{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,7]],"date-time":"2026-04-07T13:05:05Z","timestamp":1775567105922,"version":"3.50.1"},"reference-count":36,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,2,28]],"date-time":"2026-02-28T00:00:00Z","timestamp":1772236800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,4,7]],"date-time":"2026-04-07T00:00:00Z","timestamp":1775520000000},"content-version":"vor","delay-in-days":38,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["BMC Bioinformatics"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    Accurate prediction of MHC-peptide binding affinity remains a challenge for immunotherapeutic development. Existing methods struggle to jointly model functional semantics of polymorphic residues, evolutionary conservation constraints, and structural dynamic. We propose the Contrast learning-based Multi-feature Heterogeneous Subgraph model (CMHS) with sequence and structural representation. For sequence representation, we introduce LoRA fine-tuning to obtain the MHC-exclusive sequence representation from ESM2, then jointly BLOSUM50 to capture long-range functional dependencies and evolutionarily conserved residues. For structural representation, we use the biophysics-guided heterogeneous graph network. Constructing an MHC-peptide graph with a novel trainable Gaussian noise layer guided by crystallographic B-factors to dynamically simulate electron density uncertainty, coupled with a three-stage message-passing framework with subgraph aggregation, subgraph extraction and heterogeneous. Finally, to align sequence and graph representation spaces, we use contrastive learning to obtain a more comprehensive representation and to enhance the ability of model prediction. Evaluations on 16 HLA allele benchmarks show average SRCC improvements of 8.7\n                    <jats:inline-formula>\n                      <jats:tex-math>$$\\%$$<\/jats:tex-math>\n                    <\/jats:inline-formula>\n                    , with improvements of average AUC of 7.6\n                    <jats:inline-formula>\n                      <jats:tex-math>$$\\%$$<\/jats:tex-math>\n                    <\/jats:inline-formula>\n                    . This work establishes a new paradigm for predicting hypervariable immune interactions. The corresponding code can be founded in github.\n                  <\/jats:p>","DOI":"10.1186\/s12859-026-06407-1","type":"journal-article","created":{"date-parts":[[2026,2,28]],"date-time":"2026-02-28T12:32:47Z","timestamp":1772281967000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Multimodal learning on heterogeneous subgraphs and LLMs representation for MHC-peptide binding affinity prediction"],"prefix":"10.1186","volume":"27","author":[{"given":"Ruimeng","family":"Li","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ying","family":"Wang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Haozhou","family":"Li","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Biyi","family":"Zhou","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Qinke","family":"Peng","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"297","published-online":{"date-parts":[[2026,2,28]]},"reference":[{"issue":"1","key":"6407_CR1","doi-asserted-by":"publisher","first-page":"51","DOI":"10.1146\/annurev.immunol.17.1.51","volume":"17","author":"JW Yewdell","year":"1999","unstructured":"Yewdell JW, Bennink JR. Immunodominance in major histocompatibility complex class I-restricted t lymphocyte responses. Annu Rev Immunol. 1999;17(1):51\u201388.","journal-title":"Annu Rev Immunol"},{"issue":"10","key":"6407_CR2","doi-asserted-by":"publisher","first-page":"1194","DOI":"10.1038\/s41587-020-0505-4","volume":"38","author":"H Huang","year":"2020","unstructured":"Huang H, Wang C, Rubelt F, Scriba TJ, Davis MM. Analyzing the mycobacterium tuberculosis immune response by T-cell receptor clustering with GLIPH2 and genome-wide antigen screening. Nat Biotechnol. 