-DRB1 * 15:02, HLA-DRB1 * 15:03, HLA-DRB1 * 16:01, HLA-DRB3 * 01:01, HLA-DRB3 * 02:02, HLA-DRB3 * 03:01, HLA-DRB4 * 01:01 and HLA-DRB5 * 01:01.
[0239] In some embodiments, peptides presented by HLA proteins include peptides identified by comparing the MS / MS spectrum of an HLA-peptide with the MS / MS spectrum of one or more HLA-peptides in a peptide database.
[0240] In some embodiments, the mutation is selected from the group consisting of a point mutation, a splice site mutation, a frameshift mutation, a readthrough mutation, and a gene fusion mutation.
[0241] In some embodiments, the peptides presented by the HLA proteins have a length of 15-40 amino acids.
[0242] In some embodiments, peptides presented by HLA proteins include peptides identified by the steps of: (a) isolating one or more HLA complexes from a cell line expressing a single HLA class II allele; (b) isolating one or more HLA-peptides from the one or more isolated HLA complexes; (c) obtaining MS / MS spectra for the one or more isolated HLA-peptides; and (d) obtaining peptide sequences corresponding to the MS / MS spectra of the one or more isolated HLA-peptides from a peptide database, wherein the one or more sequences obtained in step (d) identify the sequence of the one or more isolated HLA-peptides.
[0243] In some embodiments, the personalized cancer treatment further comprises an adjuvant.
[0244] In some embodiments, the personalized cancer treatment further comprises an immune checkpoint inhibitor.
[0245] In some embodiments, the training data includes structured data, time series data, unstructured data, relational data, or any combination thereof.
[0246] In some embodiments, the unstructured data includes image data.
[0247] In some embodiments, the relational data includes data from customer systems, enterprise systems, operational systems, websites, web-accessible application program interfaces (APIs), or any combination thereof.
[0248] In some embodiments, the training data is uploaded to a cloud-based database.
[0249] In some embodiments, the training step is performed using a convolutional neural network.
[0250] In some embodiments, the convolutional neural network includes at least two convolutional layers.
[0251] In some embodiments, the convolutional neural network (CNN) includes at least one batch normalization step.
[0252] In some embodiments, the convolutional neural network includes at least one spatial dropout step.
[0253] In some embodiments, the convolutional neural network includes at least one global max pooling step.
[0254] In some embodiments, the convolutional neural network includes at least one dense layer.
[0255] In some embodiments, the step of identifying a peptide sequence comprises identifying a peptide sequence with a mutation expressed in cancer cells of the subject.
[0256] In some embodiments, the step of identifying a peptide sequence comprises identifying a peptide sequence that is not expressed in normal cells of the subject.
[0257] In some embodiments, the step of identifying peptide sequences comprises identifying overexpressed peptide sequences.
[0258] In some embodiments, identifying the peptide sequence comprises identifying a viral peptide sequence. In one aspect, a method for identifying HLA class II specific peptides for subject specific immunotherapy includes the steps of: identifying candidate peptides comprising an epitope; inputting amino acid information of a plurality of peptide sequences, each comprising the epitope, using a computer processor into a machine learning HLA-peptide presentation prediction model to generate a set of HLA presentation predictions of the peptide sequences to immune cells, each presentation prediction representing a probability that one or more proteins encoded by HLA class II alleles of the subject's cells will present a given peptide sequence comprising the epitope, the positive predictive value of the prediction model being at least 0.1%, between 0.1% and 50%, or at most 50% recall; and inputting amino acid information of a plurality of peptide sequences, each comprising the epitope, into a machine learning HLA-peptide presentation prediction model to generate a set of HLA presentation predictions of the peptide sequences to immune cells, each presentation prediction representing a probability that one or more proteins encoded by HLA class II alleles of the subject's cells will present a given peptide sequence comprising the epitope, the positive predictive value of the prediction model being at least 0.1, between 0.1% and 50%, or at most 50% recall. Provided herein is a method comprising the steps of: selecting a protein predicted by a predictive model to bind to a candidate peptide from one or more proteins that have a probability greater than a threshold value of a predicted presentation probability value for presentation of the candidate peptide to an immune cell; contacting the candidate peptide with a protein encoded by an HLA class II allele, thus causing the candidate peptide to compete with a placeholder peptide associated with the protein encoded by an HLA class II allele; and identifying the candidate peptide as a peptide for immunotherapy specific for the protein encoded by an HLA class II allele based on whether the candidate peptide replaces the placeholder peptide.
[0259] In some embodiments, the immunotherapy is cancer immunotherapy.
[0260] In some embodiments, the identifying step comprises 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. In some embodiments, the epitope is a cancer-specific epitope.
[0261] In some embodiments, the at least one protein encoded by the HLA class II allele comprises at least alpha 1 and beta 1 subunits of the HLA protein present in dimeric form or fragments thereof. In some embodiments, the placeholder peptide is a CLIP peptide. In some embodiments, the placeholder peptide is a CMV peptide. In some embodiments, the method further comprises measuring the IC50 of the replacement of the placeholder peptide with the target peptide. In some embodiments, the IC50 of the replacement of the placeholder peptide with the target peptide is less than 500 nM. In some embodiments, at least one protein from the one or more proteins encoded by the HLA class II alleles of the subject's 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 the HLA class II alleles of the subject's cell is a recombinant protein. In some embodiments, the at least one protein encoded by the HLA class II alleles of the subject's cell is expressed in a eukaryotic cell.
[0262] In one aspect, provided herein is an assay method for verifying the specificity of a candidate peptide for binding to an HLA class II protein, the method comprising the steps of expressing in a eukaryotic cell a polynucleic acid construct comprising a nucleic acid sequence encoding an HLA class II protein comprising an alpha chain and a beta chain or a portion thereof capable of binding to a peptide comprising an MHC-II binding epitope, wherein the expressed HLA class II protein or a portion thereof remains associated with a placeholder peptide; isolating the HLA class II protein or a portion thereof expressed in the eukaryotic cell; and performing a peptide exchange assay by (a) adding increasing amounts of the candidate peptide to determine whether the candidate peptide replaces the placeholder peptide associated with the HLA class II protein or a portion thereof, and (b) calculating the IC50 of the replacement reaction to determine the affinity of the candidate peptide for the HLA class II protein or a portion thereof compared to the placeholder peptide, thereby verifying the specificity of the candidate peptide for binding to the HLA class II protein.
[0263] 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.
[0264] In some embodiments, the peptide exchange assay comprises detection of a peptide fluorescent probe or tag, hi some embodiments, the placeholder peptide is a CLIP peptide.
[0265] In some embodiments, the polynucleic acid construct comprises an expression vector further comprising one or more of a promoter, a linker, one or more protease cleavage sites, a secretion signal, a dimerization factor, a ribosome skipping sequence, one or more tags for purification and / or detection.
[0266] In one aspect, provided herein is a method for assaying immunogenicity of an MHC class II binding peptide, comprising the steps of: selecting a protein encoded by an HLA class II allele 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%, 0.1%-50%, or at most 50% recall, and the protein has a probability greater than a threshold of a presentation predicted probability value for presentation of the identified peptide sequence; contacting the peptide with the protein encoded by the selected HLA class II allele, thus allowing the peptide to compete with a placeholder peptide associated with the protein encoded by the selected HLA class II allele and displacing the placeholder peptide, 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 assaying for one or more of the activation parameters of the CD4+ T cells selected from induction of cytokines, induction of chemokines, and expression of cell surface markers.
[0267] 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-gamma.
[0268] In one aspect, a method for inducing CD4+ T cell activation in a subject for cancer immunotherapy comprises the steps of: identifying a peptide sequence associated with cancer and comprising a cancer mutation, the step comprising comparing a DNA, RNA or protein sequence from a cancer cell of the subject with a DNA, RNA or protein sequence from a normal cell of the subject; and selecting a protein encoded by an HLA class II allele that is normally expressed by a cell of the subject and that is 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 with a recall of at least 0.1%, between 0.1% and 50%, or up to 50%. Provided herein is a method comprising the steps of: contacting the identified peptide with a protein encoded by a selected HLA class II allele, the protein having a probability greater than a threshold value of a predicted probability of presentation for the identified peptide sequence; verifying whether the identified peptide competes with and replaces a placeholder peptide associated with the protein encoded by the selected HLA class II allele with an IC50 value of less than 500 nM; purifying the identified peptide; and administering an effective amount of the identified peptide to a subject.
[0269] In one aspect, provided herein is a method for producing HLA class II tetramers or multimers, comprising expressing in a eukaryotic cell a vector comprising nucleic acid sequences encoding the alpha and beta chains of an HLA protein, a linker, one or more protease cleavage sites, a secretion signal, a biotinylation motif and at least one tag for identification or purification, thus secreting each HLA protein alpha1 and beta1 heterodimer in a dimeric form, where the heterodimer is associated with a placeholder peptide; purifying the secreted heterodimer from the cell culture medium; verifying peptide binding activity using a peptide exchange assay; adding streptavidin, thereby conjugating the heterodimer to a tetramer; purifying the tetramer; and obtaining a yield of greater than 1 mg / L.
[0270] In some embodiments, the vector comprises a CMV promoter. In some embodiments, the vector comprises a sequence encoding a placeholder peptide linked to the beta 1 chain via a cleavable site. In some embodiments, the peptide exchange assay involves pre-cleaving the placeholder peptide from the beta 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 by any one of column chromatography, batch chromatography, ion exchange chromatography, molecular sieve chromatography, affinity chromatography, or LC-MS.
[0271] In one aspect, provided herein is a composition comprising an HLA class II tetramer comprising either an HLA-DR heterodimer, or an HLA-DP heterodimer, or an HLA-DQ heterodimer, each heterodimer comprising an alpha chain and a beta chain, purified, and present at a concentration greater than 0.25 mg / L. In some embodiments, the HLA class II tetramer comprises a heterodimeric pair selected from the group consisting of an HLA-DR protein and a protein that can be selected from the group consisting of an HLA-DP protein or an HLA-DQ protein. In some embodiments, the HLA protein is 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) The snowflake is on the ground.
[0272] In some embodiments, the heterodimer pair is expressed in eukaryotic cells. In some embodiments, the heterodimer pair is encoded by a vector. In some embodiments, the vector comprises a nucleic acid sequence encoding the alpha and beta chains of HLA proteins, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, so that each HLA protein alpha1 and beta1 heterodimer is secreted in a dimerized state, and the secreted heterodimer is associated with a placeholder peptide. In some embodiments, the vector comprises a nucleic acid sequence encoding the alpha and beta chains of HLA proteins, a secretion signal, a biotinylation motif, and at least one tag for identification or purification, so that each HLA protein alpha1 and beta1 heterodimer is secreted in a dimerized state, and the secreted heterodimer is associated with a placeholder peptide.
[0273] In some embodiments, the HLA class II heterodimers are secreted from eukaryotic cells into cell culture medium and purified by any one of column or batch chromatography, ion exchange chromatography, molecular sieve chromatography, affinity chromatography, or LC-MS.
[0274] In one aspect, a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject includes the steps of inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that an epitope sequence of the given peptide sequence will be presented by one or more proteins encoded by HLA class I or class II alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model including a plurality of predictor variables identified based on at least training data, the training data including a plurality of predictor variables identified based on at least training data, the plurality of predictor variables including at least one predictor variable that is expressed in the cell. Provided herein is a method comprising: (a) receiving, based on a set of presentation predictions, sequence information of peptide sequences presented by HLA proteins expressed in a cell and identified by mass spectrometry; training peptide sequence information including amino acid position information, the training peptide sequence information being associated with HLA proteins expressed in a cell; and a function expressing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and predictor variables; (b) determining or predicting that each of the peptide sequences of the polypeptide sequence is not immunogenic to the subject based on the set of presentation predictions; and (c) administering to the subject a composition comprising a drug.
[0275] In one aspect, provided herein is a method of screening a drug comprising a polypeptide sequence for immunogenicity in a subject, the method comprising: (a) inputting amino acid information of a peptide sequence of the polypeptide sequence into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the peptide sequence, each presentation prediction representing a probability that an epitope sequence of the given peptide sequence will be presented by one or more proteins encoded by HLA class I or class II alleles of a cell of the subject, the machine learning HLA-peptide presentation prediction model including a plurality of predictor variables identified based on at least training data, the training data including: sequence information of sequences of peptides presented by HLA proteins expressed in the cell and identified by mass spectrometry; training peptide sequence information including amino acid position information, the training peptide sequence information associated with the HLA proteins expressed in the cell; and a function representing a 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 (b) determining or predicting that at least one of the peptide sequences of the polypeptide sequence is immunogenic to the subject based on the set of presentation predictions.
[0276] In one embodiment, the method further comprises deciding not to administer the drug to the subject.
[0277] In one embodiment, the drug comprises an antibody or a binding fragment thereof.
[0278] In one embodiment, the peptide sequences of the polypeptide sequence include each contiguous peptide sequence of the polypeptide sequence having a length of 8, 9, 10, 11 or 12 amino acids, and the protein encoded by an HLA class I or class II allele of the subject's cell is a protein encoded by an HLA class I allele of the subject's cell.
[0279] In one embodiment, the peptide sequences of the polypeptide sequence include each contiguous 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 the protein encoded by an HLA class I or class II allele of the subject's cell is a protein encoded by a class II MHC allele of the subject's cell.
[0280] In one aspect, provided herein is a method of treating a subject having an autoimmune disease or condition, the method comprising: (a) identifying or predicting an epitope of an expressed protein that is presented by HLA class I or class II on a cell of the subject, wherein a complex comprising the identified or predicted epitope and HLA class I or class II is targeted to CD8 T cells 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 from the subject or allogeneic regulatory T cells; and (d) administering the regulatory T cells expressing the TCR to the subject.
[0281] In one embodiment, the autoimmune disease or condition is diabetes.
[0282] In one embodiment, the cell is a pancreatic islet cell.
[0283] In one aspect, provided herein is a method of treating a subject having an autoimmune disease or condition, comprising administering to the subject a regulatory T cell that expresses a T cell receptor (TCR) that binds to a complex comprising (i) an epitope of an expressed protein identified or predicted to be presented by HLA class I or class II on cells of the subject, and (ii) HLA class I or class II, wherein the complex is targeted to CD8 T cells or CD4 T cells of the subject.
[0284] Further aspects and advantages of the present disclosure will become readily apparent to those skilled in the art from the following detailed description, in which merely exemplary embodiments of the present disclosure are shown and described. As will be understood, the present disclosure is capable of other and different embodiments, and its several details can be modified in various obvious respects, all without departing from the present disclosure. Accordingly, the drawings and description should be regarded as illustrative in nature, and not restrictive.
[0285] MAPTAC™ can be used for high throughput peptide binding assays, where LC-MS / MS is used to measure peptides bound to HLA class II after isolation with MAPTAC™ constructs at different time points and under different conditions, e.g., heating at 37° C., to obtain sequences of populations of peptides with different stability.
[0286] In one aspect, a method for treating cancer in a subject comprising the steps of identifying peptide sequences associated with the cancer, the step comprising comparing DNA, RNA or protein sequences from cancer cells of the subject to DNA, RNA or protein sequences from normal cells of the subject; and using a computer processor, inputting 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 a probability that one or more proteins encoded by HLA class II alleles of the subject's cells will present a given sequence of the identified peptide sequences, the machine learning HLA-peptide presentation prediction model comprising a plurality of predictor variables identified based on at least training data, the training data comprising: Provided herein is a method comprising the steps of: 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, the training peptide sequence information associated with HLA proteins expressed in cells; and a function expressing a relationship between the amino acid position information received as input and a presentation probability generated as output based on the amino acid position information and predictor variables; selecting a subset of peptide sequences for preparing a personalized cancer treatment identified based on the set of presentation predictions; and administering to a subject a composition comprising one or more of the peptides, wherein the predictive model has a positive predictive value of at least 0.1%, 0.1%-50%, or up to 50% recall.
[0287] In some embodiments, the machine learning HLA-peptide presentation predictive model comprises sequence information of sequences of peptides presented by HLA proteins expressed in cells and identified by mass spectrometry after performing reverse phase offline fractionation.
[0288] In some embodiments, the predictive model shows a 1.1-fold to 100-fold improvement over NetMHCIIpan. In some embodiments, the predictive model shows a 1.1-fold, 2-fold, 3-fold, 4-fold, 5-fold, 6-fold, 7-fold, 7.4-fold, 8-fold, 9-fold, 10-fold, 11-fold, 12-fold, 13-fold, 14-fold, 15-fold, 16-fold, 17-fold, 18-fold, 20-fold, 21-fold, 22-fold, 23-fold, 24-fold, 25-fold, 26-fold, 27-fold, 28-fold, 29-fold, 30-fold, 31-fold, 32-fold, 33-fold, 34-fold, 35-fold, 36-fold, 37-fold, 38-fold, 39-fold, 40-fold, 41-fold, 42-fold, 43x, 44x, 45x, 50x, 55x, 60x, 61x, 62x, 63x, 64x, 65x, 66x, 67x, 68x, 69x, 70x, 71x, 72x, 73x, 74x, 75x, 76x, 77x, 78x, 79x, 80x, 81x, 8x, 83x, 84x, 85x, 86x, 87x, 88x, 89x, 90x, 91x, 92x, 93x, 94x, 95x, 96x, 97x, 98x, 99x, 100x or greater improvement.
[0289] Incorporation by Reference All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference. To the extent that the publications and patents or patent applications incorporated by reference conflict with the disclosure contained herein, the present specification supersedes and / or takes precedence over any such conflicting material.
[0290] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the invention are utilized, and the accompanying drawings (also referred to herein as "drawings"), in which: [Brief description of the drawings]
[0291] [Figure 1A]FIG. 1A is a diagram depicting a peptide docked with an MHC class I protein.
[0292] [Figure 1B] FIG. 1B shows an exemplary diagram depicting a peptide docked with an MHC class II protein.
[0293] [Diagram 2] FIG. 2 shows an exemplary experimental approach for generating single-allele HLA class II binding peptide data. HLA class II peptides are introduced into any cell, including cells that do not express HLA class II, resulting in the expression of a specific HLA class II allele(s) in the cells. A population of engineered HLA-expressing cells is harvested, lysed, and their HLA-peptide complexes are tagged (e.g., biotinylated) and immunopurified (e.g., using biotin-streptavidin interactions). Single-HLA-specific HLA-associated peptides can be eluted from their tagged (e.g., biotinylated) complexes and evaluated (e.g., sequenced using high-resolution LC-MS / MS).
[0294] [Diagram 3] Figure 3 shows exemplary sequence logo representations of HLA class II-DRB1*11:01 associated peptides across Neon BAP, Expi293 cell line; Neon BAP, A375 cell line; IEDB, affinity <50 nM; and pan-HLA Class II Ab, homozygous LCL. Figure 3 shows that examples of MS-derived motifs match known patterns and show consistency across transfected cell lines.
[0295] [Figure 4]FIG. 4 is an exemplary depiction of the performance of HLA class II binding predictors. FIG. 4 is a bar plot showing the performance of the binding predictor (neonmhc2) and NetMHCIIpan applied to a validation dataset consisting of observed mass spectrometry peptides and decoy peptides generated at a ratio of 1:19 (hit:decoy) by randomly shuffling hit peptides. For the NEON binding predictor neonmhc2, a separate model is built for each MHC II allele shown. The height of the bar indicates the positive predictive value (PPV), defined as the fraction of predicted binders in the validation set that were actually hit peptides. Alleles are sorted by the performance of the model when predicting for that allele.
[0296] [Diagram 5] Figure 5 shows an exemplary effect of the scored peak intensity (SPI) threshold on the validation of binding predictors. Figure 5 shows the performance of HLA class II binding predictors when trained / validated on sets of peptides with different scored peak intensity (SPI) cutoffs. For each trained allele-specific model, the performance of the model is shown in three settings: training and evaluation on a data set using observed MS hit peptides greater than or equal to 70 SPI, training on peptides greater than or equal to 50 SPI and validating on peptides greater than or equal to 70 SPI, and training and validating on peptides greater than or equal to 50 SPI.
[0297] [Figure 6]Figure 6 shows an exemplary bar plot showing representative data from the number of peptides observed by allele profiling by LC-MS / MS that are greater than or equal to the 70 scored peak intensity (SPI) cutoff. Each bar represents the total number of observed peptides for an allele. There is collected data for 35 HLA-DR alleles. The collected data for the 35 HLA-DR alleles has >95% population coverage for HLA-DR (USA allele frequency).
