{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,18]],"date-time":"2026-08-18T02:56:03Z","timestamp":1787021763455,"version":"build-2736575974"},"update-to":[{"DOI":"10.1371\/journal.pcbi.1012511","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2024,11,21]],"date-time":"2024-11-21T00:00:00Z","timestamp":1732147200000}}],"reference-count":45,"publisher":"Public Library of Science (PLoS)","issue":"11","license":[{"start":{"date-parts":[[2024,11,11]],"date-time":"2024-11-11T00:00:00Z","timestamp":1731283200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100010661","name":"Horizon 2020 Framework Programme","doi-asserted-by":"publisher","award":["964998"],"award-info":[{"award-number":["964998"]}],"id":[{"id":"10.13039\/100010661","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>Neoantigen immunogenicity prediction is a highly challenging problem in the development of personalised medicines. Low reactivity rates in called neoantigens result in a difficult prediction scenario with limited training datasets. Here we describe ImmugenX, a modular protein language modelling approach to immunogenicity prediction for CD8+ reactive epitopes. ImmugenX comprises of a pMHC encoding module trained on three pMHC prediction tasks, an optional TCR encoding module and a set of context specific immunogenicity prediction head modules. Compared with state-of-the-art models for each task, ImmugenX\u2019s encoding module performs comparably or better on pMHC binding affinity, eluted ligand prediction and stability tasks. ImmugenX outperforms all compared models on pMHC immunogenicity prediction (Area under the receiver operating characteristic curve = 0.619, average precision: 0.514), with a 7% increase in average precision compared to the next best model. ImmugenX shows further improved performance on immunogenicity prediction with the integration of TCR context information. ImmugenX performance is further analysed for interpretability, which locates areas of weakness found across existing immunogenicity models and highlight possible biases in public datasets.<\/jats:p>","DOI":"10.1371\/journal.pcbi.1012511","type":"journal-article","created":{"date-parts":[[2024,11,11]],"date-time":"2024-11-11T13:42:12Z","timestamp":1731332532000},"page":"e1012511","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":13,"title":["A modular protein language modelling approach to immunogenicity prediction"],"prefix":"10.1371","volume":"20","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-6829-5306","authenticated-orcid":true,"given":"Hugh","family":"O\u2019Brien","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Max","family":"Salm","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Laura T.","family":"Morton","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Maciej","family":"Szukszto","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Felix","family":"O\u2019Farrell","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Charlotte","family":"Boulton","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Laurence","family":"King","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8558-8197","authenticated-orcid":true,"given":"Supreet Kaur","family":"Bola","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Pablo D.","family":"Becker","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Andrew","family":"Craig","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Morten","family":"Nielsen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yardena","family":"Samuels","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Charles","family":"Swanton","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Marc R.","family":"Mansour","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sine Reker","family":"Hadrup","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9763-1700","authenticated-orcid":true,"given":"Sergio A.","family":"Quezada","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2024,11,11]]},"reference":[{"issue":"4","key":"pcbi.1012511.ref001","doi-asserted-by":"crossref","first-page":"646","DOI":"10.1016\/j.ccell.2023.03.003","article-title":"Adoptive cell transfer immunotherapy for patients with solid epithelial cancers","volume":"41","author":"SA Rosenberg","year":"2023","journal-title":"Cancer Cell"},{"issue":"3","key":"pcbi.1012511.ref002","doi-asserted-by":"crossref","first-page":"818","DOI":"10.1016\/j.cell.2020.09.015","article-title":"Key