2020;38(10):1194\u2013202.","journal-title":"Nat Biotechnol"},{"issue":"24","key":"6407_CR3","doi-asserted-by":"publisher","first-page":"3265","DOI":"10.1093\/bioinformatics\/btm471","volume":"23","author":"C Lundegaard","year":"2007","unstructured":"Lundegaard C, Lund O, Ke\u015fmir C, Brunak S, Nielsen M. Modeling the adaptive immune system: predictions and simulations. Bioinformatics. 2007;23(24):3265\u201375.","journal-title":"Bioinformatics"},{"issue":"2","key":"6407_CR4","doi-asserted-by":"publisher","first-page":"127","DOI":"10.1016\/j.artmed.2010.05.003","volume":"50","author":"W Zhang","year":"2010","unstructured":"Zhang W, Liu J, Niu Y. Quantitative prediction of MHC-II binding affinity using particle swarm optimization. Artif Intell Med. 2010;50(2):127\u201332.","journal-title":"Artif Intell Med"},{"issue":"49","key":"6407_CR5","doi-asserted-by":"publisher","first-page":"17408","DOI":"10.1021\/ja109042w","volume":"132","author":"B Gulbakan","year":"2010","unstructured":"Gulbakan B, Yasun E, Shukoor MI, Zhu Z, You M, Tan X, et al. A dual platform for selective analyte enrichment and ionization in mass spectrometry using aptamer-conjugated graphene oxide. J Am Chem Soc. 2010;132(49):17408\u201310.","journal-title":"J Am Chem Soc"},{"issue":"10","key":"6407_CR6","doi-asserted-by":"publisher","first-page":"1504","DOI":"10.1038\/s41588-021-00935-7","volume":"53","author":"Y Luo","year":"2021","unstructured":"Luo Y, Kanai M, Choi W, Li X, Sakaue S, Yamamoto K, et al. A high-resolution HLA reference panel capturing global population diversity enables multi-ancestry fine-mapping in HIV host response. Nat Genet. 2021;53(10):1504\u201316.","journal-title":"Nat Genet"},{"issue":"8","key":"6407_CR7","doi-asserted-by":"publisher","first-page":"796","DOI":"10.1371\/journal.pone.0000796","volume":"2","author":"M Nielsen","year":"2007","unstructured":"Nielsen M, Lundegaard C, Blicher T, Lamberth K, Harndahl M, Justesen S, et al. NetMHCpan, a method for quantitative predictions of peptide binding to any HLA-A and-B locus protein of known sequence. PLoS ONE. 2007;2(8):796.","journal-title":"PLoS ONE"},{"key":"6407_CR8","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1186\/s13073-016-0288-x","volume":"8","author":"M Nielsen","year":"2016","unstructured":"Nielsen M, Andreatta M. NetMHCpan-3.0; improved prediction of binding to MHC class I molecules integrating information from multiple receptor and peptide length datasets. Genome Med. 2016;8:1\u20139.","journal-title":"Genome Med"},{"issue":"9","key":"6407_CR9","doi-asserted-by":"publisher","first-page":"3360","DOI":"10.4049\/jimmunol.1700893","volume":"199","author":"V Jurtz","year":"2017","unstructured":"Jurtz V, Paul S, Andreatta M, Marcatili P, Peters B, Nielsen M. NetMHCpan-4.0: improved peptide-MHC class I interaction predictions integrating eluted ligand and peptide binding affinity data. J Immunol. 2017;199(9):3360\u20138.","journal-title":"J Immunol"},{"issue":"W1","key":"6407_CR10","doi-asserted-by":"publisher","first-page":"449","DOI":"10.1093\/nar\/gkaa379","volume":"48","author":"B Reynisson","year":"2020","unstructured":"Reynisson B, Alvarez B, Paul S, Peters B, Nielsen M. NetMHCpan-4.1 and netMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Res. 2020;48(W1):449\u201354.","journal-title":"Nucleic Acids Res"},{"issue":"1","key":"6407_CR11","doi-asserted-by":"publisher","first-page":"129","DOI":"10.1016\/j.cels.2018.05.014","volume":"7","author":"TJ O\u2019Donnell","year":"2018","unstructured":"O\u2019Donnell TJ, Rubinsteyn A, Bonsack M, Riemer AB, Laserson U, Hammerbacher J. MHCflurry: open-source class I MHC binding affinity prediction. Cell Syst. 2018;7(1):129\u201332.","journal-title":"Cell Syst"},{"issue":"4","key":"6407_CR12","doi-asserted-by":"publisher","first-page":"321","DOI":"10.1093\/bib\/bbaf321","volume":"26","author":"J Ma","year":"2025","unstructured":"Ma J, Wang Z, Tong C, Yang Q, Zhang L, Liu H. pMHChat, characterizing the interactions between major histocompatibility complex class II molecules and peptides with large language models and deep hypergraph learning. Brief Bioinform. 