[0298] [Figure 7A] Figure 7A shows the PPV of the model when applied to the test partition of data for the indicated HLA class II alleles. The decoy peptides used were scrambled sequences of the positive (hit) peptide sequences with a hit to decoy ratio of 1:19. The PPV was determined by identifying the top scoring 5% of peptides in the test partition and determining the fraction of them that were positive for binding to the protein encoded by each HLA class II allele.
[0299] [Figure 7B-1] Figure 7B shows an exemplary predictive performance as a function of training set size (curves obtained by artificially downsampling the training set). Figure 7B shows that, in general, for the 35 HLA-DR alleles collected, increasing the size of the training set increases the value of PPV. [Figure 7B-2] Figure 7B shows an exemplary predictive performance as a function of training set size (curves obtained by artificially downsampling the training set). Figure 7B shows that, in general, for the 35 HLA-DR alleles collected, increasing the size of the training set increases the value of PPV. [Figure 7B-3]Figure 7B shows an exemplary predictive performance as a function of training set size (curves obtained by artificially downsampling the training set). Figure 7B shows that, in general, for the 35 HLA-DR alleles collected, increasing the size of the training set increases the value of PPV.
[0300] [Figure 8] Figure 8 shows an exemplary graph demonstrating that processing-related variables can further improve prediction. Random sequences of peptides observed by MS selected from protein-coding exomes can be differentiated. For the training data partition, logistic regression can be fitted using binding strengths (NetMHCIIpan or Neon predictors) and processing features (RNA-Seq expression and derived gene-level bias terms) to predict HLA class II presentation. For separate evaluation partitions, exon positions overlapping with MHC II peptides observed by MS ("hits") can be scored in parallel with random exon positions not observed by MS (1:499 ratio). The top 0.2% (1 / 500) can be called positive, and positive predictive value can be evaluated using this threshold.
[0301] [Figure 9]FIG. 9 shows an exemplary neural network architecture. Input peptides are shown as 20-mers, and shorter peptides are filled in with a "missing" character. Each peptide has a 31-dimensional embedding, so the input to the neural network is a 20×31 matrix. Feature normalization for the 20×31 matrix is performed based on the mean and standard deviation of the feature values in the training set before processing by the neural network. The first convolutional layer has a kernel of 9 amino acids and 50 filters (also called channels), and uses a Rectified Linear Unit (ReLU) activation function. This is followed by batch normalization, then spatial dropout with a dropout rate of 20%. This is followed by another convolutional layer with a kernel of 3 amino acids and 20 filters, and uses a ReLU activation function, then batch normalization again and spatial dropout with a dropout rate of 20%. A global max pooling is then applied taking the maximally activated neuron in each of the 20 filters, and then these 20 values are passed through a fully connected (dense) layer with a single neuron using a sigmoid activation function. We treat the output of this layer as the joint / unjoint prediction.,L2 regularization is applied to the weights of the first, second, and dense,layers with weights of 0.05, 0.1, and 0.01, respectively.,Additional models used vary the number of,convolutional layers and the kernel size of each layer.
[0302] [Figure 10] FIG. 10 illustrates an exemplary computer control system programmed or otherwise configured to implement the methods provided herein.
[0303] [Figure 11A] FIG. 11A shows an exemplary overview of the MAPTAC™ experimental workflow.
[0304] [Figure 11B] FIG. 11B shows exemplary peptide counts per allele merged across replicate experiments.
[0305] [Figure 11C] FIG. 11C shows exemplary peptide length distributions for HLA class I and HLA class II alleles profiled by MAPTAC™.
[0306] [Figure 11D] FIG. 11D shows exemplary cysteine per residue frequencies observed for MAPTAC™ and IEDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01, and DRB1*11:01), the human proteome, and multi-allelic MS data from previous publications.
[0307] [Figure 12A] FIG. 12A shows the Caucasian frequencies for HLA-DR, HLA-DP, and HLA-DQ alleles present in >1% of individuals, and counts of peptides from the indicated sources measured as strong binders (<50 nM).
[0308] [Figure 12B] FIG. 12B shows an exemplary length distribution of IEDB peptides with associated HLA class II affinity measurements.
[0309] [Figure 12C]FIG. 12C shows an exemplary Western blot of (1) Expi293, (2) HeLa, and (3) A375 cell lines individually 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. Membranes were blotted with anti-biotin ligase epitope tag to visualize biotin acceptor peptide (BAP), and anti-beta-tubulin as a loading control. Lanes corresponded to the following fractions collected during the MAPTAC™ protocol: lane 1, input; lane 2, biotinylated input; and lane 3, input after pulldown.
[0310] [Figure 12D] FIG. 12D shows exemplary amino acid frequencies per residue observed for multi-allelic MS data from MAPTAC™ and IEDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01, and DRB1*11:01), the human proteome, and previous publications.
[0311] [Figure 12E] Figure 12E shows the Caucasian frequencies for HLA-DR, HLA-DP, and HLA-DQ alleles present in >1% of individuals, and counts of peptides from the indicated sources measured as strong binders (<50 nM). This figure contains additional data to Figure 12A. Additional data was obtained from tools.iedb.org / main / datasets / .
[0312] [Figure 12F] FIG. 12F shows exemplary amino acid frequencies per residue observed for MAPTAC™ (reduced and alkylated), MAPTAC™ (untreated) and IEDB (alleles DRB1*01:01, DRB1*03:01, DRB1*09:01, and DRB1*11:01), the human proteome, and multi-allelic MS data from previous publications.
[0313] [Figure 13] Figure 13 shows an exemplary display of core binding sequence logos for MAPTAC™ and MHC II alleles per IEDB. The sequence logos are diagrammatic representations, with the height of each amino acid proportional to the frequency of occurrence of that amino acid in peptides that bind to the MHC protein encoded by the allele. The positions with the lowest entropy are shown in color, which corresponds to the amino acid characteristics. The peptides are from the datasets shown and aligned according to a CNN-based predictor (Methods). The logos represent all peptides, including those that do not closely match the overall motif (e.g., no peptides are sequestered in the "trash" cluster).
[0314] [Figure 14A] FIG. 14A shows exemplary sequence logos for HLA-A*02:01 binding peptides (ligands) analyzed in two different cell lines (A375 & expi293) using different HLA-ligand profiling techniques including binding assays, stability assays, soluble HLA (sHLA) mass spectrometry, single allele mass spectrometry, and MAPTAC™.
[0315] [Figure 14B] FIG. 14B shows exemplary fractions of MAPTAC™ peptides exhibiting 0, 1, 2, 3, and 4 heuristically defined anchors.
[0316] [Figure 14C] FIG. 14C shows an exemplary distribution of MAPTAC™ observed peptides (20 peptides per allele, each with SPI>70 and nested sets of size ≥2) and NetMHCIIpan predicted binding affinities for length-matched decoys sampled from the proteome.
[0317] [Figure 15A] Figure 15A shows an exemplary architecture of a convolutional neural network (CNN) trained to distinguish between monoallelic MHC peptides and scrambled length-balanced decoys. The schematic shows the embedding of amino acid features, the use of two convolutional layers with different filter sizes, as well as the use of global max pooling as input to the final logistic output node.
[0318] [Figure 15B] Figure 15B is an exemplary result showing Kendall's tau statistic for IEDB affinity correlation measured with binding predictions from either neonmhc2 or NetMHCIIpan. Peptides evaluated include only those posted to the IEDB in the year following the publication of NetMHCIIpan.
[0319] [Figure 16] FIG. 16 is an exemplary depiction of the performance of neonmhc2 as a function of the size of the training dataset.
[0320] [Figure 17A] FIG. 17A shows an exemplary cluster assignment for MAPTAC™ peptides (20 per allele) spiked into pan-DR and pan-class II MHC MS datasets. The datasets were deconvolved using GibbsCluster. Each box represents one MAPTAC™ peptide. The color of the box indicates the cluster to which it is assigned, and the gray bar indicates the allele from which the peptide actually originates. The total number of clusters in the Gibbs cluster solution (right side) was selected using a mutual information (MI) metric. The MI score also determines how samples are sorted; samples with high MI solutions appear at the top.
[0321] [Figure 17B]Figure 17B shows an exemplary core binding sequence logo for multi-allelic MS data deconvoluted by GibbsCluster. Each set of peptides corresponds to the cluster that best aligns to the MAPTAC™ spike-in.
[0322] [Figure 17C] Figure 17C shows representative performance of models predicting hold-out MAPTAC peptides using either MAPTAC data or deconvoluted multi-allelic data. For each allele, the larger of the two data sources (usually MAPTAC) is downsampled, so the predictors are based on an equal number of training examples. NetMHCIIpan performance is shown as a further comparison.
[0323] [Figure 17D] FIG. 17D shows exemplary core binding sequence logos derived from multi-allelic MS data from the sources indicated.
[0324] [Figure 18A] FIG. 18A shows an exemplary graph of peptide vs. source gene expression fraction (transcripts per million (TPM) for peptides observed by MS and random proteome decoys) (data replotted from Schuster et al. 2017).
[0325] [Figure 18B] Figure 18B shows exemplary observed and expected numbers of class II peptides per gene as determined by joint analysis of colorectal cancer, melanoma, and ovarian cancer datasets (Loeffler et al., 2018, and Schuster et al., 2017). The expected count is derived by multiplying the length of the gene by the expression level. The expected and observed counts were summed across relevant samples. Genes known to be present in plasma are marked according to their concentration (inset).
[0326] [Figure 18C] FIG. 18C shows an exemplary distribution of enrichment scores (ratio of observed vs. expected observations, similar to FIG. 18B) for genes related to autophagy.
[0327] [Figure 18D] Figure 18D shows an exemplary distribution of enrichment scores according to the localization of each source gene. The localization of the source genes was determined using Uniprot (uniprot_sprot.dat).
[0328] [Figure 18E] FIG. 18E shows exemplary data depicting a comparison of the predicted and observed frequencies of the fraction of the total number of peptides with MHC-II binding affinity that were sequestered based on subcellular localization characteristics.
[0329] [Figure 18F] Figure 18F shows exemplary representative data of the relative concordance of peptides in observations of two different 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 concordance of each expression profile to the peptide repertoire. Error bars correspond to 95% confidence intervals computed by bootstrap resampling.
[0330] [Figure 19A] Figure 19A shows exemplary representative data of HLA-DRB1 expression levels in five study cases. Each dot represents expression in an individual cell type of an individual patient, averaged across cells.
[0331] [Figure 19B]Figure 19B shows exemplary representative data of tumor- and stroma-derived HLA-DRB1 expression input from RNA-Seq for TCGA patients. Bars correspond to individual patients, grouped by tumor type. Patients were included if they had a mutation in an HLA class II pathway gene (CIITA, CD74, or CTSSS) as determined by DNA-based mutation calling. 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 CIITA, CD74, or CTSS that show mutations.
[0332] [Figure 19C] FIG. 19C shows exemplary representative data from an additional single-cell RNA-Seq study including pre-checkpoint blockade immunotherapy and post-checkpoint blockade immunotherapy biopsies.
[0333] [Figure 20] FIG. 20 shows exemplary representative experimental data assessing the overall performance of predictions on native donor tissues.
[0334] [Figure 21A] Figure 21A shows exemplary representative data showing the integrated presentation model predicting the cellular HLA class II ligandome. This represents the PPV at a hit-to-decoy ratio of 1:499 for the pan-DR dataset (also analyzed in Figure 30B and Figure 32E). The predictor uses binding predictions (NetMHCIIpan or neonmhc2) and optionally gene expression, gene bias (as in Figure 32A), and overlap with previously observed HLA-DQ peptides. For each candidate peptide, the binding score was calculated as the maximum value across HLA-DR alleles in the sample genotype.
[0335] [Figure 21B]FIG. 21B shows exemplary representative data showing predictive performance with and without the use of processing features, using the same hit:decoy ratios and performance metrics as in FIG. 21A, for tumor-derived peptides identified using SILAC presented by dendritic cells (analyzed from cell lysates).
[0336] [Figure 21C] FIG. 21C shows exemplary expression and gene bias scores for heavily labeled peptides (red dots, plotted according to K562 expression) compared to lightly labeled peptides (gray dots, plotted according to DC expression) observed in UV treatment experiments.
[0337] [Figure 21D] Figure 21D shows an exemplary diagram depicting overlap of heavy-labeled peptide source genes from lysate and UV treatment experiments, with gene names colored by functional class.
[0338] [Figure 22A] FIG. 22A shows an exemplary flow diagram depicting the assay protocol disclosed herein for validating HLA class II-driven CD4+ T cells and T cell responses.
[0339] [Figure 22B] FIG. 22B shows the design of an exemplary HLA protein dimer construct for a peptide exchange assay (top panel) and a diagrammatic representation of an exemplary assay workflow (bottom panel).
[0340] [Diagram 23] FIG. 23 shows an exemplary diagram of an exemplary vector design for MHC-II expression to screen for new binding peptides, and a representation of the expressed protein product.
[0341] [Figure 24]FIG. 24 shows an exemplary flow diagram of the transfection, purification and cleavage of the placeholder peptide from the beta chain.
[0342] [Figure 25A] FIG. 25A shows an exemplary diagram illustrating a vector encoding a CLIP peptide that is associated with increased secretion of expressed MHC-II peptides.
[0343] [Figure 25B] FIG. 25B shows exemplary diagrammatic representations of shorter and longer forms of nucleic acids encoding CLIP0 and CLIP1, respectively.
[0344] [Figure 25C] FIG. 25C shows an exemplary representative result of Coomassie gel analysis of the alpha and beta chains with and without the longer clip.
[0345] [Figure 26A] FIG. 26A shows an exemplary diagram of a TR-FRET assay.
[0346] [Figure 26B] FIG. 26B shows exemplary representative polarization data from an HLA class II peptide binding assay using a fluorescence resonance energy transfer (FRET) assay with specific peptides.
[0347] [Figure 26C] FIG. 26C shows exemplary representative polarization data from an HLA class II peptide binding assay using a fluorescence resonance energy transfer (FRET) assay with specific peptides.
[0348] [Figure 26D] FIG. 26D shows an exemplary percent displacement of MHC-construct bound peptides calculated from the increase in fluorescence.
[0349] [Figure 26E] FIG. 26E shows an exemplary percent displacement of MHC-construct bound peptides calculated from the increase in fluorescence.
[0350] [Figure 26F] Figure 26F shows an exemplary peptide exchange using an assay using differential scanning fluorometry (DSF). A diagram showing an exemplary mechanism for detecting peptide dissociation from MHC class II using heat, which also dissociates MHC class II heterodimers, resulting in fluorophore binding and high fluorescence, is shown. An exemplary schematic of the displacement of a placeholder peptide by an epitope peptide is also shown. An exemplary melting curve plotted against temperature is also shown.
[0351] [Figure 26G] Figure 26G shows an exemplary soluble HLA-DM construct and its use to perform MHC class II peptide exchange. The construct shown contains a CMV promoter, coding sequences for the HLA-DM beta chain and the HLA-DM alpha chain downstream of a secretory sequence (leader), and a BAP sequence at the 3' end of the beta chain coding sequence; a His tag at the 3' end of the alpha chain coding sequence. The two chains are separated by an intervening ribosomal skipping sequence. The construct was expressed in Expi-CHO cells and the protein secreted into the culture medium was purified.
[0352] [Fig. 26H] FIG. 26H shows exemplary size exclusion chromatography data using HLA-sDM to perform peptide exchange.
[0353] [Figure 27A] FIG. 27A shows an exemplary diagram of constructing an exemplary DRB tetramer repertoire.
[0354] [Figure 27B] FIG. 27B shows an exemplary diagram of constructing an exemplary class II tetramer repertoire.
[0355] [Figure 27C] FIG. 27C shows an exemplary illustration of a summary of DRB tetramer repertoire coverage for DRB1 alleles for peptide exchange.
[0356] [Figure 27D] FIG. 27D shows an exemplary coverage of human MHC class II allele production.
[0357] [Figure 27E] FIG. 27E shows exemplary results from tetramer staining of samples induced with Flu epitopes (memory response) or HIV epitopes (naive response).
[0358] [Figure 28A] Figure 28A shows an exemplary diagrammatic representation of how peptides are assessed for HLA class II restriction by a fluorescence polarization assay that allows a screening method for the rapid identification of allelic restriction for epitope peptides. The assay principle of Figure 28A allows affinity measurements and unambiguous measurements of peptide exchange.
[0359] [Figure 28B] Figure 28B shows an exemplary summary of a number of assay conditions explored in a fluorescence polarization assay with DRB1*01:01 (top panel). Also shown is a diagram of a soluble MHC class II allele and a full-length MHC class II allele with a transmembrane domain in a detergent micelle (bottom panel), both of which were constructed with a placeholder peptide with a cleavable linker for use in the assay.
[0360] [Figure 28C]Figure 28C shows an exemplary diagrammatic representation of the assay for investigating full-length and soluble alleles previously shown in the bottom panel of Figure 28B. Briefly, both full-length and soluble alleles are expressed in cells. The membrane-bound form of the full-length allele is recovered by permeabilizing the membrane, while the secreted form is recovered from the cell supernatant. The recovered class II HLA allele protein is purified by passing through a nickel (Ni2+) column.
[0361] [Figure 28D] Figure 28D shows exemplary data showing that the purification method does not affect peptide potency. On the left are shown average IC50 values from experiments using full-length HLA-DR1 purified by L243 and full-length HLA-DR1 purified by Ni2+.
[0362] [Figure 28E] Figure 28E shows exemplary data showing that the choice of soluble form (sDR1) or full-length form (fDR1) does not affect peptide potency. On the left are shown average IC50 values from experiments using sDR1 forms or fDR1. FP, fluorescence polarization.
[0363] [Figure 28F] FIG. 28F shows an exemplary illustration of an exemplary evaluation of neonmhc2-predicted peptides and NetMHCIIpan-predicted peptides in a binding assay and identification of discordant peptides.
[0364] [Figure 28G] Figure 28G shows exemplary fluorescence polarization binding screening data for evaluating peptides predicted by neonmhc2; shown as a heat map and also as percent inhibition of probe binding shown for each concentration of peptide used. Green indicates good binding proportional to the intensity of the color. Yellow indicates intermediate binding and red indicates poor binding, also shown as the corresponding percent inhibition value.
[0365] [Fig. 28H] FIG. 28H shows a summary of the evaluation of peptides predicted by neonmhc2 in an exemplary binding assay.
[0366] [Figure 29] FIG. 29 shows exemplary mean counts of peptides from average MAPTAC™ experimental replicates (50 million cells) for each HLA allele.
[0367] [Fig. 30A-B]Figures 30A-30C show exemplary binding core analysis and multi-allelic deconvolution fidelity for HLA class II MAPTAC™ alleles + / - HLA-DM. Figure 30A shows one of the representative exemplary sequence logos for HLA-DR, HLA-DQ, and HLA-DP alleles according to MAPTAC™ with or 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 higher than 10% are colored according to their chemical properties; all other amino acids are gray. Peptides were aligned according to the GibbsCluster tool (Supplementary Methods), and the logos represent all peptides, including those that do not closely match the entire motif (e.g., no peptides are sequestered in the "trash" cluster). Figure 30B shows an exemplary depiction of cluster assignments for MAPTAC™ peptides (20 per allele) spiked into a pan-DR MS dataset. The dataset was deconvolved using GibbsCluster. Each colored box represents one MAPTAC™ peptide. The color of the box indicates the cluster to which it is assigned, and the gray bar indicates the allele from which the peptide originates. Figure 30C shows an exemplary graph showing the share of peptides showing 0, 1, 2, 3, or 4 predicted residues at the anchor positions for the alleles shown in Figure 30B. The anchor positions were defined as the four positions with the lowest entropy, and "predicted" residues were defined as those with a frequency of 10% or higher at those positions. [Figure 30C]Figures 30A-30C show exemplary binding core analysis and multi-allelic deconvolution fidelity for HLA class II MAPTAC™ alleles + / - HLA-DM. Figure 30A shows one of the representative exemplary sequence logos for HLA-DR, HLA-DQ, and HLA-DP alleles according to MAPTAC™ with or 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 higher than 10% are colored according to their chemical properties; all other amino acids are gray. Peptides were aligned according to the GibbsCluster tool (Supplementary Methods), and the logos represent all peptides, including those that do not closely match the entire motif (e.g., no peptides are sequestered in the "trash" cluster). Figure 30B shows an exemplary depiction of cluster assignments for MAPTAC™ peptides (20 per allele) spiked into a pan-DR MS dataset. The dataset was deconvolved using GibbsCluster. Each colored box represents one MAPTAC™ peptide. The color of the box indicates the cluster to which it is assigned, and the gray bar indicates the allele from which the peptide originates. Figure 30C shows an exemplary graph showing the share of peptides showing 0, 1, 2, 3, or 4 predicted residues at the anchor positions for the alleles shown in Figure 30B. The anchor positions were defined as the four positions with the lowest entropy, and "predicted" residues were defined as those with a frequency of 10% or higher at those positions.