Parameters of Tumor Epitope Immunogenicity Revealed Through a Consortium Approach Improve Neoantigen Prediction","volume":"183","author":"DK Wells","year":"2020","journal-title":"Cell"},{"issue":"1","key":"pcbi.1012511.ref003","doi-asserted-by":"crossref","first-page":"33","DOI":"10.1186\/s13073-016-0288-x","article-title":"NetMHCpan-3.0; improved prediction of binding to MHC class I molecules integrating information from multiple receptor and peptide length datasets","volume":"8","author":"M Nielsen","year":"2016","journal-title":"Genome Medicine"},{"issue":"1","key":"pcbi.1012511.ref004","doi-asserted-by":"crossref","first-page":"129","DOI":"10.1016\/j.cels.2018.05.014","article-title":"MHCflurry: Open-Source Class I MHC Binding Affinity Prediction.","volume":"7","author":"TJ O\u2019Donnell","year":"2018","journal-title":"Cell Systems"},{"issue":"W1","key":"pcbi.1012511.ref005","doi-asserted-by":"crossref","first-page":"W449","DOI":"10.1093\/nar\/gkaa379","article-title":"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","volume":"48","author":"B Reynisson","year":"2020","journal-title":"Nucleic Acids Research"},{"issue":"1","key":"pcbi.1012511.ref006","doi-asserted-by":"crossref","first-page":"42","DOI":"10.1016\/j.cels.2020.06.010","article-title":"MHCflurry 2.0: Improved PanAllele Prediction of MHC Class I-Presented Peptides by Incorporating Antigen Processing.","volume":"11","author":"TJ O\u2019Donnell","year":"2020","journal-title":"Cell Systems"},{"issue":"1","key":"pcbi.1012511.ref007","doi-asserted-by":"crossref","first-page":"72","DOI":"10.1016\/j.cels.2022.12.002","article-title":"Improved predictions of antigen presentation and TCR recognition with MixMHCpred2.2 and PRIME2.0 reveal potent SARS-CoV-2 CD8+ T-cell epitopes","volume":"14","author":"D Gfeller","year":"2023","journal-title":"Cell Systems"},{"key":"pcbi.1012511.ref008","first-page":"1","article-title":"Deep neural networks predict class I major histocompatibility complex epitope presentation and transfer learn neoepitope immunogenicity","author":"BA Albert","year":"2023","journal-title":"Nature Machine Intelligence"},{"issue":"3","key":"pcbi.1012511.ref009","doi-asserted-by":"crossref","first-page":"300","DOI":"10.1038\/s42256-022-00459-7","article-title":"A transformerbased model to predict peptide\u2013HLA class I binding and optimize mutated peptides for vaccine design","volume":"4","author":"Y Chu","year":"2022","journal-title":"Nature Machine Intelligence"},{"issue":"4","key":"pcbi.1012511.ref010","doi-asserted-by":"crossref","first-page":"bbac259","DOI":"10.1093\/bib\/bbac259","article-title":"A comprehensive analysis of the IEDB MHC class-I automated benchmark","volume":"23","author":"R Trevizani","year":"2022","journal-title":"Briefings in Bioinformatics"},{"issue":"4","key":"pcbi.1012511.ref011","doi-asserted-by":"crossref","first-page":"1517","DOI":"10.4049\/jimmunol.1600582","article-title":"Pan-Specific Prediction of Peptide-MHC Class I Complex Stability, a Correlate of T Cell Immunogenicity","volume":"197","author":"M Rasmussen","year":"2016","journal-title":"Journal of Immunology (Baltimore, Md: 1950)."},{"issue":"6","key":"pcbi.1012511.ref012","doi-asserted-by":"crossref","first-page":"357","DOI":"10.1007\/s00251-010-0441-4","article-title":"NetCTLpan: pan-specific MHC class I pathway epitope predictions","volume":"62","author":"T Stranzl","year":"2010","journal-title":"Immunogenetics"},{"issue":"5","key":"pcbi.1012511.ref013","doi-asserted-by":"crossref","first-page":"563","DOI":"10.1038\/s43018-021-00197-6","article-title":"A machine learning model for ranking candidate HLA class I neoantigens based on known neoepitopes from multiple human tumor types","volume":"2","author":"JJ Gartner","year":"2021","journal-title":"Nature Cancer"},{"key":"pcbi.1012511.ref014","article-title":"Machine learning methods and harmonized datasets improve immunogenic neoantigen prediction","author":"M Mu\u00a8ller","year":"2023","journal-title":"Immunity"},{"key":"pcbi.1012511.ref015","author":"C Dens","year":"2023","journal-title":"The pitfalls of negative data bias for the T-cell epitope specificity challenge"},{"key":"pcbi.1012511.ref016","author":"BPY Kwee","year":"2023","journal-title":"STAPLER: Efficient learning of TCR-peptide specificity prediction from fulllength TCR-peptide