2025;26(4):321.","journal-title":"Brief Bioinform"},{"issue":"23","key":"6407_CR13","doi-asserted-by":"publisher","first-page":"4946","DOI":"10.1093\/bioinformatics\/btz427","volume":"35","author":"Y Hu","year":"2019","unstructured":"Hu Y, Wang Z, Hu H, Wan F, Chen L, Xiong Y, et al. Acme: pan-specific peptide-MHC class I binding prediction through attention-based deep neural networks. Bioinformatics. 2019;35(23):4946\u201354.","journal-title":"Bioinformatics"},{"key":"6407_CR14","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1186\/s12859-019-2892-4","volume":"20","author":"P Phloyphisut","year":"2019","unstructured":"Phloyphisut P, Pornputtapong N, Sriswasdi S, Chuangsuwanich E. MHCSeqNet: a deep neural network model for universal MHC binding prediction. BMC Bioinform. 2019;20:1\u201310.","journal-title":"BMC Bioinform"},{"key":"6407_CR15","doi-asserted-by":"crossref","unstructured":"Sefer E. Anomaly detection via graph contrastive learning. In: Proceedings of the 2025 SIAM International Conference on Data Mining (SDM). Society for Industrial and Applied Mathematics. 2025;51\u201360.","DOI":"10.1137\/1.9781611978520.6"},{"key":"6407_CR16","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2025.112563","volume":"172","author":"X Bi","year":"2025","unstructured":"Bi X, Ma W, Jiang H, Lu W, Nie J, Wei Z, et al. Protein interaction pattern recognition using heterogeneous semantics mining and hierarchical graph representation. Pattern Recogn. 2025;172:112563.","journal-title":"Pattern Recogn"},{"key":"6407_CR17","doi-asserted-by":"publisher","first-page":"22","DOI":"10.1109\/TNNLS.2025.3599927","volume":"37","author":"X Bi","year":"2025","unstructured":"Bi X, Ma W, Jiang H, Lu W, Wei Z, Zhang S. SSPPI: cross-modality enhanced protein-protein interaction prediction from sequence and structure perspectives. IEEE Trans Neural Netw Learn Syst. 2025;37:22\u201336.","journal-title":"IEEE Trans Neural Netw Learn Syst"},{"issue":"1","key":"6407_CR18","doi-asserted-by":"publisher","first-page":"785","DOI":"10.1093\/bioinformatics\/btad785","volume":"40","author":"X Wang","year":"2024","unstructured":"Wang X, Wu T, Jiang Y, Chen T, Pan D, Jin Z, et al. RPEMHC: improved prediction of MHC-peptide binding affinity by a deep learning approach based on residue-residue pair encoding. Bioinformatics. 2024;40(1):785.","journal-title":"Bioinformatics"},{"issue":"3","key":"6407_CR19","doi-asserted-by":"publisher","first-page":"133","DOI":"10.1093\/bib\/bbae133","volume":"25","author":"L Zhang","year":"2024","unstructured":"Zhang L, Song W, Zhu T, Liu Y, Chen W, Cao Y. ConvNeXt-MHC: improving MHC-peptide affinity prediction by structure-derived degenerate coding and the ConvNeXt model. Brief Bioinform. 2024;25(3):133.","journal-title":"Brief Bioinform"},{"issue":"1","key":"6407_CR20","doi-asserted-by":"publisher","first-page":"3174","DOI":"10.1038\/s41467-021-23225-6","volume":"12","author":"AC McShan","year":"2021","unstructured":"McShan AC, Devlin CA, Morozov GI, Overall SA, Moschidi D, Akella N, et al. TAPBPR promotes antigen loading on MHC-I molecules using a peptide trap. Nat Commun. 2021;12(1):3174.","journal-title":"Nat Commun"},{"issue":"1","key":"6407_CR21","doi-asserted-by":"publisher","first-page":"301","DOI":"10.1146\/annurev-genom-091212-153455","volume":"14","author":"J Trowsdale","year":"2013","unstructured":"Trowsdale J, Knight JC. Major histocompatibility complex genomics and human disease. Annu Rev Genomics Hum Genet. 