[0368] [Figure 31A]Figures 31A-31F show an exemplary architecture and benchmarking of neonmhc2 binding prediction algorithms. Figure 31A shows an exemplary architecture of a convolutional neural network (CNN) trained to distinguish between monoallelic HLA class II peptides and scrambled, length-balanced decoys. The schematic shows an embedding layer of amino acid features, two convolutional layers of width 6, the presence of skip connections all the way to the end, and the use of a combination of average and max pooling operations as input to the final logistic output node. Figure 31B shows exemplary positive predictive values (PPV) for NetMHCIIpan and neonmhc2 evaluated on partitions of MAPTAC™ data that were not used for training or hyperparameter optimization. For each allele, n peptides observed by MS were scored together with 19n length-balanced decoys sampled from the same set of source genes, and the top-ranked peptides (e.g., top 5%) n of each predictor were called positive. This evaluation protocol ensures that the number of false positives and false negatives are equal, so PPV is the same as recall. Figure 31C shows exemplary NetMHCIIpan PPV and neonmhc2 PPV for the TGEM dataset. For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. Figure 31D shows exemplary ex vivo T cell induction results for neoantigen peptides. Peptides were selected based on high neonmhc2 scores and weak NetMHCIIpan scores for HLA-DRB1*11:01. Figure 31E shows a comparison of models trained on monoallelic MAPTAC data with deconvoluted multi-allelic data evaluated against hold-out monoallelic data. Values are as shown for neonmhc2 when the training dataset was downsampled to match the size of the deconvoluted training set.Figure 31F shows PPV for NetMHCIIpan-v3.1, a predictor trained by deconvolution, and neonmhc2 TGEM dataset (with and without downsampling). For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. [Figure 31B]Figures 31A-31F show an exemplary architecture and benchmarking of neonmhc2 binding prediction algorithms. Figure 31A shows an exemplary architecture of a convolutional neural network (CNN) trained to distinguish between monoallelic HLA class II peptides and scrambled, length-balanced decoys. The schematic shows an embedding layer of amino acid features, two convolutional layers of width 6, the presence of skip connections all the way to the end, and the use of a combination of average and max pooling operations as input to the final logistic output node. Figure 31B shows exemplary positive predictive values (PPV) for NetMHCIIpan and neonmhc2 evaluated on partitions of MAPTAC™ data that were not used for training or hyperparameter optimization. For each allele, n peptides observed by MS were scored together with 19n length-balanced decoys sampled from the same set of source genes, and the top-ranked peptides (e.g., top 5%) n of each predictor were called positive. This evaluation protocol ensures that the number of false positives and false negatives are equal, so PPV is the same as recall. Figure 31C shows exemplary NetMHCIIpan PPV and neonmhc2 PPV for the TGEM dataset. For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. Figure 31D shows exemplary ex vivo T cell induction results for neoantigen peptides. Peptides were selected based on high neonmhc2 scores and weak NetMHCIIpan scores for HLA-DRB1*11:01. Figure 31E shows a comparison of models trained on monoallelic MAPTAC data with deconvoluted multi-allelic data evaluated against hold-out monoallelic data. Values are as shown for neonmhc2 when the training dataset was downsampled to match the size of the deconvoluted training set.Figure 31F shows PPV for NetMHCIIpan-v3.1, a predictor trained by deconvolution, and neonmhc2 TGEM dataset (with and without downsampling). For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. [Figure 31C]Figures 31A-31F show an exemplary architecture and benchmarking of neonmhc2 binding prediction algorithms. Figure 31A shows an exemplary architecture of a convolutional neural network (CNN) trained to distinguish between monoallelic HLA class II peptides and scrambled, length-balanced decoys. The schematic shows an embedding layer of amino acid features, two convolutional layers of width 6, the presence of skip connections all the way to the end, and the use of a combination of average and max pooling operations as input to the final logistic output node. Figure 31B shows exemplary positive predictive values (PPV) for NetMHCIIpan and neonmhc2 evaluated on partitions of MAPTAC™ data that were not used for training or hyperparameter optimization. For each allele, n peptides observed by MS were scored together with 19n length-balanced decoys sampled from the same set of source genes, and the top-ranked peptides (e.g., top 5%) n of each predictor were called positive. This evaluation protocol ensures that the number of false positives and false negatives are equal, so PPV is the same as recall. Figure 31C shows exemplary NetMHCIIpan PPV and neonmhc2 PPV for the TGEM dataset. For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. Figure 31D shows exemplary ex vivo T cell induction results for neoantigen peptides. Peptides were selected based on high neonmhc2 scores and weak NetMHCIIpan scores for HLA-DRB1*11:01. Figure 31E shows a comparison of models trained on monoallelic MAPTAC data with deconvoluted multi-allelic data evaluated against hold-out monoallelic data. Values are as shown for neonmhc2 when the training dataset was downsampled to match the size of the deconvoluted training set.Figure 31F shows PPV for NetMHCIIpan-v3.1, a predictor trained by deconvolution, and neonmhc2 TGEM dataset (with and without downsampling). For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. [Figure 31D]Figures 31A-31F show an exemplary architecture and benchmarking of neonmhc2 binding prediction algorithms. Figure 31A shows an exemplary architecture of a convolutional neural network (CNN) trained to distinguish between monoallelic HLA class II peptides and scrambled, length-balanced decoys. The schematic shows an embedding layer of amino acid features, two convolutional layers of width 6, the presence of skip connections all the way to the end, and the use of a combination of average and max pooling operations as input to the final logistic output node. Figure 31B shows exemplary positive predictive values (PPV) for NetMHCIIpan and neonmhc2 evaluated on partitions of MAPTAC™ data that were not used for training or hyperparameter optimization. For each allele, n peptides observed by MS were scored together with 19n length-balanced decoys sampled from the same set of source genes, and the top-ranked peptides (e.g., top 5%) n of each predictor were called positive. This evaluation protocol ensures that the number of false positives and false negatives are equal, so PPV is the same as recall. Figure 31C shows exemplary NetMHCIIpan PPV and neonmhc2 PPV for the TGEM dataset. For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. Figure 31D shows exemplary ex vivo T cell induction results for neoantigen peptides. Peptides were selected based on high neonmhc2 scores and weak NetMHCIIpan scores for HLA-DRB1*11:01. Figure 31E shows a comparison of models trained on monoallelic MAPTAC data with deconvoluted multi-allelic data evaluated against hold-out monoallelic data. Values are as shown for neonmhc2 when the training dataset was downsampled to match the size of the deconvoluted training set.Figure 31F shows PPV for NetMHCIIpan-v3.1, a predictor trained by deconvolution, and neonmhc2 TGEM dataset (with and without downsampling). For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. [Fig. 31E-F]Figures 31A-31F show an exemplary architecture and benchmarking of neonmhc2 binding prediction algorithms. Figure 31A shows an exemplary architecture of a convolutional neural network (CNN) trained to distinguish between monoallelic HLA class II peptides and scrambled, length-balanced decoys. The schematic shows an embedding layer of amino acid features, two convolutional layers of width 6, the presence of skip connections all the way to the end, and the use of a combination of average and max pooling operations as input to the final logistic output node. Figure 31B shows exemplary positive predictive values (PPV) for NetMHCIIpan and neonmhc2 evaluated on partitions of MAPTAC™ data that were not used for training or hyperparameter optimization. For each allele, n peptides observed by MS were scored together with 19n length-balanced decoys sampled from the same set of source genes, and the top-ranked peptides (e.g., top 5%) n of each predictor were called positive. This evaluation protocol ensures that the number of false positives and false negatives are equal, so PPV is the same as recall. Figure 31C shows exemplary NetMHCIIpan PPV and neonmhc2 PPV for the TGEM dataset. For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set. Figure 31D shows exemplary ex vivo T cell induction results for neoantigen peptides. Peptides were selected based on high neonmhc2 scores and weak NetMHCIIpan scores for HLA-DRB1*11:01. Figure 31E shows a comparison of models trained on monoallelic MAPTAC data with deconvoluted multi-allelic data evaluated against hold-out monoallelic data. Values are as shown for neonmhc2 when the training dataset was downsampled to match the size of the deconvoluted training set.Figure 31F shows PPV for NetMHCIIpan-v3.1, a predictor trained by deconvolution, and neonmhc2 TGEM dataset (with and without downsampling). For each allele, the top ranked n peptides were called positive, where n is the number of confirmed immunogenic epitopes in the evaluated set.
[0369] [Fig. 32A-E]Figures 32A-32E show exemplary gene representation and protein processing in the HLA class II tumor peptidome. Figure 32A shows exemplary results of observed and expected number of HLA class II peptides per gene determined by joint analysis of colorectal cancer, melanoma, and ovarian cancer datasets. The expected count is derived by multiplying the length of the gene by the expression level. The expected and observed counts were summed across relevant samples. Genes known to be present in plasma were marked according to their concentration. Figure 32B shows exemplary results of predicted and observed frequency of peptides per subcellular localization. Figure 32C shows exemplary results of the distribution of enrichment scores for genes regulated by the proteasome (ratio of observed vs. expected observations, partly as in Figure 32B). Gene sets included those containing known ubiquitination sites and those whose abundance increased upon application of proteasome inhibitors. FIG. 32D shows a diagram illustrating three exemplary working models for how HLA class II peptides are processed: i) by cathepsin and other enzymes cleaving the protein into peptide fragments, which are then bound by HLA; ii) by proteins or unfolded polypeptides binding to HLA, which are then cleaved to peptide length; and iii) by proteins being partially digested, then binding, and then further trimming after binding. Each model corresponds to a different prediction approach. FIG. 32E shows the absolute increase in PPV observed for a logistic regression model including processing-related variables and neonmhc2 binding predictions, compared to a model using only binding predictions. Evaluation was performed on 11 samples profiled with HLA-DR antibodies (same samples as analyzed in FIG. 30B); each point corresponds to one sample. Asterisks indicate significant improvement according to a two-tailed paired t-test (*: p<0.01, **: p<0.001, ***: p<0.0001). The same analysis is shown in Figure 40B, but instead using NetMHCIIpan as the base predictor. The methods for decoy selection and PPV calculation are identical to those used in Figure 31B.
[0370] [Figure 33A-C]Figures 33A-33G show exemplary results of identification and prediction of tumor antigens presented by dendritic cells. Figure 33A shows an exemplary diagrammatic representation of the experimental workflow for identifying DC-presented HLA-II ligands derived from cancer cells (K562). Cancer cells were grown in SILAC medium for complete incorporation, lysed or irradiated, and then plated together with monocyte-derived dendritic cells. Presented peptides were isolated by pan-DR antibodies and sequenced by LC-MS / MS. Figure 33B shows exemplary data representing prediction performance using the same hit-to-decoy ratio and performance metrics as in Figure 21A for tumor-derived peptides presented by dendritic cells. Performance is shown for NetMHCIIpan-based and neonmhc2-based models with and without the use of processing features. Figure 33C shows an exemplary gene expression distribution for source genes of heavy-labeled peptides (red curve, plotted according to K562 expression) compared to source genes of light-labeled peptides (gray curve, plotted according to DC expression) observed in UV treatment experiments. Figure 33D shows an exemplary graph of PPV at a hit-to-decoy ratio of 1:499 for predicting presented tumor antigens with and without processing features using NetMHCIIpan-based and neonmhc2-based models. Data points represent, from left to right, samples: Donor 1 HOCl treated cells: NetMHCIIpan sequential expression; NetMHCIIpan sequential expression + gene bias; NetMHCIIpan sequential expression + gene bias + DQ overlap, full processing scheme; Donor 1, UV treated: neonmhc2; neonmhc2 + threshold expression; neonmhc2 + sequential expression; neonmhc2 + sequential expression + gene bias; neonmhc2 + sequential expression + gene bias + DQ overlap. Figure 33E shows the significance of different gene localizations and functional classes in predicting heavy (K562-derived) and light (DC-derived) peptides, respectively. P values are calculated according to a logistic regression controlling for neonmhc2 binding score and source gene expression. Bar colors indicate the signs associated with the coefficients in the regression.Figure 33F shows an exemplary graphical representation of results showing overlap of tumor cell-derived peptide source genes in UV-treated and HOCl-treated experiments (colored by functional class). Figure 33G shows exemplary data showing PPV for predicting tumor antigens presented in a second donor using a logistic model fitted to heavy labeled peptides observed in a first donor. Models were fitted using neonmhc2 binding alone; binding and expression; or binding, expression, and a binary variable indicating whether the peptide is derived from a mitochondrial gene. [Fig. 33D-E]Figures 33A-33G show exemplary results of identification and prediction of tumor antigens presented by dendritic cells. Figure 33A shows an exemplary diagrammatic representation of the experimental workflow for identifying DC-presented HLA-II ligands derived from cancer cells (K562). Cancer cells were grown in SILAC medium for complete incorporation, lysed or irradiated, and then plated together with monocyte-derived dendritic cells. Presented peptides were isolated by pan-DR antibodies and sequenced by LC-MS / MS. Figure 33B shows exemplary data representing prediction performance using the same hit-to-decoy ratio and performance metrics as in Figure 21A for tumor-derived peptides presented by dendritic cells. Performance is shown for NetMHCIIpan-based and neonmhc2-based models with and without the use of processing features. Figure 33C shows an exemplary gene expression distribution for source genes of heavy-labeled peptides (red curve, plotted according to K562 expression) compared to source genes of light-labeled peptides (gray curve, plotted according to DC expression) observed in UV treatment experiments. Figure 33D shows an exemplary graph of PPV at a hit-to-decoy ratio of 1:499 for predicting presented tumor antigens with and without processing features using NetMHCIIpan-based and neonmhc2-based models. Data points represent, from left to right, samples: Donor 1 HOCl treated cells: NetMHCIIpan sequential expression; NetMHCIIpan sequential expression + gene bias; NetMHCIIpan sequential expression + gene bias + DQ overlap, full processing scheme; Donor 1, UV treated: neonmhc2; neonmhc2 + threshold expression; neonmhc2 + sequential expression; neonmhc2 + sequential expression + gene bias; neonmhc2 + sequential expression + gene bias + DQ overlap. Figure 33E shows the significance of different gene localizations and functional classes in predicting heavy (K562-derived) and light (DC-derived) peptides, respectively. P values are calculated according to a logistic regression controlling for neonmhc2 binding score and source gene expression. Bar colors indicate the signs associated with the coefficients in the regression.Figure 33F shows an exemplary graphical representation of results showing overlap of tumor cell-derived peptide source genes in UV-treated and HOCl-treated experiments (colored by functional class). Figure 33G shows exemplary data showing PPV for predicting tumor antigens presented in a second donor using a logistic model fitted to heavy labeled peptides observed in a first donor. Models were fitted using neonmhc2 binding alone; binding and expression; or binding, expression, and a binary variable indicating whether the peptide is derived from a mitochondrial gene. [Fig. 33F-G]Figures 33A-33G show exemplary results of identification and prediction of tumor antigens presented by dendritic cells. Figure 33A shows an exemplary diagrammatic representation of the experimental workflow for identifying DC-presented HLA-II ligands derived from cancer cells (K562). Cancer cells were grown in SILAC medium for complete incorporation, lysed or irradiated, and then plated together with monocyte-derived dendritic cells. Presented peptides were isolated by pan-DR antibodies and sequenced by LC-MS / MS. Figure 33B shows exemplary data representing prediction performance using the same hit-to-decoy ratio and performance metrics as in Figure 21A for tumor-derived peptides presented by dendritic cells. Performance is shown for NetMHCIIpan-based and neonmhc2-based models with and without the use of processing features. Figure 33C shows an exemplary gene expression distribution for source genes of heavy-labeled peptides (red curve, plotted according to K562 expression) compared to source genes of light-labeled peptides (gray curve, plotted according to DC expression) observed in UV treatment experiments. Figure 33D shows an exemplary graph of PPV at a hit-to-decoy ratio of 1:499 for predicting presented tumor antigens with and without processing features using NetMHCIIpan-based and neonmhc2-based models. Data points represent, from left to right, samples: Donor 1 HOCl treated cells: NetMHCIIpan sequential expression; NetMHCIIpan sequential expression + gene bias; NetMHCIIpan sequential expression + gene bias + DQ overlap, full processing scheme; Donor 1, UV treated: neonmhc2; neonmhc2 + threshold expression; neonmhc2 + sequential expression; neonmhc2 + sequential expression + gene bias; neonmhc2 + sequential expression + gene bias + DQ overlap. Figure 33E shows the significance of different gene localizations and functional classes in predicting heavy (K562-derived) and light (DC-derived) peptides, respectively. P values are calculated according to a logistic regression controlling for neonmhc2 binding score and source gene expression. Bar colors indicate the signs associated with the coefficients in the regression.Figure 33F shows an exemplary graphical representation of results showing overlap of tumor cell-derived peptide source genes in UV-treated and HOCl-treated experiments (colored by functional class). Figure 33G shows exemplary data showing PPV for predicting tumor antigens presented in a second donor using a logistic model fitted to heavy labeled peptides observed in a first donor. Models were fitted using neonmhc2 binding alone; binding and expression; or binding, expression, and a binary variable indicating whether the peptide is derived from a mitochondrial gene.
[0371] [Fig. 34A-B] Figures 34A-34B show exemplary characterization of MAPTAC™ data related to Figure 29. Figure 34A shows exemplary HLA cell surface analysis by FACS of Expi293 cell lines transfected with MAPTAC™ constructs encoding affinity tagged HLA-A*02:01-BAP. Figure 34B shows exemplary HLA cell surface analysis by FACS of Expi293 cell lines transfected with MAPTAC™ constructs encoding affinity tagged HLA-DRB1*11:01-BAP (bottom). HLA cell surface expression of transfected Expi293 cells (orange) was compared to stained untransfected Expi293 (blue), unstained untransfected Expi293 (red), stained PBMCs (dark green), and unstained PBMCs (light green). W6 / 32 (pan-HLA class I) was used for all HLA class I staining, whereas REA332 (pan-HLA class II) was used for HLA class II staining.
[0372] [Diagram 35]Figure 35 shows an exemplary comparison of MAPTAC™ and IEDB logos, related to Figure 30A. Measured affinities and NetMHCIIpan predicted affinities for peptides observed by MS that did not show good NetMHCIIpan scores but were well supported by MS (scored peak intensity >70 and nested set size >1).
[0373] [Figure 36A] Figures 36A-36C show exemplary analyses of HLA-DR1 MAPTAC™ data fidelity, related to Figures 30A-30C. Figure 36A shows exemplary NetMHCIIpan3.1 scores for HLA-DR1 MAPTAC™ peptides (green) (length 12-23) for common alleles compared to 50,000 length-matched decoy peptides (blue) randomly sampled from the proteome. Figure 36B shows exemplary measured affinities and NetMHCIIpan predicted affinities for exemplary MS-observed peptides that do not show good NetMHCIIpan scores but are well supported in MS (scored peak intensity >70 and nested set size >1). Figure 36C shows exemplary HLA class II sequence logos for HLA-DRB1 alleles determined by MAPTAC™ in different cell types. [Figure 36B]Figures 36A-36C show exemplary analyses of HLA-DR1 MAPTAC™ data fidelity, related to Figures 30A-30C. Figure 36A shows exemplary NetMHCIIpan3.1 scores for HLA-DR1 MAPTAC™ peptides (green) (length 12-23) for common alleles compared to 50,000 length-matched decoy peptides (blue) randomly sampled from the proteome. Figure 36B shows exemplary measured affinities and NetMHCIIpan predicted affinities for exemplary MS-observed peptides that do not show good NetMHCIIpan scores but are well supported in MS (scored peak intensity >70 and nested set size >1). Figure 36C shows exemplary HLA class II sequence logos for HLA-DRB1 alleles determined by MAPTAC™ in different cell types. [Figure 36C] Figures 36A-36C show exemplary analyses of HLA-DR1 MAPTAC™ data fidelity, related to Figures 30A-30C. Figure 36A shows exemplary NetMHCIIpan3.1 scores for HLA-DR1 MAPTAC™ peptides (green) (length 12-23) for common alleles compared to 50,000 length-matched decoy peptides (blue) randomly sampled from the proteome. Figure 36B shows exemplary measured affinities and NetMHCIIpan predicted affinities for exemplary MS-observed peptides that do not show good NetMHCIIpan scores but are well supported in MS (scored peak intensity >70 and nested set size >1). Figure 36C shows exemplary HLA class II sequence logos for HLA-DRB1 alleles determined by MAPTAC™ in different cell types.