data"},{"key":"pcbi.1012511.ref017","doi-asserted-by":"crossref","first-page":"1055151","DOI":"10.3389\/fimmu.2022.1055151","article-title":"NetTCR-2.1: Lessons and guidance on how to develop models for TCR specificity predictions","volume":"13","author":"A Montemurro","year":"2022","journal-title":"Frontiers in Immunology"},{"issue":"12","key":"pcbi.1012511.ref018","doi-asserted-by":"crossref","first-page":"1618","DOI":"10.1038\/s43018-023-00675-z","article-title":"Breaking the performance ceiling for neoantigen immunogenicity prediction","volume":"4","author":"H O\u2019Brien","year":"2023","journal-title":"Nature Cancer"},{"key":"pcbi.1012511.ref019","first-page":"1","article-title":"Can we predict T cell specificity with digital biology and machine learning?","author":"D Hudson","year":"2023","journal-title":"Nature Reviews Immunology"},{"issue":"5","key":"pcbi.1012511.ref020","doi-asserted-by":"crossref","first-page":"656","DOI":"10.1038\/s41587-021-01072-6","article-title":"Sensitive identification of neoantigens and cognate TCRs in human solid tumors","volume":"40","author":"M Arnaud","year":"2022","journal-title":"Nature Biotechnology"},{"issue":"D1","key":"pcbi.1012511.ref021","doi-asserted-by":"crossref","first-page":"D339","DOI":"10.1093\/nar\/gky1006","article-title":"The Immune Epitope Database (IEDB): 2018 update.","volume":"47","author":"R Vita","year":"2019","journal-title":"Nucleic Acids Research"},{"issue":"1","key":"pcbi.1012511.ref022","doi-asserted-by":"crossref","first-page":"1935","DOI":"10.1038\/s41467-022-29342-0","article-title":"Neoantigen-specific CD8 T cell responses in the peripheral blood following PDL1 blockade might predict therapy outcome in metastatic urothelial carcinoma.","volume":"13","author":"JS Holm","year":"2022","journal-title":"Nature Communications."},{"key":"pcbi.1012511.ref023","doi-asserted-by":"crossref","first-page":"100024","DOI":"10.1016\/j.immuno.2023.100024","article-title":"Benchmarking solutions to the T-cell receptor epitope prediction problem: IMMREP22 workshop report.","volume":"9","author":"P Meysman","year":"2023","journal-title":"ImmunoInformatics."},{"key":"pcbi.1012511.ref024","author":"J Devlin","year":"2019","journal-title":"BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding"},{"issue":"2","key":"pcbi.1012511.ref025","doi-asserted-by":"crossref","first-page":"155","DOI":"10.1093\/protein\/4.2.155","article-title":"Correlation between stability of a protein and its dipeptide composition: a novel approach for predicting in vivo stability of a protein from its primary sequence","volume":"4","author":"K Guruprasad","year":"1990","journal-title":"Protein Engineering"},{"key":"pcbi.1012511.ref026","first-page":"4768","article-title":"A unified approach to interpreting model predictions.","author":"SM Lundberg","year":"2017","journal-title":"Proceedings of the 31st International Conference on Neural Information Processing Systems."},{"key":"pcbi.1012511.ref027","doi-asserted-by":"crossref","first-page":"100027","DOI":"10.1016\/j.immuno.2023.100027","article-title":"Interpretable deep learning to uncover the molecular binding patterns determining TCR\u2013epitope interaction predictions.","volume":"11","author":"C Dens","year":"2023","journal-title":"ImmunoInformatics."},{"issue":"4","key":"pcbi.1012511.ref028","doi-asserted-by":"crossref","first-page":"bbaa318","DOI":"10.1093\/bib\/bbaa318","article-title":"Current challenges for unseen-epitope TCR interaction prediction and a new perspective derived from image classification","volume":"22","author":"P Moris","year":"2021","journal-title":"Briefings in Bioinformatics"},{"issue":"1","key":"pcbi.1012511.ref029","doi-asserted-by":"crossref","first-page":"219","DOI":"10.1016\/j.cell.2019.08.032","article-title":"UVBInduced Tumor Heterogeneity Diminishes Immune Response in Melanoma","volume":"179","author":"Y Wolf","year":"2019","journal-title":"Cell"},{"key":"pcbi.1012511.ref030","author":"Z Lin","year":"2022","journal-title":"Language models of protein sequences at the scale of evolution enable accurate structure prediction"},{"key":"pcbi.1012511.ref031","author":"PH Bradley","year":"2022","journal-title":"Structure-based prediction of T cell receptor:peptide-MHC interactions"},{"issue":"2","key":"pcbi.1012511.ref032","doi-asserted-by":"crossref","first-page":"539","DOI":"10.4049\/jimmunol.1901173","article-title":"Impact