2013;14(1):301\u201323.","journal-title":"Annu Rev Genomics Hum Genet"},{"issue":"9","key":"6407_CR22","doi-asserted-by":"publisher","first-page":"810","DOI":"10.2174\/0115748936283134240109054157","volume":"19","author":"NN Soylu","year":"2024","unstructured":"Soylu NN, Sefer E. DeepPTM: protein post-translational modification prediction from protein sequences by combining deep protein language model with vision transformers. Curr Bioinform. 2024;19(9):810\u201324.","journal-title":"Curr Bioinform"},{"issue":"3","key":"6407_CR23","doi-asserted-by":"publisher","first-page":"2177","DOI":"10.1109\/TCBB.2023.3237769","volume":"20","author":"NN Soylu","year":"2023","unstructured":"Soylu NN, Sefer E. Bert2ome: Prediction of 2$$^{\\prime }$$-o-methylation modifications from RNA sequence by transformer architecture based on bert. IEEE\/ACM Trans Comput Biol Bioinf. 2023;20(3):2177\u201389.","journal-title":"IEEE\/ACM Trans Comput Biol Bioinf"},{"issue":"6637","key":"6407_CR24","doi-asserted-by":"publisher","first-page":"1123","DOI":"10.1126\/science.ade2574","volume":"379","author":"Z Lin","year":"2023","unstructured":"Lin Z, Akin H, Rao R, Hie B, Zhu Z, Lu W, et al. Evolutionary-scale prediction of atomic-level protein structure with a language model. Science. 2023;379(6637):1123\u201330.","journal-title":"Science"},{"key":"6407_CR25","unstructured":"Hu EJ, Shen Y, Wallis P, Allen-Zhu Z, Li Y, Wang S, Wang L, Chen W. Lora: low-rank adaptation of large language models. 2021. arXiv:2106.09685 [cs.CL]."},{"issue":"5","key":"6407_CR26","doi-asserted-by":"publisher","first-page":"766","DOI":"10.1038\/s41592-024-02240-7","volume":"21","author":"B McMaster","year":"2024","unstructured":"McMaster B, Thorpe C, Ogg G, Deane CM, Koohy H. Can Alphafold\u2019s breakthrough in protein structure help decode the fundamental principles of adaptive cellular immunity? Nat Methods. 2024;21(5):766\u201376.","journal-title":"Nat Methods"},{"key":"6407_CR27","doi-asserted-by":"publisher","first-page":"1855","DOI":"10.1038\/s41592-024-02377-5","volume":"21","author":"J Schwab","year":"2024","unstructured":"Schwab J, Kimanius D, Burt A, Dendooven T, Scheres SHW. DynaMight: estimating molecular motions with improved reconstruction from cryo-EM images. Nat Methods. 2024;21:1855\u201362.","journal-title":"Nat Methods"},{"issue":"3","key":"6407_CR28","doi-asserted-by":"publisher","first-page":"173","DOI":"10.1016\/S1359-0278(97)00024-2","volume":"2","author":"I Bahar","year":"1997","unstructured":"Bahar I, Atilgan AR, Erman B. Direct evaluation of thermal fluctuations in proteins using a single-parameter harmonic potential. Fold Des. 1997;2(3):173\u201381.","journal-title":"Fold Des"},{"issue":"48","key":"6407_CR29","first-page":"1471","volume":"25","author":"A Dehghan","year":"2024","unstructured":"Dehghan A, Abbasi K, Razzaghi P, Banadkuki H, Gharaghani S. CCL-DTI: contributing the contrastive loss in drug-target interaction prediction. BMC Bioinform. 2024;25(48):1471\u20132105.","journal-title":"BMC Bioinform"},{"issue":"4","key":"6407_CR30","doi-asserted-by":"publisher","first-page":"511","DOI":"10.1093\/bioinformatics\/btv639","volume":"32","author":"M Andreatta","year":"2016","unstructured":"Andreatta M, Nielsen M. Gapped sequence alignment using artificial neural networks: application to the MHC class I system. Bioinformatics. 2016;32(4):511\u20137.","journal-title":"Bioinformatics"},{"issue":"4","key":"6407_CR31","doi-asserted-by":"publisher","first-page":"418","DOI":"10.1016\/j.cels.2020.09.001","volume":"11","author":"TJ O\u2019Donnell","year":"2020","unstructured":"O\u2019Donnell TJ, Rubinsteyn A, Laserson U. MHCflurry 2.0: improved pan-allele prediction of MHC class I-presented peptides by incorporating antigen processing. Cell Syst. 2020;11(4):418\u20139.","journal-title":"Cell Syst"},{"issue":"13","key":"6407_CR32","doi-asserted-by":"publisher","first-page":"2174","DOI":"10.1093\/bioinformatics\/btv123","volume":"31","author":"T Trolle","year":"2015","unstructured":"Trolle T, Metushi IG, Greenbaum JA, Kim Y, Sidney J, Lund O, et al. Automated benchmarking of peptide-MHC class I binding predictions. Bioinformatics. 