[0374] [Fig. 37A-B]Figures 37A-37C show additional exemplary analyses of MAPTAC™ motifs related to Figures 30A-30C. Figure 37A shows sequence logos derived from MAPTAC™ for experiments with and without HLA-DM co-transfection (expi293 cell line). Figure 37B shows sequence logos for MAPTAC™ and several HLA class I alleles according to IEDB. Note that A*32:01 does not show high frequency Q at P2 and C*03:03 does not show high frequency Y at P9, which differs from previous studies using multi-allelic deconvolution; the logo for B*52:01 has not been previously published. Figure 37C shows an exemplary alignment of peptides observed by MAPTAC™ to the gene sequence of CD74. [Figure 37C] Figures 37A-37C show additional exemplary analyses of MAPTAC™ motifs related to Figures 30A-30C. Figure 37A shows sequence logos derived from MAPTAC™ for experiments with and without HLA-DM co-transfection (expi293 cell line). Figure 37B shows sequence logos for MAPTAC™ and several HLA class I alleles according to IEDB. Note that A*32:01 does not show high frequency Q at P2 and C*03:03 does not show high frequency Y at P9, which differs from previous studies using multi-allelic deconvolution; the logo for B*52:01 has not been previously published. Figure 37C shows an exemplary alignment of peptides observed by MAPTAC™ to the gene sequence of CD74.
[0375] [Figure 38Ai]Figures 38Ai-38D show exemplary neonmhc2 performance statistics and T cell flow staining related to Figures 31A-31D. Figure 38Ai shows the performance of neonmhc2 as a function of the size of an exemplary training dataset. PPV was evaluated in the same manner as in Figure 31B and using the same evaluation peptides; however, the training data was randomly downsampled to mimic a smaller training dataset. Figure 38Aii shows an exemplary sequence logo for peptide clusters derived from the biallelic HLA-DR ligandome using GibbsCluster (default settings; "trash clusters allowed"). Figure 38B shows an exemplary representative flow cytometry plot of IFN-γ expression by CD4+ cells from induced samples re-recalled with neoantigen peptide prediction using neonmhc2. Delta values were calculated by subtracting the percent of CD4+ cells expressing IFN-γ when re-recalled with neoantigen (+peptide) from the percent of CD4+ expressing IFN-γ that were re-recalled in the absence of neoantigen (no peptide). The two flow plots on the left show the CD4+ Figure 38C shows an exemplary delta value from a well recalled with a single neonmhc2 neoantigen peptide. A peptide was considered an inducing hit if it had a positive response (above 3% delta response, highlighted). Figure 38D shows an exemplary sequence logo for a peptide cluster derived for a multi-allelic HLA-DR ligandome using GibbsCluster (default settings; "trash" cluster tolerance). [Figure 38Aii]Figures 38Ai-38D show exemplary neonmhc2 performance statistics and T cell flow staining related to Figures 31A-31D. Figure 38Ai shows the performance of neonmhc2 as a function of the size of an exemplary training dataset. PPV was evaluated in the same manner as in Figure 31B and using the same evaluation peptides; however, the training data was randomly downsampled to mimic a smaller training dataset. Figure 38Aii shows an exemplary sequence logo for peptide clusters derived from the biallelic HLA-DR ligandome using GibbsCluster (default settings; "trash clusters allowed"). Figure 38B shows an exemplary representative flow cytometry plot of IFN-γ expression by CD4+ cells from induced samples re-recalled with neoantigen peptide prediction using neonmhc2. Delta values were calculated by subtracting the percent of CD4+ cells expressing IFN-γ when re-recalled with neoantigen (+peptide) from the percent of CD4+ expressing IFN-γ that were re-recalled in the absence of neoantigen (no peptide). The two flow plots on the left show the CD4+ Figure 38C shows an exemplary delta value from a well recalled with a single neonmhc2 neoantigen peptide. A peptide was considered an inducing hit if it had a positive response (above 3% delta response, highlighted). Figure 38D shows an exemplary sequence logo for a peptide cluster derived for a multi-allelic HLA-DR ligandome using GibbsCluster (default settings; "trash" cluster tolerance). [Fig. 38B-C]Figures 38Ai-38D show exemplary neonmhc2 performance statistics and T cell flow staining related to Figures 31A-31D. Figure 38Ai shows the performance of neonmhc2 as a function of the size of an exemplary training dataset. PPV was evaluated in the same manner as in Figure 31B and using the same evaluation peptides; however, the training data was randomly downsampled to mimic a smaller training dataset. Figure 38Aii shows an exemplary sequence logo for peptide clusters derived from the biallelic HLA-DR ligandome using GibbsCluster (default settings; "trash clusters allowed"). Figure 38B shows an exemplary representative flow cytometry plot of IFN-γ expression by CD4+ cells from induced samples re-recalled with neoantigen peptide prediction using neonmhc2. Delta values were calculated by subtracting the percent of CD4+ cells expressing IFN-γ when re-recalled with neoantigen (+peptide) from the percent of CD4+ expressing IFN-γ that were re-recalled in the absence of neoantigen (no peptide). The two flow plots on the left show the CD4+ Figure 38C shows an exemplary delta value from a well recalled with a single neonmhc2 neoantigen peptide. A peptide was considered an inducing hit if it had a positive response (above 3% delta response, highlighted). Figure 38D shows an exemplary sequence logo for a peptide cluster derived for a multi-allelic HLA-DR ligandome using GibbsCluster (default settings; "trash" cluster tolerance). [Figure 38D]Figures 38Ai-38D show exemplary neonmhc2 performance statistics and T cell flow staining related to Figures 31A-31D. Figure 38Ai shows the performance of neonmhc2 as a function of the size of an exemplary training dataset. PPV was evaluated in the same manner as in Figure 31B and using the same evaluation peptides; however, the training data was randomly downsampled to mimic a smaller training dataset. Figure 38Aii shows an exemplary sequence logo for peptide clusters derived from the biallelic HLA-DR ligandome using GibbsCluster (default settings; "trash clusters allowed"). Figure 38B shows an exemplary representative flow cytometry plot of IFN-γ expression by CD4+ cells from induced samples re-recalled with neoantigen peptide prediction using neonmhc2. Delta values were calculated by subtracting the percent of CD4+ cells expressing IFN-γ when re-recalled with neoantigen (+peptide) from the percent of CD4+ expressing IFN-γ that were re-recalled in the absence of neoantigen (no peptide). The two flow plots on the left show the CD4+ Figure 38C shows an exemplary delta value from a well recalled with a single neonmhc2 neoantigen peptide. A peptide was considered an inducing hit if it had a positive response (above 3% delta response, highlighted). Figure 38D shows an exemplary sequence logo for a peptide cluster derived for a multi-allelic HLA-DR ligandome using GibbsCluster (default settings; "trash" cluster tolerance).
[0376] [Figure 39A-C]Figures 39A-39C show additional exemplary cell origin analysis for HLA class II, related to Figures 32A-32E. Figure 39A shows exemplary percentile neonmhc2 scores for HLA class II peptides observed in four PBMC samples profiled with pan-DR antibodies (RG1248, RG1104, RG1095, and HDSC in Figure 30B) according to whether the peptide source gene was present in human plasma. For each peptide, the best (worst) percentile across alleles present in the donor was used. Scores for random length-balanced proteome decoys are shown for comparison. Box plots show the 5th, 25th, 50th, 75th, and 95th percentiles. Figure 39B shows exemplary counts of observed and predicted peptides per gene for HLA class I using the same methodology as in Figure 32A. Data corresponds to the same tumor types (colorectal, ovarian, and melanoma). Genes present in human plasma are highlighted in blue and sized according to their concentration. Figure 39C shows the relative concordance of exemplary peptide observations for two different gene expression profiles. For each sample, gene-level peptide counts were modeled as a linear combination of bulk tumor gene expression and professional APC gene expression profiles. The ratio of coefficients determines the relative concordance with the peptide repertoire of each expression profile. Error bars correspond to 95% confidence intervals computed by bootstrap resampling.
[0377] [Fig. 40A-B]Figures 40A-40B show additional exemplary analyses of processing motifs related to Figures 32A-32E. Figure 40A shows exemplary amino acid frequencies near the N- and C-terminal peptide cuts relative to the average proteome frequency (applying upstream positions U3-U1 and downstream positions D1-D3) or to the average peptide frequency (applying internal positions N1-C1) observed in donor PBMCs, monocyte-derived dendritic cells, colorectal cancer, melanoma, ovarian cancer, and expi293 cell lines (used for the majority of MAPTAC™ data generation). Figure 40B shows the same analysis as in Figure 32E, but using NetMHCIIpan as the primary predictor. For eight samples profiled with HLA-DR antibodies (the same samples analyzed in Figure 31B), an absolute increase in PPV was observed for logistic regression models that included variables related to processing in addition to predictions by NetMHCIIpan (compared to models with NetMHCIIpan only). Asterisks indicate significant improvement (*: p<0.01, **: p<0.001, ***: p<0.0001) according to two-tailed paired t-test.
[0378] [Diagram 41] FIG. 41 shows an exemplary nomenclature system used to refer to positions upstream, within, and downstream of a peptide.
[0379] [Figure 42A] FIG. 42A shows a diagram depicting an exemplary workflow for the analysis of endogenously processed and HLA-1 and HLA class II presented peptides by nLC-MS / MS.
[0380] [Figure 42B] Figure 42B shows a graph depicting exemplary experimental results from nLC-MS / MS analysis of tryptic peptides with or without FAIMS. Representative overlap in detection of HLA-I and HLA class II peptides by nLC-MS / MS analysis with or without FAIMS at the analytical scales shown is also shown.
[0381] [Figure 43A] FIG. 43A shows exemplary HLA class I acidic and basic reverse phase fractionation peptide detection with and without FAIMS.
[0382] [Figure 43B] FIG. 43B shows exemplary experimental results showing detection of unique peptides bound to HLA class I plotted against retention time.
[0383] [Figure 44A] FIG. 44A shows exemplary HLA class II acidic and basic reverse phase fractionation peptide detection with and without FAIMS.
[0384] [Figure 44B] FIG. 44B shows exemplary experimental results showing detection of unique peptides bound to HLA class II plotted against retention time.
[0385] [Figure 45-1] FIG. 45 shows an exemplary graph of crossover sizes of HLA class I binding peptides detected using the methods shown (left) and a Venn diagram of exemplary standard and optimized workflows for LC-MS / MS detection of HLA class I binding peptides (right). [Figure 45-2] FIG. 45 shows an exemplary graph of crossover sizes of HLA class I binding peptides detected using the methods shown (left) and a Venn diagram of exemplary standard and optimized workflows for LC-MS / MS detection of HLA class I binding peptides (right).
[0386] [Figure 46-1]FIG. 46 shows an exemplary graph of crossover sizes of HLA class II binding peptides detected using the methods shown (left) and a Venn diagram of exemplary standard and optimized workflows for LC-MS / MS detection of HLA class II binding peptides (right). [Figure 46-2] FIG. 46 shows an exemplary graph of crossover sizes of HLA class II binding peptides detected using the methods shown (left) and a Venn diagram of exemplary standard and optimized workflows for LC-MS / MS detection of HLA class II binding peptides (right). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0387] Detailed Description All terms are intended to be understood as would be understood by one of ordinary skill 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.
[0388] The section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.
[0389] Although various features of the present disclosure may be described in the context of a single embodiment, the features may also be provided separately or in any suitable combination. Conversely, although the present disclosure may for clarity be described herein in the context of separate embodiments, the present disclosure may also be practiced in a single embodiment.
[0390] The present disclosure is based on the important observation that the presentation of antigens, specifically cancer antigens, by specific HLA class II alpha and beta chain pairs can be predicted with a high degree of confidence using a new computer-based machine learning HLA-peptide presentation prediction model that enables the use of HLA class II specific peptides for improved immunotherapy.
[0391] In one aspect, the present disclosure provides a method for predicting peptides that can accurately pair or bind to specific HLA class II alpha and beta chain heterodimers, ensuring that high fidelity binding of peptides to HLA class II proteins (composed of alpha and beta chain heterodimers) ensures that specific peptides are presented to T lymphocytes, thereby eliciting specific immune responses and avoiding any cross-reactivity or indiscriminate immunity. Some recent studies have also shown that CD4+ T cells can recognize ligands presented by HLA class II and contribute to tumor control. Ideally, cancer vaccines and other immunotherapies would utilize CD4+ T cell responses to direct, but current attempts have completely ignored HLA class II antigen prediction due to the insufficient accuracy of current prediction tools.
[0392] In one aspect, the present disclosure provides a method for predicting peptides that may accurately bind to a particular HLA class II protein such that peptides can be used to activate a more sustained and robust immune response when administered therapeutically to a subject expressing the particular cognate HLA class II protein due to the ability of the HLA class II protein to activate CD4+ T cells and stimulate immune memory. In some embodiments, the methods provided herein show an improvement over currently available predictors for the particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 1.1-fold improvement over currently available predictors for the particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 2-fold improvement over currently available predictors for the particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 3-fold improvement over currently available predictors for the particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 4-fold improvement over currently available predictors for the particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 5-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 6-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 7-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about an 8-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 9-fold improvement over currently available predictors for a particular HLA class II protein prediction.In some embodiments, the methods provided herein show at least about a 10-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 15-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 20-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 30-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 40-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 50-fold improvement over currently available predictors for a particular HLA class II protein prediction. In some embodiments, the methods provided herein show at least about a 60-fold improvement over currently available predictors for a particular HLA class II protein prediction.
[0393] In one aspect, the present invention provides a method for immunotherapy that is tailored or personalized for a specific subject.Every subject or patient expresses a specific array of HLA class I and HLA class II proteins.HLA typing is a well-known technique that allows determining the specific repertoire of HLA proteins expressed by a subject.Once the HLA heterodimer expressed by a specific subject is known, the improved, sophisticated and reliable method described herein for predicting the peptide that can bind to a specific HLA class II alpha chain and beta chain heterodimer with high fidelity ensures that a specific immune response that is tailored specifically for a subject can be generated.
[0394] In this application, the use of the singular includes the plural unless otherwise specified. It should be noted that as used herein, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. In this application, the use of "or" means "and / or" unless otherwise specified. Furthermore, the use of the term "including" and other forms such as "include," "includes," and "included" are not intended to be limiting. The term "one or more" or "at least one", such as one or more or at least one member(s) of a group of members, is clear in itself and, by way of further illustration, the term specifically encompasses reference to any one of said members, or any two or more of said members, such as any >3, >4, >5, >6 or >7 of said members, and up to all of said members.
[0395] References herein to "some embodiments," "an embodiment," "one embodiment," or "other embodiments" mean that a feature, structure, or characteristic described in connection with an embodiment is included in at least some embodiments of the present disclosure, but not necessarily in all embodiments.
[0396] As used in the specification and claim(s), the words "comprising" (and any form of comprising, such as "comprise" and "comprises"), "having" (and any form of having, such as "have" and "has"), "including" (and any form of including, such as "includes" and "include"), or "containing" (and any form of containing, such as "contains" and "contain") are inclusive or open-ended and do not exclude additional, unrecited elements or method steps. It is intended that any embodiment discussed herein can be implemented with respect to any method or composition of the disclosure, and vice versa. Additionally, the compositions of the disclosure can be used to achieve the methods of the disclosure.
[0397] The terms "about" or "approximately," when used herein in reference to a measurable value, such as a parameter, amount, duration, and the like, are intended to encompass variations of + / - 20% or less, + / - 10% or less, + / - 5% or less, or + / - 1% or less from the specified value, insofar as such variations are appropriate to the practice of the present disclosure. It is to be understood that the value to which the modifier "about" or "approximately" refers is itself also specifically disclosed.
[0398] The term "immune response" includes T cell-mediated and / or B cell-mediated immune responses that are influenced by modulation of T cell costimulation. Exemplary immune responses include T cell responses, such as cytokine production, and cellular cytotoxicity. Additionally, the term immune response includes immune responses that are indirectly influenced by T cell activation, such as antibody production (humoral response) and activation of cytokine-responsive cells, such as macrophages.
[0399] "Receptor" should be understood to mean a biological molecule or group of molecules that can bind to a ligand. Receptors can be useful in transmitting information in a cell, cell formation, or organism. A receptor includes at least one receptor unit and can contain two or more receptor units, where each receptor unit can consist of a protein molecule, e.g., a glycoprotein molecule. A receptor can have a structure that complements the structure of a ligand and form a complex with the ligand as a binding partner. Signaling information can be transmitted by a change in the conformation of the receptor after binding to a ligand on the surface of a cell. According to the present disclosure, a receptor can refer to a protein of MHC class I and class II that can form a receptor / ligand complex with a ligand, e.g., a peptide or peptide fragment of appropriate length. Class I and class II MHC peptides encoded by HLA class I and class II alleles are often referred to herein as HLA class I and HLA class II peptides, respectively, 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 class II molecules, or such common variants thereof, as would be well understood within the context of the skilled artisan's considerations.
[0400] A "ligand" is a molecule that can form a complex with a receptor. According to the present disclosure, a ligand is understood to mean, for example, a peptide or peptide fragment that has a suitable length and has a suitable binding motif in its amino acid sequence, and thus can bind to and form a complex with MHC class I or MHC class II proteins (i.e., HLA class I and HLA class II proteins).
[0401] An "antigen" is a molecule capable of stimulating an immune response and may be produced by cancer cells or 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 class II proteins on the surface of the cell. During the course of a naturally occurring immune response, antigens recognized in 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 on HLA class I molecules. Antigens that give rise to peptides recognized in association with HLA class I MHC molecules are generally peptides produced within the cell, which are processed and associated with class I MHC molecules. It is now understood that peptides that associate with a given HLA class I or class II molecule are characterized by having a common binding motif, and many different HLA class I and class II molecule binding motifs have been determined.Synthetic peptides that correspond to the amino acid sequence of a given antigen and contain the binding motif of a given HLA class I or class II molecule can also be synthesized.These peptides can then be added to appropriate APCs, and the APCs can be used to stimulate T helper cell or CTL responses in vitro or in vivo.The binding motifs, methods for synthesizing peptides, and methods for stimulating T helper cell or CTL responses are all known and readily available to those skilled in the art.
[0402] The term "peptide" is used interchangeably herein with "mutant peptide" and "neo-antigenic peptide." Similarly, the term "polypeptide" is used interchangeably herein with "mutant polypeptide" and "neo-antigenic polypeptide." "Neo-antigen" or "neoepitope" refers to a class of tumor antigens or tumor epitopes that arise from tumor-specific mutations in expressed proteins. The present disclosure further includes peptides that include tumor-specific mutations, peptides that include known tumor-specific mutations, and mutant polypeptides or fragments thereof identified by the methods of the present disclosure. These peptides and polypeptides are referred to herein as "neo-antigenic peptides" or "neo-antigenic polypeptides." The polypeptides or peptides may be of various lengths, may be in either neutral (uncharged) or salt form, and may be free or contain modifications such as glycosylation, side chain oxidation, phosphorylation, or any post-translational modification, and may be subject to conditions where the modifications do not destroy the biological activity of the polypeptides described herein. In some embodiments, the neo-antigenic peptides of the disclosure may comprise, for HLA Class I, 22 residues or less in length, e.g., from about 8 to about 22 residues, from about 8 to about 15 residues, or 9 or 10 residues, and for HLA Class II, 40 residues or less in length, e.g., from about 8 to about 40 residues, from about 8 to about 24 residues, from about 12 to about 19 residues, or from about 14 to about 18 residues. In some embodiments, the neo-antigenic peptide or neo-antigenic polypeptide comprises a neo-epitope.