of Cysteine Residues on MHC Binding Predictions and Recognition by Tumor-Reactive T Cells","volume":"205","author":"A Sachs","year":"2020","journal-title":"The Journal of Immunology"},{"key":"pcbi.1012511.ref033","first-page":"1","article-title":"High-throughput, targeted MHC class I immunopeptidomics using a functional genetics screening platform","author":"PM Bruno","year":"2023","journal-title":"Nature Biotechnology"},{"key":"pcbi.1012511.ref034","first-page":"15","article-title":"IMPROVE: a feature model to predict neoepitope immunogenicity through broad-scale validation of T-cell recognition.","author":"A Borch","year":"2024","journal-title":"Frontiers in Immunology"},{"key":"pcbi.1012511.ref035","author":"A Paszke","year":"2019","journal-title":"PyTorch: An Imperative Style, High-Performance Deep Learning Library;"},{"key":"pcbi.1012511.ref036","article-title":"Attention Is All You Need.","author":"A Vaswani","year":"2017","journal-title":"arXiv:170603762 [cs]."},{"issue":"8","key":"pcbi.1012511.ref037","doi-asserted-by":"crossref","first-page":"e796","DOI":"10.1371\/journal.pone.0000796","article-title":"NetMHCpan, a Method for Quantitative Predictions of Peptide Binding to Any HLA-A and -B Locus Protein of Known Sequence.","volume":"2","author":"M Nielsen","year":"2007","journal-title":"PLoS ONE"},{"key":"pcbi.1012511.ref038","first-page":"11","article-title":"A Systematic, Unbiased Mapping of CD8+ and CD4+ T Cell Epitopes in Yellow Fever Vaccinees.","author":"A Stryhn","year":"2020","journal-title":"Frontiers in Immunology"},{"issue":"9","key":"pcbi.1012511.ref039","doi-asserted-by":"crossref","first-page":"1017","DOI":"10.1038\/s41592-022-01578-0","article-title":"VDJdb in the pandemic era: a compendium of T cell receptors specific for SARSCoV-2.","volume":"19","author":"M Goncharov","year":"2022","journal-title":"Nature Methods."},{"issue":"18","key":"pcbi.1012511.ref040","doi-asserted-by":"crossref","first-page":"2924","DOI":"10.1093\/bioinformatics\/btx286","article-title":"McPAS-TCR: a manually curated catalogue of pathology-associated T cell receptor sequences","volume":"33","author":"N Tickotsky","year":"2017","journal-title":"Bioinformatics"},{"issue":"7","key":"pcbi.1012511.ref041","doi-asserted-by":"crossref","first-page":"1316","DOI":"10.1101\/gr.080531.108","article-title":"The consensus coding sequence (CCDS) project: Identifying a common protein-coding gene set for the human and mouse genomes.","volume":"19","author":"KD Pruitt","year":"2009","journal-title":"Genome Research"},{"key":"pcbi.1012511.ref042","article-title":"The Cancer Epitope Database and Analysis Resource (CEDAR).","volume":"gkac902","author":"salo\u02d8glu-Yal Ko\u00b8","year":"2022","journal-title":"Nucleic Acids Research"},{"issue":"5","key":"pcbi.1012511.ref043","doi-asserted-by":"crossref","first-page":"659","DOI":"10.1038\/ng.3822","article-title":"Immunosequencing identifies signatures of cytomegalovirus exposure history and HLA-mediated effects on the T cell repertoire","volume":"49","author":"RO Emerson","year":"2017","journal-title":"Nature Genetics"},{"issue":"Database issue","key":"pcbi.1012511.ref044","doi-asserted-by":"crossref","first-page":"D256","DOI":"10.1093\/nar\/gki010","article-title":"IMGT\/GENE-DB: a comprehensive database for human and mouse immunoglobulin and T cell receptor genes","volume":"33","author":"V Giudicelli","year":"2005","journal-title":"Nucleic Acids Research"},{"key":"pcbi.1012511.ref045","article-title":"Scikit-learn: Machine Learning in Python.","author":"F Pedregosa","year":"2012","journal-title":"arxiv"}],"updated-by":[{"DOI":"10.1371\/journal.pcbi.1012511","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2024,11,21]],"date-time":"2024-11-21T00:00:00Z","timestamp":1732147200000}}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012511","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,11,21]],"date-time":"2024-11-21T14:01:52Z","timestamp":1732197712000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012511"}},"subtitle":[],"editor":[{"given":"Yang","family":"Lu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2024,11,11]]},"references-count":45,"journal-issue":{"issue":"11","published-online":{"date-parts":[[2024,11,11]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1012511","relation":{},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,11,11]]}}}