2015;31(13):2174\u201381.","journal-title":"Bioinformatics"},{"issue":"3","key":"6407_CR33","doi-asserted-by":"publisher","first-page":"177","DOI":"10.1007\/s00251-011-0579-8","volume":"64","author":"E Karosiene","year":"2012","unstructured":"Karosiene E, Lundegaard C, Lund O, Nielsen M. NetMHCcons: a consensus method for the major histocompatibility complex class I predictions. Immunogenetics. 2012;64(3):177\u201386.","journal-title":"Immunogenetics"},{"issue":"10","key":"6407_CR34","doi-asserted-by":"publisher","first-page":"1293","DOI":"10.1093\/bioinformatics\/btp137","volume":"25","author":"H Zhang","year":"2009","unstructured":"Zhang H, Lund O, Nielsen M. The pickpocket method for predicting binding specificities for receptors based on receptor pocket similarities: application to MHC-peptide binding. Bioinformatics. 2009;25(10):1293\u20139.","journal-title":"Bioinformatics"},{"issue":"14","key":"6407_CR35","doi-asserted-by":"publisher","first-page":"1765","DOI":"10.1093\/bioinformatics\/btg247","volume":"19","author":"B Peters","year":"2003","unstructured":"Peters B, Tong W, Sidney J, Sette A, Weng Z. Examining the independent binding assumption for binding of peptide epitopes to MHC-I molecules. Bioinformatics. 2003;19(14):1765\u201372.","journal-title":"Bioinformatics"},{"issue":"1","key":"6407_CR36","doi-asserted-by":"publisher","first-page":"394","DOI":"10.1186\/1471-2105-10-394","volume":"10","author":"Y Kim","year":"2009","unstructured":"Kim Y, Sidney J, Pinilla C, Sette A, Peters B. Derivation of an amino acid similarity matrix for peptide: MHC binding and its application as a Bayesian prior. BMC Bioinform. 2009;10(1):394.","journal-title":"BMC Bioinform"}],"container-title":["BMC Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/article\/10.1186\/s12859-026-06407-1","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12859-026-06407-1.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12859-026-06407-1.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,7]],"date-time":"2026-04-07T12:27:18Z","timestamp":1775564838000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1186\/s12859-026-06407-1"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,28]]},"references-count":36,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["6407"],"URL":"https:\/\/doi.org\/10.1186\/s12859-026-06407-1","relation":{},"ISSN":["1471-2105"],"issn-type":[{"value":"1471-2105","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,28]]},"assertion":[{"value":"10 November 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"13 February 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"28 February 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"Not applicable. The datasets in our research are from IEDB datasets, including the training MHC alleles datasets downloaded from\n                      \n                      , the weekly datasets downloaded from\n                      \n                      . All externally submitted data are from NIH epitope contracts whose projects undergo ethical screening prior to data collection, especially when human or live specimens are in question. Therefore, there is no further assessment to be done by the IEDB.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Consent to participate"}},{"value":"Not applicable.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Consent for publication"}},{"value":"The authors declare that they have no competing interest.","order":4,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interest"}}],"article-number":"82"}}