[0403] The term "epitope" includes any protein determinant capable of specific binding to an antibody, antibody peptide, and / or antibody-like molecule as defined herein (including, but not limited to, a T-cell receptor). Epitope determinants generally consist of chemically active surface groupings of molecules such as amino acids or sugar side chains and generally have specific three dimensional structural characteristics, as well as specific charge characteristics.
[0404] A "T cell epitope" is a peptide sequence to which class I or class II MHC molecules can bind in the form of a peptide-presenting MHC molecule or MHC complex, which in turn is recognized in these forms by and bound by cytotoxic T lymphocytes or helper T cells, respectively.
[0405] The term "antibody" as used herein includes IgG (including IgG1, IgG2, IgG3, and IgG4), IgA (including IgA1 and IgA2), IgD, IgE, IgM, and IgY, and is intended to encompass whole antibodies, including single-chain whole antibodies, and antigen-binding (Fab) fragments thereof. Antigen-binding antibody fragments include, but are not limited to, Fab, Fab', and F(ab')2, Fd (consisting of VH and CH1), single-chain variable fragments (scFv), single-chain antibodies, disulfide-linked variable fragments (dsFv), and fragments containing either the VL or VH domain. Antibodies may be from any animal source. Antigen-binding antibody fragments, including single-chain antibodies, may include the variable region(s) alone or in combination with the whole or parts of the following: hinge region, CH1 domain, CH2 domain, and CH3 domain. Any combination of variable region(s) and hinge region, CH1 domain, CH2 domain, and CH3 domain is also included. The antibody can be, for example, a monoclonal, polyclonal, chimeric, humanized, and human monoclonal and polyclonal antibody that specifically binds to an HLA-associated polypeptide or an HLA-HLA binding peptide (HLA-peptide) complex. Those skilled in the art will appreciate that various immunoaffinity techniques are suitable for enriching soluble proteins, such as soluble HLA-peptide complexes or membrane-associated HLA-associated polypeptides, e.g., those proteolytically cleaved from the membrane. These include techniques in which (1) one or more antibodies capable of specifically binding to soluble proteins are immobilized on a fixed or mobile substrate (e.g., plastic wells or resins, latex or paramagnetic beads), and (2) a solution containing soluble proteins from a biological sample is passed over the substrate coated with the antibodies, thereby binding the soluble proteins to the antibodies. The substrate with the antibodies and bound soluble proteins is separated from the solution, and if necessary the antibodies and soluble proteins are separated, for example, by varying the pH and / or ionic strength and / or ionic composition of the solution in which the antibodies bathe.Alternatively, immunoprecipitation techniques can be used in which antibodies are combined with soluble proteins to form macromolecular aggregates, which can be separated from the solution by molecular sieving techniques or by centrifugation.
[0406] The term "immunopurification (IP)" (or immunoaffinity purification or immunoprecipitation) is a well-known and widely used process in the art for isolating a desired antigen from a sample. In general, the process involves contacting a sample containing the desired antigen with an affinity matrix in which an antibody against the antigen is covalently attached to a solid phase. The antigen in the sample becomes bound to the affinity matrix through immunochemical binding. The affinity matrix is then washed to remove any unbound species. The antigen is removed from the affinity matrix by changing the chemical composition of the solution in contact with the affinity matrix. Immunopurification can be performed on a column containing the affinity matrix, in which case the solution is the eluent. Alternatively, immunopurification can be in a batch process, in which case the affinity matrix is maintained as a suspension in the solution. The key step in the process is the removal of the antigen from the matrix. This is generally achieved by increasing the ionic strength of the solution in contact with the affinity matrix, for example by adding inorganic salts. Changing the pH can also be effective to dissociate the immunochemical binding of the antigen to the affinity matrix.
[0407] An "agent" is any small molecule chemical compound, an antibody, a nucleic acid molecule, or a polypeptide, or a fragment thereof.
[0408] An "alteration" or "change" is an increase or decrease. The alteration may be as little as 1%, 2%, 3%, 4%, 5%, 10%, 20%, 30%, or 40%, 50%, 60%, or even as much as 70%, 75%, 80%, 90%, or 100%.
[0409] A "biological sample" is any tissue, cell, fluid, or other material derived from an organism. As used herein, the term "sample" includes biological samples, such as any tissue, cell, fluid, or other material derived from an organism. "Specifically binds" refers to a compound (e.g., a peptide) that recognizes and binds to a molecule (e.g., a polypeptide) in a sample, e.g., a biological sample, but does not substantially recognize or bind to other molecules.
[0410] "Capture reagent" refers to a reagent that specifically binds to a molecule (eg, a nucleic acid molecule or a polypeptide) in order to select or isolate the molecule (eg, a nucleic acid molecule or a polypeptide).
[0411] As used herein, the terms "determining," "assessing," "assaying," "measuring," "detecting," and their grammatical equivalents refer to both quantitative and qualitative determinations, and thus the term "determining" is used interchangeably herein with "assaying," "measuring," and the like. When a quantitative determination is intended, phrases such as "determining the amount" of an analyte are used. When a qualitative and / or quantitative determination is intended, phrases such as "determining the level" of an analyte or "detecting" an analyte are used.
[0412] A "fragment" is a portion of a protein or nucleic acid that is substantially identical to a reference protein or nucleic acid. In some embodiments, the portion retains at least 50%, 75%, or 80%, or 90%, 95%, or even 99% of the biological activity of the reference protein or nucleic acid described herein.
[0413] The terms "isolated," "purified," "biologically pure," and their grammatical equivalents refer to the free, to varying degrees, of a material from components that are normally associated with it as found in its native state. "Isolate" indicates some degree of separation from the original source or surroundings. "Purify" indicates a degree of separation greater than isolation. A "purified" or "biologically pure" protein is sufficiently free of other materials such that no impurities substantially affect the biological properties of the protein or cause other deleterious consequences. That is, a nucleic acid or peptide of the present disclosure is purified if it is substantially free of cellular material, viral material, or culture medium if produced by recombinant DNA techniques, or chemical precursors or other chemicals if chemically synthesized. Purity and homogeneity are generally determined using analytical chemistry techniques, such as polyacrylamide gel electrophoresis or high performance liquid chromatography. The term "purified" may indicate that the nucleic acid or protein gives rise to essentially one band in an electrophoretic gel. For proteins that can be subject to modifications, such as phosphorylation or glycosylation, the different modifications can give rise to different isolated proteins, which can be purified separately.
[0414] An "isolated" polypeptide (e.g., a peptide from an HLA-peptide complex) or polypeptide complex (e.g., an HLA-peptide complex) is a polypeptide or polypeptide complex of the disclosure that has been separated from naturally associated components. Generally, a polypeptide or polypeptide complex is isolated if it is at least 60%, by weight, free from naturally associated proteins and naturally occurring organic molecules. A preparation can be at least 75%, at least 90%, or at least 99%, by weight, a polypeptide or polypeptide complex of the disclosure. An isolated polypeptide or polypeptide complex of the disclosure can be obtained, for example, by extraction from a natural source, by expressing a recombinant nucleic acid encoding one or more components of such a polypeptide or polypeptide complex, or by chemically synthesizing one or more components of the polypeptide or polypeptide complex. Purity can be measured by any reasonable method, for example, by column chromatography, polyacrylamide gel electrophoresis, or HPLC analysis. In some cases, MHC class II proteins (ie, MHC class II peptides) encoded by HLA alleles are referred to interchangeably herein as HLA class II proteins (or HLA class II peptides).
[0415] The term "vector" refers to a nucleic acid molecule capable of transporting or mediating the expression of heterologous nucleic acid. Plasmids are one of the genera encompassed by the term "vector". A vector generally refers to a nucleic acid sequence that contains 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 an operably linked gene and / or nucleic acid sequence is referred to herein as an "expression vector". In general, useful expression vectors are often in the form of "plasmids", which in their vector form are not bound to a chromosome and generally refer to circular double-stranded DNA molecules that contain the entities necessary for stable or transient expression of the encoded DNA. 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 be integrated into the genome of the host or can replicate autonomously in the cell. The vector can be a DNA vector or an RNA vector. Other forms of expression vectors that serve equivalent functions known to those of skill in the art can also be used, e.g., autonomously replicating extrachromosomal vectors or vectors capable of integrating into a host genome. Exemplary vectors are those that are capable of autonomous replication and / or expression of nucleic acids to which they are linked.
[0416] The term "spacer" or "linker", when used in reference to a fusion protein, refers to a peptide that joins the proteins that make up the fusion protein. In general, a spacer has no specific biological activity other than joining proteins or RNA sequences or preserving some minimum distance or other spatial relationship between the proteins or RNA sequences. However, in some embodiments, the amino acids that are constituents of the spacer can be selected to affect some property of the molecule, such as the folding, net charge, or hydrophobicity of the molecule. Linkers suitable for use in certain embodiments of the present disclosure are well known to those of skill in the art and include, but are not limited to, linear or branched carbon linkers, heterocyclic carbon linkers, or peptide linkers. Linkers are used to separate two antigenic peptides by a sufficient distance to ensure that, in some embodiments, each antigenic peptide folds properly. Exemplary peptide linker sequences adopt a flexible, extended conformation and do not tend to give rise to ordered secondary structures. Exemplary amino acids in flexible protein regions include Gly, Asn, and Ser. Virtually any permutation of an amino acid sequence containing Gly, Asn, and Ser is predicted to meet the above criteria for a linker sequence. Other near-neutral amino acids, such as Thr and Ala, can also be used in the linker sequence. Still other amino acid sequences that can be used as linkers are described 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 No. 4,935,233; and U.S. Patent No. 4,751 , No. 180.
[0417] The term "neoplasia" refers to any disease caused by or resulting from inappropriately high levels of cell division, inappropriately low levels of apoptosis, or both. Glioblastoma is one non-limiting example of neoplasia or cancer. The term "cancer" or "tumor" or "hyperproliferative disorder" refers to the presence of cells that have characteristics typical of cancer-causing cells, such as uncontrolled proliferation, immortality, metastatic potential, rapid growth and proliferation rate, and certain characteristic morphological features. Cancer cells are often in the form of tumors, but such cells may exist alone in an animal or may be non-tumorigenic cancer cells, such as leukemia cells. Cancers include, but are not limited to, B cell cancers (e.g., multiple myeloma, Waldenstrom's macroglobulinemia), heavy chain diseases (e.g., alpha chain disease, gamma chain disease, and mu chain disease), benign monoclonal gammopathy, and immune cell amyloidosis, melanoma, breast cancer, lung cancer, bronchial cancer, colorectal cancer, prostate cancer (e.g., metastatic prostate cancer, hormone refractory prostate cancer), pancreatic cancer, gastric cancer, ovarian cancer, urinary bladder cancer, brain or central nervous system cancer, peripheral nervous system cancer, esophageal cancer, cervical cancer, uterine or uterine cancer, oral or pharyngeal cancer, liver cancer, kidney cancer, testicular cancer, biliary tract cancer, small intestine or appendix cancer, salivary gland cancer, thyroid cancer, adrenal cancer, osteosarcoma, chondrosarcoma, and cancer of hematological tissues.Other non-limiting examples of types of cancer applicable to the methods encompassed by the present disclosure include human sarcomas and carcinomas, such as fibrosarcoma, myxosarcoma, liposarcoma, chondrosarcoma, osteogenic sarcoma, chordoma, angiosarcoma, endothelioma, lymphangiosarcoma, lymphangioendothelioma, synovium, mesothelioma, Ewing's tumor, leiomyosarcoma, rhabdomyosarcoma, 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 adenocarcinoma, cystadenocarcinoma, medullary carcinoma, bronchogenic lung carcinoma, renal cell carcinoma, hepatoma, cholangiocarcinoma, 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, pinealoma, hemangioblastoma, acoustic neuroma, oligodendrocyte glioma, meningioma, melanoma, neuroblastoma, retinoblastoma; leukemias, such as acute lymphocytic leukemia and acute myelocytic leukemia (myeloblastic leukemia, proostotic leukemia, myelomonocytic leukemia, monocytic leukemia, and erythroleukemia); chronic leukemias (chronic myelocytic (granulocytic) leukemia and chronic lymphocytic leukemia); and polycythemia vera, lymphoma (Hodgkin's disease and non-Hodgkin's disease), multiple myeloma, Waldenstrom's macroglobulinemia, and heavy chain disease. In some embodiments, the cancer is an epithelial cancer, such as, but not limited to, bladder cancer, breast cancer, cervical cancer, colon cancer, gynecological cancer, renal 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 cancer is non-small cell lung cancer, non-papillary renal cell carcinoma, cervical cancer, ovarian cancer (e.g., serous ovarian cancer), or breast cancer. Epithelial cancers can be characterized in a variety of other ways, including, but not limited to, serous, endometrioid, mucinous, clear cell, Brenner, or anaplastic. In some embodiments, the present disclosure is used to treat, diagnose, and / or prognose lymphoma or subtypes thereof, including, but not limited to, mantle cell lymphoma. Lymphoproliferative disorders are also considered to be proliferative diseases.
[0418] The term "vaccine" should be understood to mean a composition for generating immunity for the prevention and / or treatment of disease (e.g., neoplasia / tumor / infectious agent / autoimmune disease). Thus, a vaccine is a medicine that contains an antigen and is intended to be used to generate specific defense and protection by vaccination in humans or animals. A "vaccine composition" may include a pharma-ceutically acceptable excipient, carrier, or diluent. Aspects of the present disclosure relate to the use of the 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 embodiments, an antigen-based vaccine contains at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 16, at least 17, at least 18, at least 19, at least 20, at least 21, at least 22, at least 23, at least 24, at least 25, at least 26, at least 27, at least 28, at least 29, at least 30, or more antigenic peptides.In some embodiments, the antigen-based vaccine comprises 2-100, 2-75, 2-50, 2-25, 2-20, 2-19, 2-18, 2-17, 2-16, 2-15, 2-14, 2-13, 2-12, 2-10, 2-9, 2-8, 2-7, 2-6, 2-5, 2-4, 3-100, 3-75, 3-50, 3-25, 3-20, 3-19, 3-18, 3-17, 3-16, 3-15, 3-14, 3-13, 3-12, 3-10, 3-9, 3-8, 3- The antigenic peptides include 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 types of antigenic peptides. In some embodiments, the antigen-based vaccine contains 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 antigenic peptides. In some cases, the antigenic peptide is a neo-antigenic peptide. In some cases, the antigenic peptide comprises one or more neo-epitopes.
[0419] The term "pharmaceutically acceptable" refers to being approved or approvable by a federal or state government regulatory agency or listed in the United States Pharmacopoeia or other generally recognized pharmacopoeias for use in animals, including humans. A "pharmaceutically acceptable excipient, carrier, or diluent" refers to an excipient, carrier, or diluent that can be administered to a subject together with an agent, does not destroy the pharmacological activity of the agent, and is non-toxic when administered in a dose sufficient to deliver a therapeutic amount of the agent. The "pharmaceutically acceptable salt" of the pooled disease-specific antigens described herein may be an acid or base salt generally considered in the art to be suitable for use in contact with human or animal tissues without excessive toxicity, irritation, allergic response, or other problems or complications. Such salts include inorganic and organic acid salts of basic residues such as amines, and alkali or organic salts of acidic residues such as carboxylic acids. Particular 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, sulfamic acid, sulfanilic acid, formic acid, toluenesulfonic acid, methanesulfonic acid, benzenesulfonic acid, ethanedisulfonic acid, 2-hydroxyethylsulfonic acid, nitric acid, benzoic acid, 2-acetoxybenzoic acid, citric acid, tartaric acid, lactic acid, stearic acid, salicylic acid, glutamic acid, ascorbic acid, pamoic acid, succinic acid, fumaric acid, maleic acid, propionic acid, hydroxymaleic acid, hydroiodic acid, phenylacetic acid, alkanoic acids, such as acetic acid, HOOC-(CH2)n-COOH, where n is 0-4, and the like. Similarly, pharma- ceutically acceptable cations include, but are not limited to, sodium, potassium, calcium, aluminum, lithium, and ammonium. Those of skill in the art will recognize from this disclosure and knowledge in the art additional pharma- ceutically acceptable salts for the pooled disease-specific antigens provided herein, including those listed by Remington's Pharmaceutical Sciences, 17th ed., Mack Publishing Company, Easton, PA, p. 1418 (1985).In general, pharma- ceutically acceptable acid or base salts can be synthesized from a parent compound that contains a basic or acidic moiety by any conventional chemical method. Briefly, such salts can be prepared by reacting the free acid or base form of these compounds with a stoichiometric amount of the appropriate base or acid in an appropriate solvent.
[0420] Nucleic acid molecules useful in the methods of the present disclosure include any nucleic acid molecule encoding a polypeptide or fragment thereof of the present disclosure. Such nucleic acid molecules do not need to be 100% identical to an endogenous nucleic acid sequence, but generally exhibit substantial identity. A polynucleotide having substantial identity to an endogenous sequence can generally hybridize with at least one strand of a double-stranded nucleic acid molecule. "Hybridize" refers to the pairing of nucleic acid molecules between complementary polynucleotide sequences or portions thereof to form a double-stranded molecule under various stringency conditions (see, e.g., Wahl, GM and SL Berger (1987) Methods Enzymol. 152: 399; Kimmel, AR (1987) Methods Enzymol. 152: 507). For example, stringent salt concentrations are usually about 750 mM. The preferred organic solvents may be less than about 100 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 stringency hybridization can be achieved without organic solvents, such as formamide, while high stringency hybridization can be achieved in the presence of at least about 35% formamide, or at least about 50% formamide. Stringent temperature conditions can usually include temperatures of at least about 30°C, at least about 37°C, or at least about 42°C. Additional parameters that vary, such as hybridization time, concentration of detergent, such as sodium dodecyl sulfate (SDS), and inclusion or exclusion of carrier DNA, are well known to those skilled in the art. By combining these various conditions as necessary, various levels of stringency can be achieved. In an exemplary embodiment, hybridization can be performed at 30° C., 750 mM NaCl, 75 mM trisodium citrate, and 1% SDS. In another exemplary embodiment, hybridization can be performed at 37° C., 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 be performed at 42° C., 250 mM NaCl, 25 mM trisodium citrate, 1% SDS, 50% formamide, and 200 μg / ml ssDNA. Useful variations of these conditions will be readily apparent to those of skill in the art. For most applications, stringency can also be varied with respect to post-hybridization washing steps. Wash stringency conditions can be defined by salt concentration and by temperature. As noted above, wash stringency can be increased by decreasing salt concentration or increasing temperature. For example, stringent salt concentrations for the wash steps can be less than about 30 mM NaCl and 3 mM trisodium citrate, or less than about 15 mM NaCl and 1.5 mM trisodium citrate.Stringent temperature conditions for the wash step may include a temperature of at least about 25° C., at least about 42° C., or at least about 68° C. In an exemplary embodiment, the wash step may be performed at 25° C., 30 mM NaCl, 3 mM trisodium citrate, and 0.1% SDS. In another exemplary embodiment, the wash step may be performed at 42° C., 15 mM NaCl, 1.5 mM trisodium citrate, and 0.1% SDS. In another exemplary embodiment, the wash step may be performed at 68° C., 15 mM NaCl, 1.5 mM trisodium citrate, and 0.1% SDS. Additional variations of these conditions will be readily apparent to one of skill in the art. Hybridization techniques are well known to those of skill in the art, see, e.g., 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.
[0421] "Substantially identical" refers to a polypeptide or nucleic acid molecule exhibiting at least 50% identity to a reference amino acid sequence (e.g., any one of the amino acid sequences described herein) or a reference nucleic acid sequence (e.g., any one of the nucleic acid sequences described herein). Such a sequence may be at least 60%, 80% or 85%, 90%, 95%, 96%, 97%, 98%, or even 99% or greater identical at the amino acid or nucleic acid level to the sequence used for comparison. Sequence identity is generally determined using sequence analysis software (e.g., the Sequence Analysis Software Package of the Genetics Computer The degree of identity is measured using the BLAST, BESTFIT, GAP, or PILEUP / PRETTYBOX programs (Biotechnology and Biotechnology Research Group, University of Wisconsin Biotechnology Center, 1710 University Avenue, Madison, Wis. 53705). Such software matches identical or similar sequences by assigning degrees of homology to various substitutions, deletions, and / or other modifications. Conservative substitutions generally include substitutions within the following groups: glycine, alanine; valine, isoleucine, leucine; aspartic acid, glutamic acid, asparagine, glutamine; serine, threonine; lysine, arginine; and phenylalanine, tyrosine. In an exemplary approach to determine the degree of identity, the BLAST program can be used with a probability score between e-3 and em°, which indicates closely related sequences. The "reference" is the comparison standard.
[0422] The term "subject" or "patient" refers to an animal that is the object of treatment, observation, or experiment. By way of example only, a subject can include, but is not limited to, a human or a non-human mammal, such as a mammal, including but not limited to a non-human primate, mouse, bovine, equine, canine, ovine, or feline.
[0423] The terms "treat", "treated", "treating", "treatment" and the like refer to reducing, preventing, or reversing a disorder and / or its associated symptoms (e.g., neoplasia or tumor, or infectious agent, or autoimmune disease). "Treating" may refer to administering therapy to a subject after onset or suspected onset of a disease (e.g., cancer, or infection by an infectious agent, or autoimmune disease). "Treating" includes the concept of "alleviating", which refers to reducing the frequency or severity of occurrence or recurrence of any symptoms or other pathological effects associated with a disease, and / or the side effects associated with treatment. The term "treating" also encompasses the concept of "managing", which refers to reducing the severity of a disease or disorder in a patient, e.g., extending the lifespan or survival of a patient with a disease, or delaying its recurrence, e.g., lengthening the period of remission in a patient suffering from a disease. Although not intended to be exclusive, it is understood that treating a disorder or condition does not seek to completely eliminate the disorder, condition, or its associated symptoms.
[0424] The terms "prevent," "preventing," "prevention," and their grammatical equivalents, as used herein, refer to avoiding or delaying the onset of symptoms associated with a disease or condition in a subject who has not experienced such symptoms at the time that administration of an agent or compound is initiated.
[0425] The term "therapeutic effect" refers to some degree of alleviation of one or more symptoms of a disorder (e.g., neoplasia, tumor, or infection by an infectious agent or autoimmune disease) or pathological deviations associated therewith. "Therapeutically effective amount" as used herein refers to an amount of an agent that, when administered in a single or multiple doses to a cell or subject, is effective to prolong the survival of a patient with such disorder, reduce, prevent or delay one or more signs or symptoms of the disorder, etc., beyond that expected in the absence of such treatment. "Therapeutically effective amount" is intended to limit the amount necessary to achieve a therapeutic effect. A physician or veterinarian of ordinary skill in the art can readily determine and prescribe the "therapeutically effective amount" (e.g., ED50) of the pharmaceutical composition required. For example, a physician or veterinarian can start a dose of the disclosed compound for use in a pharmaceutical composition at a level lower than that required to achieve the desired therapeutic effect, and gradually increase the dosage until the desired effect is achieved. Disease, condition, and disorder are used interchangeably herein.
[0426] Those skilled in the art will understand that the terms "peptide tag", "affinity tag", "epitope tag" or "affinity acceptor tag" are used interchangeably herein. As used herein, the term "affinity acceptor tag" refers to an amino acid sequence that allows the tagged protein to be easily detected or purified, for example, by affinity purification. Affinity acceptor tags are generally (but not necessarily) located at or near the N-terminus or C-terminus of an HLA allele. A variety of peptide tags are well known in the art. Non-limiting examples include poly-histidine tags (e.g., 4-15 consecutive His residues, such as 8 consecutive His residues); poly-histidine-glycine tags; HA tags (e.g., Field et al., Mol. Cell. Biol., 8: 2159, 1988); c-myc tags (e.g., Evans et al., Mol. Cell. Biol., 5: 3610, 1985); herpes simplex virus glycoprotein D (gD) tags (e.g., Paborsky et al., Protein Engineering, 3: 547, 1990); FLAG tags (e.g., See, e.g., Hopp et al., BioTechnology, 6: 1204, 1988; U.S. Pat. No. 4,703,000. 04 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 tag (e.g. For example, Lutz-Freyemuth et al., Proc. Natl. Acad. Sci. USA, 87: 6393, 1990; streptavidin tags (StrepTag™ or StrepTagII); (trademark); see, for example, Schmidt et al., J. Mol. Biol., 255 (5):753-766, 1996 or U.S. Pat. No. 5,506,121; also commercially available from Sigma-Genosys); or the VSV-G epitope tag derived from the Vesicular Stomatis virus glycoprotein; or the V5 tag derived from a small epitope (Pk) found in the P and V proteins of the Simian Virus 5 (SV5) paramyxovirus. In some embodiments, the affinity acceptor tag is an "epitope tag," which is a type of peptide tag that adds a recognizable epitope (antibody binding site) to an HLA-protein, resulting in binding of the corresponding antibody, thereby allowing identification or affinity purification of the tagged protein. Non-limiting examples of epitope tags are Protein A or Protein G, which bind IgG. In some embodiments, the matrix of IgG Sepharose 6 Fast Flow chromatography resin is covalently coupled to human IgG. This resin allows for high flow rates for rapid and convenient purification of protein A tagged proteins. Numerous other tag moieties are known and can be envisioned by those of skill in the art and are contemplated herein. Any peptide tag can be used as long as it can be expressed as an element of an affinity acceptor tagged HLA-peptide complex.
[0427] As used herein, the term "affinity molecule" refers to a molecule or ligand that binds to an affinity acceptor peptide with chemical specificity. Chemical specificity is the ability of a protein's binding site to bind a specific ligand. The fewer ligands a protein can bind, the greater its specificity. Specificity describes the strength of binding between a given protein and a ligand. This relationship can be described by the dissociation constant (KD), which characterizes the equilibrium between bound and unbound states for a protein-ligand system.
[0428] The term "affinity acceptor-tagged HLA-peptide complex" refers to a complex comprising an HLA class I or class II associated peptide or portion thereof, specifically bound to a single allele recombinant HLA class I or class II peptide comprising an affinity acceptor peptide.
[0429] The terms "specific binding" or "specifically binding," when used in reference to the interaction of an affinity molecule and an affinity acceptor tag or epitope and an HLA peptide, mean that the interaction is dependent on the presence of a particular structure (e.g., an antigenic determinant or epitope) on the protein; in other words, the affinity molecule recognizes and binds to a specific affinity acceptor peptide structure, and not to proteins in general.
[0430] 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 acceptor tag" and an "affinity molecule" and an HLA-binding peptide and an HLA class I or class II molecule. KD is the dissociation constant and has units of molality. Affinity constant is the inverse of the dissociation constant. Affinity constant is sometimes used as a general term to describe this chemical entity. Affinity constant is a direct measure of the energy of binding. Affinity can be experimentally determined, for example, by surface plasmon resonance (SPR) using a commercially available Biacore SPR unit. Affinity can also be expressed as the inhibitory concentration 50 (IC50), which is the concentration at which 50% of the peptide is displaced. Similarly, lnIC50 refers to the natural logarithm of IC50. K off refers to the dissociation rate constant for dissociation of an affinity molecule from an HLA-peptide complex tagged with, for example, an affinity acceptor.
[0431] In some embodiments, affinity acceptor tagged HLA-peptide complexes contain biotin acceptor peptides (BAPs) and are immunopurified from complex cell mixtures using streptavidin / NeutrAvidin beads. Biotin-avidin / streptavidin binding is the strongest non-covalent interaction known in nature. This property is exploited as a biological tool for a wide range of applications, such as immunopurification of proteins with covalently attached biotin. In an exemplary embodiment, a biotin acceptor peptide (BAP) is incorporated into a nucleic acid sequence encoding an HLA allele as an affinity acceptor tag for immunopurification. BAPs can be specifically biotinylated at a single lysine residue in the tag in vivo or in vitro (e.g., U.S. Patent Nos. 5,723,584; 5,874,239; and 5,932,433; and UK Patent No. GB2370039). BAPs are generally 15 amino acids long and contain a single lysine as a biotin acceptor residue. In some embodiments, the BAP is placed at or near the N-terminus or C-terminus of a single allelic HLA peptide. In some embodiments, the BAP is placed between the heavy chain domain and the β2 microglobulin domain of an HLA class I peptide. In some embodiments, the BAP is placed between the β-chain domain and the α-chain domain of an HLA class II peptide. In some embodiments, the BAP is placed in the loop region between the α1, α2, and α3 domains of the heavy chain of HLA class I, or between the α1 and α2 domains and between the β1 and β2 domains of the α and β chains, respectively, of HLA class II. Exemplary constructs designed for HLA class I and class II expression incorporating BAP for biotinylation and immunopurification are described in FIG. 2.
[0432] As used herein, the term "biotin" refers to the compound biotin itself as well as its analogs, derivatives and variants. Thus, the term "biotin" includes biotin (cis-hexahydro-2-oxo-1H-thieno[3,4]imidazole-4-pentanoic acid) and any derivatives and analogs thereof, including biotin-like compounds. Such compounds include, for example, biotin-eN-lysine, biocytin hydrazide, amino or sulfhydryl derivatives of 2-iminobiotin and biotinyl-E-aminocaproic acid-N-hydroxysuccinimide ester, sulfosuccinimide iminobiotin, biotin bromoacetyl hydrazide, p-diazobenzoyl biocytin, 3-(N-maleimidopropionyl) biocytin, desthiobiotin, and the like. The term "biotin" also includes biotin variants that can specifically bind to one or more of the following: rizavidin, avidin, streptavidin, tamavidin moieties, or other avidin-like peptides.
[0433] As used herein, a "PPV determination method" may refer to a presentation PPV determination method. For example, a "PPV determination method" may include (a) processing amino acid information of a plurality of test peptide sequences using an HLA peptide presentation prediction model, such as a machine learning HLA peptide presentation prediction model, to generate a plurality of test presentation predictions, each of which may result in a class II HLA allele of a cell, such as a class II HLA allele of a cell of a subject; The method includes the steps of: (a) identifying or calling a top percentage of the plurality of test peptide sequences, such as, for example, 0.2% of the top plurality of test peptide sequences, as being presented by class II HLA alleles of the cell; and (b) calculating a PPV of an HLA peptide presentation prediction model, the PPV being determined to be present by one or more proteins encoded by HLA alleles of the cell, the plurality of test peptide sequences including at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, and at least 500 test peptide sequences including at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, and at least 499 decoy peptide sequences contained within a protein encoded by a genome of an organism, such as an organism of the same species as the subject, the plurality of test peptide sequences including a ratio of the number of hit peptide sequences to the number of decoy peptide sequences of less than 1, such as, for example, a ratio of the at least one hit peptide sequence to the at least 499 decoy peptide sequences of 1:499; (b) identifying or calling a top percentage of the plurality of test peptide sequences, such as, for example, 0.2% of the top plurality of test peptide sequences, as being presented by class II HLA alleles of the cell; and (c) calculating a PPV of an HLA peptide presentation prediction model, the PPV being determined to be present by one or more proteins encoded by HLA alleles of the cell, the plurality of test peptide sequences including at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, the plurality of test peptide sequences including at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, the plurality of test peptide sequences including at least one decoy peptide sequence contained within a protein encoded by a genome of an organism, such as an organism of the same species as the subject. and the fraction of test peptide sequences identified or called as presented by class II HLA alleles of the cell, which are peptides observed as presented by HLA alleles. In some embodiments, the decoy peptides are of the same length, i.e., contain 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 compared to the hit peptides. In some embodiments, the decoy peptides are peptides that are endogenous peptides. In some embodiments, the decoy peptides are synthetic peptides. In some embodiments,Decoy peptide is an endogenous peptide identified by mass spectrometry to bind to a first MHC class I or class II protein, and the first MHC class I or class II protein is distinct from the second MHC class I or class II protein that binds to the hit peptide. In some embodiments, the decoy peptide can be a scrambled peptide, for example, the decoy peptide can include an amino acid sequence in which the amino acid positions within the peptide length are rearranged relative to the amino acid positions of the hit 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 about 1:10, 1:20, 1:50, 1:100, 1:250, 1:500, 1:1000, 1:1500, 1:2000, 1:2500, 1:5000, 1:7500, 1:10000, 1:25000, 1:50000 or 1:100000. In some embodiments, at least one hit peptide sequence is 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 decoy peptide sequences are at least 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700,4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400, 8500, 8600, 87 00, 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, 320 00, 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, 87 The decoy peptide sequence may be 500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000. In some embodiments, the at least 500 test peptide sequences are at least 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100,4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800, 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200 , 8300, 8400, 8500, 8600, 8700, 8800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 10000, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000, 19000, 20000, 21000, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000 , 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41000, 42000, 43000, 44000, 45000, 46000, 47000, 48000, 49000, 50000, 52500, 55000, 57500, 60000, 62500, 65000, 67500, 70000, 72500, 75000, 77500, 80000, 82500, The peptides include 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 test peptide sequences. In some embodiments, identifying or calling a top percentage of the plurality of test peptide sequences as presented by class II HLA alleles of the cell comprises identifying or calling a top percentage of the plurality of test peptide sequences as presented by class II HLA alleles of the cell, the top percentage being selected from the group consisting of the top 0.20%, 0.30%, 0.40%, 0.50%, 0.60%, 0.70%, 0.80%, 0.90%, 1.00%, 1.10%, 1.20%, 1.30%, 1.40%, 1.50%, 1.60%, 1.70%,1.80%, 1.90%, 2.00%, 2.10%, 2.20%, 2.30%, 2.40%, 2.50%, 2.60%, 2.70%, 2.80%, 2.90%, 3.00%, 3.10%, 3.20%, 3.30%, 3.40%, 3.50%, 3.60%, 3.70%, 3.80%, 3.90%, 4.00 %, 4.10%, 4.20%, 4.30%, 4.40%, 4.50%, 4.60%, 4.70%, 4.80%, 4.90%, 5.00%, 5.10%, 5.20%, 5.30%, 5.40%, 5.50%, 5.60%, 5.70%, 5.80%, 5.90%, 6.00%, 6.10%, 6.20%, 6.3 0%, 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 In some embodiments, the method includes identifying or calling 0.60%, 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% of the total number of HLA alleles of the cell as being presented by a class II HLA allele of the cell. In some embodiments, the cell is a monoallelic cell.
[0434] As used herein, a "PPV determination method" may refer to a binding PPV determination method. For example, a "PPV determination method" may include (a) processing amino acid information of a plurality of test peptide sequences using an HLA peptide binding prediction model, such as a machine learning HLA peptide binding prediction model, to generate a plurality of test binding predictions, each of which is associated with a class II HLA allele of a cell, such as a class II HLA allele of a cell of a subject; The method includes the steps of: indicating a likelihood that one or more proteins encoded by HLA alleles bind to a given test peptide sequence of a plurality of test peptide sequences, the plurality of test peptide sequences comprising (i) at least one hit peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, and (ii) at least 20 test peptide sequences comprising at least 19 decoy peptide sequences contained within a protein comprising at least one peptide sequence identified by mass spectrometry as being presented by an HLA protein expressed in the cell, the plurality of test peptide sequences comprising a ratio of the number of hit peptide sequences to the number of decoy peptide sequences of less than one, such as a 1:19 ratio of at least one hit peptide sequence to at least 19 decoy peptide sequences; (b) identifying or calling a top percentage of the plurality of test peptide sequences, such as 5% of the top plurality of test peptide sequences, as binding to the HLA protein; and (c) calculating a PPV of the HLA peptide binding prediction model, the PPV being a peptide observed by mass spectrometry to be presented by a class II HLA allele of the cell. This may refer to a method in which the fraction of test peptide sequences identified or called as binding to an HLA allele is determined. In some embodiments, the ratio of the number of hit peptide sequences to the number of decoy peptide sequences is about 1:2, 1:3, 1:4, 1:5, 1:10, 1:20, 1:25, 1:30, 1:40, 1:50, 1:75, 1:100, 1:200, 1:250, 1:500, or 1:1000. In some embodiments, at least one hit peptide sequence is 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, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89,20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100 hit peptide sequences. In some embodiments, the at least 19 decoy peptide sequences are at least 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200 , 3300, 3400, 3500, 3600, 3700, 3800, 3900, 4000, 4100, 4200, 4300, 4400, 4500, 4600, 4700, 4800, 4900, 5000, 5100, 5200, 5300, 5400, 5500, 5600, 5700, 5800 , 5900, 6000, 6100, 6200, 6300, 6400, 6500, 6600, 6700, 6800, 6900, 7000, 7100, 7200, 7300, 7400, 7500, 7600, 7700, 7800, 7900, 8000, 8100, 8200, 8300, 8400 , 8500, 8600, 8700, 8800, 8900, 9000, 9100, 9200, 9300, 9400, 9500, 9600, 9700, 9800, 9900, 10000, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000,19000, 20000, 21000, 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, 7000 The decoy peptide sequence may include 0, 72500, 75000, 77500, 80000, 82500, 85000, 87500, 90000, 92500, 95000, 97500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 decoy peptide sequences. In some embodiments, the at least 20 test peptide sequences are at least 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500 600, 700, 800, 900, 1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, 3000, 3100, 3200, 3300, 3400, 3500, 3600, 37 00, 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, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 4000 0, 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, 97 The peptides include 500, 100000, 125000, 150000, 175000, 200000, 225000, 250000, 275000, 300000, 325000, 350000, 375000, 400000, 425000, 450000, 475000, 500000, 600000, 700000, 800000, 900000 or 1000000 test peptide sequences. In some embodiments, identifying or calling a top percentage of the plurality of test peptide sequences as being presented by class II HLA alleles of the cell comprises identifying or calling the top 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, 20%, 21%, 22%, 23%, 24%, 25%, 26%, 27%, 28%, 29%, 30%, 31%, 32%, 33%, 34%, 35%, 36%, 37%, 38%, 39%, or 40% as being presented by class II HLA alleles of the cell. In some embodiments, the cell is a monoallelic cell.
[0435] Human leukocyte antigen (HLA) system The immune system can be categorized 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, and the majority of potential pathogens are quickly neutralized by the innate immune system before they can cause, for example, a notable infection. The adaptive immune system responds to molecular structures, called antigens, of the invading organism. Unlike the innate immune system, the adaptive immune system is highly specific for pathogens. Adaptive immunity can also provide long-lasting protection; for example, a person who recovers from measles is then protected from measles for the rest of their life. There are two types of adaptive immune responses, which include humoral and cell-mediated immune responses. In humoral immune responses, antibodies secreted into body fluids by B cells bind to antigens from pathogens, which leads to the elimination of the pathogen through various mechanisms, for example, complement-mediated lysis. In cell-mediated immune responses, T cells are activated that can destroy other cells. For example, if disease-related proteins are present in cells, they are proteolytically fragmented into peptides within the cell. Specific intracellular proteins then attach themselves to the antigens or peptides thus formed and transport them to the surface of the cell, where they are presented to the body's molecular defense mechanisms within T cells. Cytotoxic T cells recognize these antigens and kill the cells that bear them.
[0436] The terms "major histocompatibility complex (MHC)", "MHC molecule" or "MHC protein" refer to proteins that represent potential T cell epitopes that can bind to peptides resulting from proteolytic cleavage of protein antigens, transport them to the cell surface and present the peptides to specific cells, e.g., cytotoxic T lymphocytes or helper T cells. Human MHC is also referred to as the HLA complex. Thus, the terms "human leukocyte antigen (HLA) system", "HLA molecule" or "HLA protein" refer to the genetic complex that codes for MHC proteins in humans. The term MHC is referred to as the "H-2" complex in mouse species. It will be understood by those skilled in the art that the terms "major histocompatibility complex (MHC)", "MHC molecule", "MHC protein" and "human leukocyte antigen (HLA) system", "HLA molecule", "HLA protein" are used interchangeably herein.
[0437] HLA proteins are classified into two types, called HLA class I and HLA class II. The structures of the proteins of the two HLA classes are very similar, but they have completely different functions. HLA class I proteins are present on the surface of almost all cells of the body, including the majority of tumor cells. HLA class I proteins are usually loaded with antigens originating from endogenous proteins or pathogens present inside the cell, and 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. HLA class II proteins mainly present peptides processed from external antigen sources, such as the outside of cells, to helper T cells. The majority of peptides bound by HLA class I proteins originate from cytoplasmic proteins produced in healthy host cells of the organism itself, and do not usually stimulate an immune response.
[0438] HLA class I molecules (Figure 1) consist of two non-covalently linked polypeptide chains, the HLA-encoded α-chain (heavy, 44-47 kD) and a non-HLA-encoded subunit (12 kD) called β2 microglobulin (or β2m). The α-chain has three extracellular domains, α1, α2 and α3, and a transmembrane domain, and the α1 and α2 regions can bind peptides of about 7-13 amino acids (e.g., about 8-11 amino acids, or 9 or 10 amino acids). HLA class I molecules bind peptides with appropriate binding motifs and present them to cytotoxic T lymphocytes. HLA class I heavy chains can be the protein product of the HLA-A allele, also called HLA-A monomer, or the protein product of the HLA-B allele (also HLA-B monomer) or the protein product of the HLA-C allele (HLA-C monomer), each of which forms a complex with β-2-microglobulin. α1 rests on the non-HLA protein β2m; β2m is encoded by the beta2-microglobulin gene located on human chromosome 15. The α3 domain connects with the transmembrane region, which connects the HLA class I molecule to the cell membrane. The peptide to be presented is held at the bottom of the peptide-binding groove in the central region of the α1 / α2 heterodimer (a molecule composed of two non-identical subunits). HLA class IA, HLA class IB, or HLA class IC are highly polymorphic. Each of the HLA class 1-A genes (referred to as HLA-A genes), HLA class 1-B genes (referred to as HLA-B genes) and HLA class 1-C genes (referred to as HLA-C genes) contains eight exons, with exon 1 encoding the leader peptide, exons 2 and 3 encoding the α1 and α2 domains, exon 5 encoding the transmembrane region, and exons 6 and 7 encoding the cytoplasmic tail. Polymorphisms in exon 2 and exon 3 are responsible for the peptide binding specificity of each class 1 molecule. The HLA class IB genes (HLA-B) have many possible variations, expression patterns and antigens presented.This group is subdivided into those encoded within the HLA locus, e.g., HLA-E, HLA-F, HLA-G, and non-stress ligands, e.g., ULBP, Rae1, and H60. The antigens / ligands for many of these molecules remain unknown, but they can interact with CD8+ T cells, NKT cells, and NK cells, respectively.
[0439] In some embodiments, the present disclosure utilizes non-classical HLA class IE alleles. HLA-E molecules are recognized by natural killer (NK) cells and CD8+ T cells. HLA-E is expressed in almost all tissues, including lung, liver, skin and placental cells. HLA-E expression is also detected in solid tumors (e.g., osteosarcoma and melanoma). HLA-E molecules bind to TCRs expressed on CD8+ T cells, resulting in T cell activation. HLA-E is also known to bind to CD94 / NKG2 receptors expressed on NK cells and CD8+ T cells. CD94 can pair with several different isoforms of NKG2 to form receptors with the potential to inhibit cell activation (NKG2A, NKG2B) or promote cell activation (NKG2C). HLA-E can bind peptides derived from amino acid residues 3-11 of the leader sequences of the majority of HLA-A, HLA-B, HLA-C, and HLA-G molecules, but not its own leader peptide. HLA-E has also been shown to present peptides derived from endogenous proteins, as well as HLA-A, HLA-B, and HLA-C alleles. Under physiological conditions, the association of CD94 / NKG2A with HLA-E loaded with peptides derived from HLA class I leader sequences usually induces inhibitory signals. Cytomegalovirus (CMV) exploits a mechanism to escape from NK cell immune surveillance by expressing the UL40 glycoprotein, which mimics the HLA-A leader. 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, playing a role in defense against CMV. Several studies have revealed several important functions of HLA-E in infectious diseases and cancer.
[0440] Peptide antigens are presented on the cell surface after they attach themselves to the molecules of HLA class I in the endoplasmic reticulum by competitive affinity binding. Here, the affinity of each peptide antigen is directly linked to its amino acid sequence and the presence of specific binding motifs at defined positions in the amino acid sequence. If the sequence of such peptides is known, it is possible to manipulate the immune system against diseased cells, for example, by using peptide vaccines.
[0441] MHC molecules are highly polymorphic, i.e., many MHC variants exist. Each variant is encoded by a variation in a gene that codes for a protein, and each such variant gene is called an allele. For humans, MHC is known as human leukocyte antigen (HLA) and involves three types of HLA class II molecules: DP, DQ, and DR. HLA class II peptide (Figure 1) has two chains, α and β, each with two domains - α1 and α2 and β1 and β2, and each chain has a transmembrane domain, α2 and β2, respectively, that connects the HLA class II molecule to the cell membrane. A heterodimer of α1 and β1 forms the peptide-binding groove. The most widely studied HLA-DR molecule has DRA and DRB, which correspond to the α and β domains, respectively. The DRB is diverse, and the DRA is nearly identical. Thus, the binding specificity of the DRB allele indicates the binding specificity of the corresponding HLA-DR. Each MHC protein has its own binding specificity, i.e., the set of peptide binding specificities for an MHC molecule can be different from that for another MHC molecule. Classical molecules present peptides to CD4+ lymphocytes. Nonclassical molecules, which are accessory factors with intracellular functions, are not exposed on the cell membrane but are located inside the membrane of lysosomes and usually load antigenic peptides onto classical HLA class II molecules.
[0442] In the HLA class II system, phagocytes such as macrophages and immature dendritic cells ingest entities by phagocytosis into phagosomes (although B cells show a more general endocytosis into endosomes), which fuse with lysosomes, where lysosomal acid enzymes cleave the ingested proteins into many different peptides. Autophagy is another source of HLA class II peptides. Due to physicochemical dynamics in molecular interactions with host-derived HLA class II variants encoded in the host genome, certain peptides exhibit immunodominance and are loaded onto HLA class II molecules. These are transported to the cell surface and externalized. The most studied subclasses of HLA class II genes are HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA-DQB1, HLA-DRA, and HLA-DRB1.
[0443] Presentation of peptides to CD4+ helper T cells by HLA class II molecules is necessary for immune responses to foreign antigens (Roche and Furuta, 2015). CD4+ T cells Once activated, they 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 other immune cells and induce their differentiation. HLA class II molecules are heterodimers of α and β chains that interact to form a peptide-binding groove that is more open than the HLA class I peptide-binding groove (Unanue et al., 2016). Peptide molecules bound to HLA class II are thought to have a 9-amino acid binding core, either N- or C-terminal, with flanking residues protruding out of the groove (Jardetzky et al., 1996;Stern et al., 1994). These peptides are typically 12–16 amino acids in length and often contain 3–4 anchor residues at positions P1, P4, P6 / 7, and P9 of the binding register (Rossjohn et al., 2015).
[0444] HLA alleles are expressed in a codominant manner, i.e., alleles (variants) inherited from both parents are expressed equally. For example, each person has two alleles of each of the three class I genes (HLA-A, HLA-B and HLA-C), and therefore can express six different types of HLA class II. At the HLA class II locus, each person inherits a pair of HLA-DP genes (DPA1 and DPB1, coding for the α and β chains), HLA-DQ (DQA1 and DQB1, for the α and β chains), one gene HLA-DRα (DRA1), and one or more genes HLA-DRβ (DRB1 and DRB3, DRB4 or DRB5). For example, HLA-DRB1 has almost more than 400 known alleles. This means that one heterozygous individual can inherit six or eight functional HLA class II alleles: three or more from each parent. Thus, HLA genes are highly polymorphic; many different alleles exist in individuals within different populations. Genes that code for HLA proteins have many possible variations, allowing each person's immune system to react to a wide range of foreign invaders. Some HLA genes have hundreds of identified versions (alleles), each of which is given a specific number. In some embodiments, HLA class I alleles are HLA-A * 02:01, HLA-B * 14:02, HLA-A * 23:01, HLA-E * 01:01 (non-classical). In some embodiments, the HLA class II allele is HLA-DRB * 01:01, HLA-DRB * 01:02, HLA-DRB * 11:01, HLA-DRB * 15:01, and HLA-DRB * It is 07:01.
[0445] The subject's subject-specific HLA allele or HLA genotype can be determined by any method known in the art.In an exemplary embodiment, the HLA genotype is determined by any method described in International Patent Application No. PCT / US2014 / 068746, published on June 11, 2015 as W02015085147, which is incorporated herein by reference in its entirety. Briefly, the method includes determining a polymorphism genotype, which may include generating an alignment of reads extracted from a sequencing dataset to a genetic reference set comprising allelic variants of the polymorphic gene; determining a first posterior probability or posterior probability-derived score for each allelic variant in the alignment; identifying an allelic variant with a maximum first posterior probability or posterior probability-derived score as a first allelic variant; identifying one or more overlapping reads that are aligned with the first allelic variant and one or more other allelic variants; determining a second posterior probability or posterior probability-derived score for the one or more other allelic variants using a weighting factor; identifying a second allelic variant by selecting an allelic variant with a maximum second posterior probability or posterior probability-derived score, wherein the first allelic variant and the second allelic variant define a genotype of the polymorphic gene; and providing an output of the first allelic variant and the second allelic variant.
[0446] In some embodiments, the MHC class II peptide:antigenic peptide binding and presentation prediction methods described herein are capable of predicting binders from a large repertoire 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-matched peptides. In some embodiments, the large database of mass spectrometry-validated HLA-matched peptides is 1.2×10 6The large database of HLA-matched peptides validated by mass spectrometry includes more than 150 HLA alleles, including both MHC class I and class II allelic subtypes. In some embodiments, the database includes at least 95% of the US population of HLA-I and HLA-II (DR subtypes).
[0447] As described herein, there is abundant evidence in both animals and humans that mutant epitopes are effective in inducing immune responses, and that cases of spontaneous tumor regression or long-term survival correlate with CD8+ T cell responses to mutant epitopes, and that "immune editing" can be followed to alter the expression of dominant mutant antigens in mice and humans.
[0448] Sequencing techniques have revealed that each tumor contains numerous patient-specific mutations that alter the protein-coding content of genes. Such mutations create altered proteins ranging from single amino acid changes (caused by missense mutations) to the addition of long stretches of novel amino acid sequence due to frameshifts, read-through of termination codons or translation of intronic regions (novel open reading frame mutations; neoORFs). These mutant proteins are beneficial targets of the host immune response against the tumor because, unlike the native proteins, they are not subject to the immune-debilitating effects of self-tolerance. Thus, mutant proteins are more likely to be immunogenic and more specific for 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.
[0449] In some embodiments, the algorithms driving the prediction methods can be further utilized for mutation calling on the peptide. In some embodiments, the prediction methods can be used to determine the driver mutation status and / or RNA expression status and / or truncation predictions in the peptide.
[0450] The term "T cells" includes CD4+ T cells and CD8+ T cells. The term T cells also includes both T helper type 1 T cells and T helper type 2 T cells. T cells, as used herein, are generally classified by function and cell surface antigens (cluster classifier antigens, or CDs) that facilitate the binding of the T cell receptor to antigens, and are also divided into two major classes: helper T (TH) cells and cytotoxic T lymphocytes (CTLs).
[0451] Mature helper T (TH) cells express the surface protein CD4 and are referred to as CD4+ T cells. After T cell development, mature naive T cells leave the thymus and begin to spread throughout the body, including lymph nodes. Naive T cells are T cells that have never been exposed to the antigen to which they are programmed to respond. Like all T cells, naive T cells express the T cell receptor-CD3 complex. The T cell receptor (TCR) consists of both constant and variable regions. The variable region determines which antigens the T cell can respond to. CD4+ T cells have TCRs with affinity for MHC class II, and proteins and CD4 are involved in determining MHC affinity during maturation in the thymus. MHC class II proteins are generally only found on the surface of specialized antigen-presenting cells (APCs). Specialized antigen-presenting cells (APCs) are primarily dendritic cells, macrophages, and B cells, with dendritic cells being the only cell group that constitutively (at all times) express MHC class II. Some APCs, such as follicular dendritic cells, also bind native (or unprocessed) antigens to their surface, but unprocessed antigens do not interact with T cells and are not involved in T cell activation. Peptide antigens that bind to HLA class I proteins are generally shorter than peptide antigens that bind to HLA class II proteins.
[0452] Cytotoxic T lymphocytes (CTL) are cytotoxic T cells, cytolytic T cells, CD8+ T cells, also known as killer T cells, refer to lymphocytes that induce apoptosis in targeted cells. CTLs form antigen-specific conjugates with target cells through the interaction of the TCR with processed antigens (Ag) on the target cell surface, resulting in apoptosis of the targeted cells. The apoptotic bodies are eliminated by macrophages. The term "CTL response" is used to refer to the primary immune response mediated by CTL cells. Cytotoxic T lymphocytes have both T cell receptors (TCRs) and CD8 molecules on their surface. T cell receptors can recognize and bind to peptides complexed with molecules of HLA class I. Each cytotoxic T lymphocyte expresses a unique T cell receptor that can bind to a specific MHC / peptide complex. The majority of cytotoxic T cells express a T cell receptor (TCR) that can recognize a specific antigen. In order for the TCR to bind to HLA class I molecules, the TCR must be accompanied by a glycoprotein called CD8 that binds to the constant portion of the HLA class I molecules. These T cells are therefore referred to as CD8+ T cells. The affinity of CD8 for MHC molecules keeps the T cell and the target cell in close association during antigen-specific activation. Once activated, CD8+ T cells are recognized as T cells and are generally classified as having a defined cytotoxic role within the immune system. However, CD8+ T cells also have the ability to produce several cytokines.
[0453] A "T cell receptor (TCR)" is a cell surface receptor involved in the activation of T cells in response to the presentation of antigen. TCRs are generally composed of two chains, alpha and beta, which assemble to form a heterodimer and associate with the CD3 transducing subunit to form the T cell receptor complex present on the cell surface. Each alpha and beta chain of the TCR consists of an immunoglobulin-like N-terminal variable (V) and constant (C) region, a hydrophobic transmembrane domain, and a short intracytoplasmic region. With respect to immunoglobulin molecules, the variable regions of the alpha and beta chains are generated by V(D)J recombination, creating a large diversity of antigen specificity within a population of T cells. However, in contrast to immunoglobulins that recognize intact antigens, T cells are activated by processed peptide fragments associated with MHC molecules, introducing extra elements for antigen recognition by T cells, known as MHC restriction. Recognition of MHC mismatches between donor and recipient through the T cell receptor leads to T cell proliferation and the potential development of GVHD. Normal surface expression of the TCR has been shown to depend on the coordinated synthesis and assembly of all seven components of the complex (Ashwell and Klusner 1990). Inactivation of TCRα or TCRβ results in Loss of the TCR from the surface of T cells can result, preventing recognition of alloantigens and therefore GVHD, but TCR destruction generally results in loss of the CD3 signaling component, altering the means of further T cell expansion.
[0454] The term "HLA peptidome" refers to a pool of peptides that specifically interact with a particular HLA class and may include thousands of different sequences. HLA peptidome includes peptide diversity derived from both normal and abnormal proteins expressed in cells. Therefore, HLA peptidome can be studied to identify cancer-specific peptides, for the development of tumor immunotherapeutics, and as a source of information on protein synthesis and degradation schemes in cancer cells. In some embodiments, HLA peptidome is a pool of soluble HLA peptides (sHLA). In some embodiments, HLA peptidome is a pool of membrane-bound HLA (mHLA).
[0455] "Antigen-presenting cells" or "APCs" include professional antigen-presenting cells (e.g., B lymphocytes, macrophages, monocytes, dendritic cells, Langerhans cells), as well as other antigen-presenting cells (e.g., keratinocytes, endothelial cells, astrocytes, fibroblasts, oligodendrocytes, thymic epithelial cells, thyroid epithelial cells, glial cells (brain), pancreatic beta cells, and vascular endothelial cells). "Antigen-presenting cells" or "APCs" are cells that express major histocompatibility complex (MHC) molecules and can present foreign antigens complexed with MHC on their surface.
[0456] Monoallelic HLA cell lines A single allele cell line expressing a single HLA class I allele, a single pair of HLA class II alleles, or a single pair of HLA class I alleles and HLA class II alleles can be generated by transducing or transfecting a suitable cell population with a polynucleic acid, e.g., a vector, encoding a single HLA allele (Figure 2). Suitable cell populations include, for example, HLA class I-deficient cell lines that exogenously express a single HLA class I allele, HLA class II-deficient cell lines that express a single exogenous pair of HLA class II alleles, or class I and class II-deficient cell lines that exogenously express a single pair of HLA class I and / or class II alleles. As an exemplary embodiment, the HLA class I-deficient B cell line is B721.221. However, it is clear to one skilled in the art that other cell populations that are deficient in HLA class I and / or HLA class II can be generated. A typical method for deleting / inactivating endogenous HLA class I or HLA class II genes includes, for example, CRISPR-Cas9-mediated genome editing in THP-1 cells. In some embodiments, the population of cells is a professional antigen-presenting cell, such as a macrophage, B cell, and dendritic cell. The cell can be a B cell or a dendritic cell. In some embodiments, the cell is a tumor cell or a cell derived from a tumor cell line. In some embodiments, the cell is isolated from a patient. In some embodiments, the cell contains an infectious agent or a part thereof. In some embodiments, the population of cells comprises at least 107 cells. In some embodiments, the population of cells is further modified, for example, by increasing or decreasing the expression and / or activity of at least one gene. In some embodiments, the gene encodes a member of the immunoproteasome. The immunoproteasome is known to be involved in processing HLA class I-binding peptides and includes LMP2 (β1i), MECL-1 (β2i), and LMP7 (β5i) subunits. The immunoproteasome can also be induced by interferon-gamma.Thus, in some embodiments, the population of cells can be contacted with one or more cytokines, growth factors, or other proteins. The cells can be stimulated with inflammatory cytokines, such as interferon-gamma, IL-10, IL-6, and / or TNF-α. The population of cells can also be subjected to various environmental conditions, such as stress (heat stress, oxygen deprivation, glucose starvation, DNA damaging agents, etc.). In some embodiments, the cells are contacted with one or more of chemotherapeutic drugs, radiation, targeted therapy, or immunotherapy. Thus, the methods disclosed herein can be used to test the effects of various genes or conditions on the processing and presentation of HLA peptides. In some embodiments, the conditions used are selected to match the condition of the patient for which the population of HLA-peptides is identified.
[0457] The single HLA-alleles of the present disclosure can be encoded and expressed using a virus-based system (e.g., an adenovirus system, an adeno-associated virus (AAV) vector, a poxvirus, or a lentivirus). Plasmids that can be used for adeno-associated virus delivery, adenovirus delivery, and lentivirus delivery have been previously described (see, e.g., U.S. Patent Nos. 6,955,808 and 6,943,019, and U.S. Patent Application No. 20080254008, which are incorporated herein by reference). Among the vectors that can be used to implement the present disclosure, integration into the host genome of cells is possible using retroviral gene transfer methods, often resulting in long-term expression of the inserted transgene. In an 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 retroviruses can be modified by incorporating foreign envelope proteins, increasing the potential target population of target cells. Retroviruses can be engineered to allow the conditional expression of inserted transgenes, so that lentiviruses infect only certain cell types. Cell type-specific promoters can be used to target expression in specific cell types. Lentivirus vectors are retrovirus vectors (and therefore both lentivirus vectors and retrovirus vectors can be used to implement the present disclosure). In addition, lentivirus vectors can transduce or infect non-dividing cells, and generally produce high virus titers.
[0458] The choice of retroviral gene transfer system may depend on the target tissue. Retroviral vectors are composed of cis-acting long terminal repeats that have the capacity to package up to 6-10 kb of foreign sequence. The minimal cis-acting LTRs are sufficient for replication and vector packaging, and are therefore used to integrate the desired nucleic acid into the target cell to provide permanent expression. Widely used retroviral vectors that can be used in the practice of the present disclosure include those based on murine leukemia virus (MuLV), gibbon ape leukemia virus (GaLV), simian immunodeficiency virus (SIV), human immunodeficiency virus (HIV), and combinations thereof (see, e.g., Buchscher et al., (1992) J. Virol. 66: 2731-2739; Johann et al., (1992) J. Virol. 66: 1635-1640; Sommnerfelt ... 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 / (See US94 / 05700). Minimal non-primate lentiviral vectors, such as those based on the equine infectious anemia virus (EIAV), are also useful in practicing the present disclosure (see, e.g., Balagaan, (2006) J Gene Med; 8: 275-285, Published online 21 November 2005 in Wiley InterScience DOI: 10.1002 / jgm.845). The vectors include cytomegalovirus vectors that drive expression of the target gene. (CMV) promoter.Thus, the present disclosure contemplates viral vectors, including retroviral and lentiviral vectors, among the vectors useful in practicing the present disclosure.
[0459] Any HLA allele can be expressed in the cell population. In an exemplary embodiment, the HLA allele is an HLA class I allele. In some embodiments, the HLA class I allele is an HLA-A allele or an HLA-B allele. In some embodiments, the HLA allele is an HLA class II allele. The sequences of HLA class I and class II alleles can be found in the IPD-IMGT / HLA database. Exemplary HLA alleles include, but are not limited to, HLA-A, HLA-B, HLA-C, HLA-D, HLA-E, HLA-F ... * 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 * One example is 07:01.
[0460] In some embodiments, the HLA allele is selected to correspond to the genotype of interest.In some embodiments, the HLA allele is a mutant HLA allele, and can be an allele that does not naturally occur in diseased patients or an allele that naturally occurs.Another advantage of the method disclosed herein is the identification of HLA-binding peptides for the HLA alleles associated with various disorders as well as the alleles that occur at low frequencies.Thus, in some embodiments, the method provided herein can identify HLA alleles even if they occur at a frequency of less than 1% in a population, for example, in a Caucasian population.
[0461] In some embodiments, the nucleic acid sequence encoding the HLA allele further comprises an affinity acceptor tag that can be used to immunopurify the HLA-protein. Suitable tags are well known in the art. In some embodiments, the affinity acceptor tag is a poly-histidine tag, a poly-histidine-glycine tag, a poly-arginine tag, a poly-aspartic acid tag, a poly-cysteine tag, a poly-phenylalanine, 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 tag, a phospholipid ... Acetyltransferase (CAT) tag, Choline-binding domain (CBD) tag, Chitin-binding domain (CBD) tag, Cellulose-binding domain (CBP) tag, Dihydrofolate reductase (DHFR) tag, Galactose-binding protein (GBP) tag, Maltose-binding protein (MBP), Glutathione-S-transferase (GST), Glu-Glu (EE) tag, Human influenza hemagglutinin (HA) tag, Horseradish peroxidase (HRP) tag, NE-tag, HSV tag, Ketosteroid isomerase (KSI) tag, KT3 tag, LacZ tag, Luciferase tag, NusA tag, PDZ domain tag, AviTag, Calmodulin-tag, E-tag, S-tag, SBP-tag, Softag1, Softag3, TC tag, VSV-tag, Xpress tag, Isopeptag, SpyTag, SnoopTag, ProfinityeXact-tag, Protein C-tag, S1-tag, S-tag, Biotin-Carboxy Carrier Protein (BCCP)-tag, Green Fluorescent Protein (GFP)-tag, Small Ubiquitin-Like Modifier (SUMO)-tag, Tandem Affinity Purification (TAP)-tag, HaloTag, Nus-tag, Thioredoxin-tag, Fc-tag, CYD-tag, HPC-tag, TrpE-tag, Ubiquitin-tag, VSV-G-epitope tag derived from the Vescular Stomatis virus glycoprotein, or V5-tag derived from a small epitope (Pk) found in the P and V proteins of the paramyxovirus Simian Virus 5 (SV5). In some embodiments, the affinity acceptor tag is an "epitope tag", which is a type of peptide tag that adds a recognizable epitope (antibody-binding site) to an HLA-protein, resulting in 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, which bind to IgG. In some embodiments, affinity acceptor tags include biotin acceptor peptide (BAP) or human influenza hemagglutinin (HA) peptide sequences. Numerous other tag moieties are known to and can be envisioned by those of skill in the art and are contemplated herein. Any peptide tag can be used as long as it can be expressed as an element of an affinity acceptor-tagged HLA-peptide complex.
[0462] The methods provided herein include isolating HLA-peptide complexes from cells transfected or transduced with affinity pull-down of HLA constructs (Figure 3). In some embodiments, the complexes can be isolated using standard immunoprecipitation techniques known in the art using commercially available antibodies. The cells can be first lysed. HLA class I-peptide complexes can be isolated using HLA class I specific antibodies such as W6 / 32 antibody, while HLA class II-peptide complexes can be isolated using HLA class II specific antibodies such as M5 / 114.15.2 monoclonal antibody. In some embodiments, a single (or pair of) HLA alleles is expressed as a fusion protein with a peptide tag, and the HLA-peptide complexes are isolated using a binding molecule that recognizes the peptide tag.
[0463] The method further comprises isolating peptides from said HLA-peptide complexes and sequencing the peptides. The peptides are isolated from the complexes by any method known to those of skill in the art, such as acid elution. While any sequencing method can be used, in some embodiments, a method using mass spectrometry is utilized, such as liquid chromatography-mass spectrometry (LC-MS or LC-MS / MS, or alternatively HPLC-MS or HPLC-MS / MS). These sequencing methods are well known to those of skill in the art and are described in Medzihradszky KF and Reviewed in Chalkley RJ. Mass Spectrom Rev. 2015 Jan-Feb; 34 (1): 43-63.
[0464] In some embodiments, the population of cells expresses one or more endogenous HLA alleles. In some embodiments, the population of cells is a population of cells that lack one or more engineered endogenous HLA class I alleles. In some embodiments, the population of cells is a population of cells that lacks engineered endogenous HLA class I alleles. In some embodiments, the population of cells is a population of cells that lacks one or more engineered endogenous HLA class II alleles. In some embodiments, the population of cells is a population of cells that lacks engineered endogenous HLA class II alleles or a population of cells that lacks engineered endogenous HLA class I alleles and endogenous HLA class II alleles. In some embodiments, the population of cells comprises cells that have been enriched or sorted, for example, by fluorescence-activated cell sorting (FACS). In some embodiments, the population of cells is sorted using fluorescence-activated cell sorting (FACS). In some embodiments, the population of cells is pre-sorted by FACS for cell surface expression of either HLA class I or class II or both HLA class I and class II. For example, FACS can be used to sort a population of cells for cell surface expression of HLA class I alleles, HLA class II alleles, or a combination thereof.
[0465] Methods for preparing personalized cancer vaccines Once a cancer-specific mutation is identified, and thus a mutation exists in DNA in cancer cells but not in normal cells of the same human subject, and the mutation leads to one or more amino acid changes in the protein encoded by the DNA, the mutation can be targeted by the host immune response. The natural immune response can be directed to the mutant protein, thereby leading to the destruction of the cancer cells expressing the protein. Due to the natural tolerance response and the immunocompromised environment in cancerous tissues, immunotherapy is a clinical route that tries to strengthen such immune response to override the body's tolerance and immunosuppressive effects. Therefore, proteins or peptides containing the above mutations are suitable candidates for immunotherapy.
[0466] The mutant proteins are ingested by professional phagocytes acting as antigen-presenting cells (APCs), chopped up, and displayed on the cell surface as antigen-presenting complexes containing major histocompatibility complex (MHC) proteins as antigens for T-cell activation. Human MHC proteins are referred to as human leukocyte antigens, HLA. MHC proteins can be MHC-class I or class II proteins, and several functional differences result from the presentation of peptides by either class I or class II MHC proteins (HLA class I and HLA class II proteins), but one notable difference lies in the fact 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 and lead to sustained immune responses. CD8+ T cells are essential to the task of cell-by-cell elimination of diseased cells such as infected or tumor cells. CD4+ T cells have a more sustained effect when activated, the most important being the generation of immune memory. CD4 subsets are differentially recruited depending on the type of immunological threat, and multiple subsets with overlapping or different functions can be co-recruited. This helps balance the immunological response to pathogenic threats. In this regard, HLA class II peptide-mediated antigen presentation leads to a sustained and coordinated immune response. On the other hand, HLA class II binding to peptides can be promiscuous, and thus non-specific peptide binding and presentation to the immune system leads to ectopic immune responses such as autoimmunity.
[0467] In one aspect, the present disclosure provides a method for predicting peptides that may precisely pair or bind to specific HLA class II alpha and beta chain heterodimers, ensuring that high fidelity binding of the peptide to HLA class II proteins (composed of alpha and beta chain heterodimers) will present the particular peptide to T lymphocytes, thereby eliciting a specific immune response and avoiding any cross-reactivity or indiscriminate immunity.
[0468] In one aspect, the present disclosure provides a method for predicting peptides that can accurately bind to a particular HLA class II protein, such that peptides can be used to activate a more sustained and robust immune response when administered therapeutically to a subject expressing the particular cognate HLA class II protein, due to the ability of the HLA class II protein to activate CD4+ T cells and stimulate immune memory. In some embodiments, a given peptide predicted to bind with high specificity to an HLA class II protein is a peptide that contains a mutation, the mutation being prevalent in the cancer or tumor cells of the subject; whereas the same HLA class II protein predicted to bind to the mutant peptide (a) does not bind to the corresponding unmutated wild-type peptide, or (b) binds with a lower affinity separately compared to its affinity for binding to the mutant peptide of the subject. Preferential binding of HLA to mutant peptides is advantageous for the development of immunotherapeutics, since cells expressing wild-type peptides are spared from immune attack by T cells reactive to peptides presented by HLA. In some embodiments, a peptide predicted to bind specifically to an HLA class II protein is a peptide with post-translational modification. Exemplary post-translational modifications include, but are not limited to, phosphorylation, ubiquitination, dephosphorylation, glycosylation, methylation, or acetylation. In some embodiments, the predicted peptides are subjected to post-translational modifications before being used for immunotherapy.
[0469] In some embodiments, the immunotherapy methods and strategies disclosed herein may also be applicable to suppressing unwanted immune activation, such as in autoimmune reactions.Specifically, peptides identified as potential binders for specific HLA subtypes can be tailored to bind to specific HLA molecules and induce tolerance rather than triggering an immunogenic response.
[0470] In one aspect, the present invention provides a method for immunotherapy that is tailored or personalized for a specific subject.Every subject or patient expresses a specific array of HLA class I and HLA class II proteins.HLA typing is a well-known technique that allows determining the specific repertoire of HLA proteins expressed by a subject.Once the HLA heterodimer expressed by a specific subject is known, the improved, sophisticated and reliable method described herein for predicting the peptide that can bind to a specific HLA class II alpha chain and beta chain heterodimer with high fidelity ensures that a specific immune response that is tailored specifically for a subject can be generated.
[0471] Genes encoding HLA heterodimers are highly polymorphic, with more than 4,000 HLA class II allele variants identified across the human population. From maternal and paternal HLA haplotypes, individuals can inherit different alleles for each of the HLA class II loci, with each HLA class II heterodimer being composed of an α-chain and a β-chain. Due to the large number of α- and β-chain pairing combinations, especially for HLA-DP and HLA-DQ alleles, the population of possible HLA heterodimers is highly complex. HLA class II heterodimers are translocated to the endoplasmic reticulum (ER) and assembled into stable complexes with the invariant chain (Ii) derived from the protein CD74. Ii stabilizes the class II complex by allowing proper protein folding and also allows export of HLA class II heterodimers to endosomal / lysosomal compartments. Within these HLA class II loading compartments, Ii is proteolytically cleaved by cathepsins into a placeholder peptide called CLIP. CLIP is then exchanged for higher affinity peptides by the chaperone HLA-DM, a non-classical HLA class II heterodimer, in a low pH environment, and the high affinity peptide-loaded HLA class II complexes are then targeted to the trans-Golgi and ultimately to the cell surface for display to CD4+ T cells.
[0472] Each HLA heterodimer is estimated to bind thousands of peptides with allele-specific binding preferences. In fact, each HLA allele is estimated to bind approximately 1,000-10,000 unique peptides and present them to T cells. Given such diversity of HLA binding, accurate prediction of whether a peptide is likely to bind to a particular HLA allele is extremely challenging. Due to the heterogeneity of α-chain and β-chain pairing, the complexity of the data that limits the ability to confidently assign core binding epitopes, and the lack of immunoprecipitation-grade allele-specific antibodies required for high-resolution biochemical analysis, very little is known about the allele-specific peptide-binding characteristics of HLA class II molecules. Furthermore, analysis of peptide epitopes derived from a given HLA allele creates uncertainty when multiple HLA alleles are presented on the cell surface.
[0473] Prediction of candidate neoantigens has been performed primarily for HLA class I epitopes (as experimental data for class I prediction algorithms are available compared to class II), but CD4+ T cell responses are often observed in both preclinical and clinical personalized neoantigen vaccination trials. These findings demonstrate that HLA class II epitope processing and presentation may also play a pivotal role in cancer treatment. HLA class II prediction algorithms exist but are inaccurate due to the open-ended peptide-binding groove of HLA class II heterodimers allowing longer peptides (generally 15-40 amino acids) to bind, thereby increasing the heterogeneity and complexity of epitope presentation. Thus, further research is needed to better understand the HLA class II peptide-binding core and the cellular processes involved in class II epitope processing and presentation. The proteomics field is currently limited by the complexity of HLA class II heterodimer formation and the availability of immunoprecipitation-grade antibodies to isolate HLA class II-peptide complexes. To overcome these challenges, we developed a single allele HLA profiling workflow that relies on LC-MS / MS for characterization of allele-specific HLA class II-ligandome for class II epitope prediction method. The following definitions are supplementary to those skilled in the art and are directed to this application, and are not attributable to any related or unrelated, e.g., any co-owned patent or application. Any method and material similar or equivalent to the methods and materials described herein can be used in the practice of testing the present disclosure, but exemplary materials and methods are described herein. Therefore, the terminology used herein is only for describing specific embodiments and is not limiting.
[0474] Disclosed herein is a method for preparing personalized cancer vaccine.The method for preparing personalized cancer vaccine can include: identifying peptide sequences with the mutations expressed in cancer cells of a target; inputting the amino acid position information of the identified peptide sequences into a machine learning HLA-peptide presentation prediction model using a computer processor to generate a set of presentation predictions for the identified peptide sequences, wherein each presentation prediction represents the probability that one or more proteins encoded by the class II MHC alleles of cancer cells of a target present a given sequence among the identified peptide sequences; and selecting a subset of the identified peptide sequences based on the set of presentation predictions to prepare personalized cancer vaccine.
[0475] In some embodiments, one or more results obtained from the methods described herein provide a quantitative value or values indicative of one or more of the following: likelihood of diagnostic accuracy, likelihood of the presence of a condition in a subject, likelihood of a condition occurring in a subject, likelihood of success of a particular treatment, or any combination thereof. In some embodiments, the methods described herein can predict the risk or likelihood of a condition occurring. In some embodiments, the methods described herein can be an early diagnostic indicator of the occurrence of a condition. In some embodiments, the methods described herein can confirm the diagnosis or existence 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 effectiveness of a treatment for a condition in a subject.
[0476] Methods for identifying MHC-II peptides In one aspect, a method is provided herein for identifying one or more peptides presented by MHC-II proteins for immune activation. In some embodiments, the one or more peptides comprise an epitope. In some embodiments, the method involves computational prediction of the likelihood that a specific epitope is presented by MHC-II proteins. In some embodiments, the method involves computational prediction of the specificity of the epitope for MHC-II presentation. In some embodiments, the computational prediction method involves evaluation of peptide-MHC interactions. In some embodiments, the computational prediction method involves prediction of the allele specificity of the peptide for antigen presentation. In some embodiments, the computational prediction method involves incorporation of bioinformatics information, such as nucleotide sequences, structural motifs of biomolecules, protein-protein interaction features, and functional potency, such as immunogenicity. In some embodiments, the computational prediction method involves machine learning. A large number of immunoinformatics methods have been developed to predict peptide-MHC interactions based on machine learning techniques 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 for both MHC class I and class II. These methods can be divided into two categories: intra-allelic (allele-specific) methods and trans-allelic (pan-specific) methods. Intra-allelic methods train on a specific MHC molecule against a limited set of experimental peptide-binding data and apply peptide-binding predictions to that molecule. The extreme polymorphism of MHC molecules and the existence of thousands of allelic variants, coupled with the lack of sufficient experimental binding data, make it impossible to build predictive models for each allele.Therefore, peptide-binding data expanding across many alleles or species can be used to develop trans-allelic and general-purpose methods, such as NetMHCIIpan (Karosiene E et al., NetMHCIIpan-3.0, a common pan-specific MHC class II prediction method including all three human MHC class II isotypes, HLA-DR, HLA-DP. and HLADQ. Immunogenetics (2013) 65(10):711-24), and TEPITOPEpan (Zhang L, et al., TEPITOPEpan: extending TEPITOPE for peptide binding prediction covering over 700 HLA-DR molecules. PLoS One (2012) 7(2):e30483). A similar method for I is available.
[0477] In some embodiments, the peptide sequence may not be expressed in normal cells of the subject. In some embodiments, any and all cells of the subject may not be cancer cells. Cancer cells include, but are not limited to, thyroid cancer, adrenal cortical 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, colon and rectal cancer, uterine cancer, esophagus cancer, Ewing's family of tumors (e.g., Ewing's sarcoma), eye cancer, gallbladder cancer, gastrointestinal carcinoid tumors, gastrointestinal stromal tumors, gestational trophoblastic disease, hairy cell leukemia, Hodgkin's disease, Kaposi's sarcoma, kidney cancer, and the like. The antibody may be produced by different cancers, including cancer of the larynx or hypopharynx, acute lymphocytic leukemia, acute myeloid leukemia, childhood leukemia, chronic lymphocytic leukemia, chronic myeloid leukemia, liver cancer, lung cancer, pulmonary carcinoid tumors, non-Hodgkin's lymphoma, male breast cancer, malignant mesothelioma, multiple myeloma, myelodysplastic syndromes, myeloproliferative disorders, nasal cavity and paranasal sinus cancer, nasopharyngeal cancer, neuroblastoma, oral cavity or oropharyngeal cancer, osteosarcoma, ovarian cancer, pancreatic cancer, penile cancer, pituitary tumors, prostate cancer, retinoblastoma, rhabdomyosarcoma, salivary gland cancer, sarcoma (adult soft tissue cancer), melanoma skin cancer, non-melanoma skin cancer, stomach cancer, testicular cancer, thymic cancer, uterine cancer (e.g., uterine sarcoma), vaginal cancer, vulvar cancer, or Waldenstrom's macroglobulinemia.
[0478] The identifying step can include comparing the DNA, RNA or protein sequence from the cancer cell of the subject with the DNA, RNA or protein sequence from the normal cell of the subject.The DNA, RNA or protein sequence from the cancer cell of the subject can be different from the DNA, RNA or protein sequence from the normal cell of the subject.The identifying step can identify the nucleic acid variant with high sensitivity.
[0479] The machine learning HLA-peptide presentation prediction model may include a plurality of predictor variables identified based on at least training data. The training data may include: sequence information of peptide sequences that are presented by HLA proteins expressed in cells and identified by mass spectrometry; training peptide sequence information including amino acid position information, the training peptide sequence information being associated with HLA proteins expressed in cells; and a function that represents 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.
[0480] In some embodiments, the training data may further include structured data, time series data, unstructured data, and relational data. Unstructured data may include audio data, image data, video, mechanical data, electrical data, chemical data, and any combination thereof for use in accurately simulating or training robotics or simulations. 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 customer systems, enterprise systems, operational systems, websites, web-accessible application program interfaces (APIs), or any combination thereof. This may be done by any method where a user inputs a file or other data format into the software or system.
[0481] In some embodiments, the training data can be stored in a database. The database can be stored in a computer readable format. A computer processor can be configured to access the data stored in the computer readable memory. In some embodiments, the computer system can be used to analyze the data to obtain results. The results can be stored in a storage medium remotely or internally and communicated to personnel such as drug application specialists. In some embodiments, the computer system can be operatively coupled with components for communicating the results. The components for communication can include wired and wireless components. Examples of wired communication components can include Universal Serial Bus (USB) connections, coaxial cable connections, Ethernet cables such as Cat5 or Cat6 cables, fiber optic cables, or telephone lines. Examples of wireless communication components can include Wi-Fi receivers, components for accessing mobile data standards such as 3G or 4G LTE data signals, or Bluetooth receivers. In some embodiments, all of this data in storage media is collected and archived to build a data warehouse.
[0482] In some embodiments, the database comprises an external database. The external database may be a medical database, such as, but not limited to, Adverse Drug Reactions. Effects Database, AHFS Supplemental File, Allergen Picklist File, Average WAC Pricing File, Brand Probability File, Canadian Drug File v2、Comprehensive Price History、Controlled Substances File、Drug Allergy Cross-Reference File、Drug Application File、Drug Dosing & Administration Database、Drug Image Database v2.0 / Drug Imprint Database v2.0、Drug Inactive Date File、D...
Claims
[Claim 1] The invention described in the specification.
