{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,8,2]],"date-time":"2025-08-02T14:19:19Z","timestamp":1754144359498,"version":"3.41.2"},"reference-count":36,"publisher":"Oxford University Press (OUP)","issue":"Supplement_1","license":[{"start":{"date-parts":[[2025,7,15]],"date-time":"2025-07-15T00:00:00Z","timestamp":1752537600000},"content-version":"vor","delay-in-days":14,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000057","name":"National Institute of General Medical Sciences","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100000057","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["R35GM155417"],"award-info":[{"award-number":["R35GM155417"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,7,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:sec>\n                  <jats:title>Motivation<\/jats:title>\n                  <jats:p>Reliable TCR-epitope binding prediction models are essential for development of adoptive T cell therapy and vaccine design. These models often struggle with false positives, which can be attributed to the limited data coverage in existing negative sample datasets. Common strategies for generating negative samples, such as pairing with background T cell receptors (TCRs) or shuffling within pairs, fail to account for model-specific vulnerabilities or biologically implausible sequences.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>To address these challenges, we propose an iterative attack-and-defend framework that systematically identifies and mitigates weaknesses in TCR-epitope prediction models. During the attack phase, a reinforcement learning from AI feedback (RLAIF) framework is used to attack a prediction model by generating biologically implausible sequences that can easily deceive the model. During the defense phase, these identified false positives are incorporated into fine-tuning dataset, enhancing the model\u2019s ability to detect false positives. A comprehensive adversarial negative dataset can be obtained by iteratively attacking and defending the model. This dataset can be directly used to improve model robustness, eliminating the need for users to conduct their own attack-and-defend cycles. We apply our framework to five existing binding prediction models, spanning diverse architectures and embedding strategies to show its efficacy. Experimental results show that our approach significantly improves these models\u2019 ability to detect adversarial false positives. The combined dataset constructed from these experiments also provides a benchmarking tool to evaluate and refine prediction models. Our framework offers a new approach for improving model robustness in other biological tasks where negative sampling is inherently limited.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Availability and implementation<\/jats:title>\n                  <jats:p>The curated dataset and code are available at a public repository (https:\/\/github.com\/Lee-CBG\/BAP_Attack_n_Defend).<\/jats:p>\n               <\/jats:sec>","DOI":"10.1093\/bioinformatics\/btaf224","type":"journal-article","created":{"date-parts":[[2025,7,15]],"date-time":"2025-07-15T13:02:51Z","timestamp":1752584571000},"page":"i429-i438","source":"Crossref","is-referenced-by-count":0,"title":["Iterative attack-and-defend framework for improving TCR-epitope binding prediction models"],"prefix":"10.1093","volume":"41","author":[{"given":"Pengfei","family":"Zhang","sequence":"first","affiliation":[{"name":"School of Computing and Augmented Intelligence, Arizona State University , Tempe, AZ 85281,","place":["United States"]},{"name":"Biodesign Institute, Arizona State University , Tempe, AZ 85281,","place":["United States"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hao","family":"Mei","sequence":"additional","affiliation":[{"name":"School of Computing and Augmented Intelligence, Arizona State University , Tempe, AZ 85281,","place":["United States"]},{"name":"Biodesign Institute, Arizona State University , Tempe, AZ 85281,","place":["United States"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Seojin","family":"Bang","sequence":"additional","affiliation":[{"name":"Google DeepMind , Mountain View, CA 94043,","place":["United States"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Heewook","family":"Lee","sequence":"additional","affiliation":[{"name":"School of Computing and Augmented Intelligence, Arizona State University , Tempe, AZ 85281,","place":["United States"]},{"name":"Biodesign Institute, Arizona State University , Tempe, AZ 85281,","place":["United States"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"286","published-online":{"date-parts":[[2025,7,15]]},"reference":[{"key":"2025071509024915000_btaf224-B1","doi-asserted-by":"crossref","first-page":"893247","DOI":"10.3389\/fimmu.2022.893247","article-title":"ATM-TCR: TCR-epitope binding affinity prediction using a multi-head self-attention model","volume":"13","author":"Cai","year":"2022","journal-title":"Front Immunol"},{"key":"2025071509024915000_btaf224-B2","article-title":"Deep reinforcement learning from human preferences","volume":"30","author":"Christiano","year":"2017","journal-title":"Adv Neural Inf Process Syst"},{"key":"2025071509024915000_btaf224-B3","doi-asserted-by":"crossref","first-page":"640725","DOI":"10.3389\/fimmu.2021.640725","article-title":"TCRMatch: predicting T-cell receptor specificity based on sequence similarity to previously characterized receptors","volume":"12","author":"Chronister","year":"2021","journal-title":"Front Immunol"},{"key":"2025071509024915000_btaf224-B4","doi-asserted-by":"crossref","first-page":"89","DOI":"10.1038\/nature22383","article-title":"Quantifiable predictive features define epitope-specific T cell receptor repertoires","volume":"547","author":"Dash","year":"2017","journal-title":"Nature"},{"key":"2025071509024915000_btaf224-B5","doi-asserted-by":"crossref","first-page":"1060","DOI":"10.1038\/s42256-023-00727-0","article-title":"The pitfalls of negative data bias for the T-cell epitope specificity challenge","volume":"5","author":"Dens","year":"2023","journal-title":"Nat Mach Intell"},{"key":"2025071509024915000_btaf224-B6","doi-asserted-by":"crossref","first-page":"302","DOI":"10.1038\/embor.2013.34","article-title":"Biggest challenges in bioinformatics","volume":"14","author":"Fuller","year":"2013","journal-title":"EMBO Rep"},{"key":"2025071509024915000_btaf224-B7","doi-asserted-by":"crossref","first-page":"236","DOI":"10.1038\/s42256-023-00619-3","article-title":"Pan-peptide meta learning for T-cell receptor\u2013antigen binding recognition","volume":"5","author":"Gao","year":"2023","journal-title":"Nat Mach Intell"},{"key":"2025071509024915000_btaf224-B8","article-title":"Generative adversarial nets","volume":"27","author":"Goodfellow","year":"2014","journal-title":"Adv Neural Inf Process Syst"},{"key":"2025071509024915000_btaf224-B9","doi-asserted-by":"crossref","first-page":"10915","DOI":"10.1073\/pnas.89.22.10915","article-title":"Amino acid substitution matrices from protein blocks","volume":"89","author":"Henikoff","year":"1992","journal-title":"Proc Nat Acad Sci USA"},{"year":"2022","author":"Hesslow","key":"2025071509024915000_btaf224-B10"},{"key":"2025071509024915000_btaf224-B11","doi-asserted-by":"crossref","first-page":"511","DOI":"10.1038\/s41577-023-00835-3","article-title":"Can we predict T cell specificity with digital biology and machine learning?","volume":"23","author":"Hudson","year":"2023","journal-title":"Nat Rev Immunol"},{"key":"2025071509024915000_btaf224-B12","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3571730","article-title":"Survey of hallucination in natural language generation","volume":"55","author":"Ji","year":"2023","journal-title":"ACM Comput Surv"},{"key":"2025071509024915000_btaf224-B13","doi-asserted-by":"crossref","first-page":"bbad038","DOI":"10.1093\/bib\/bbad038","article-title":"Deep autoregressive generative models capture the intrinsics embedded in T-cell receptor repertoires","volume":"24","author":"Jiang","year":"2023","journal-title":"Brief Bioinform"},{"key":"2025071509024915000_btaf224-B14","doi-asserted-by":"crossref","first-page":"btac788","DOI":"10.1093\/bioinformatics\/btac788","article-title":"TCRconv: predicting recognition between T cell receptors and epitopes using contextualized motifs","volume":"39","author":"Jokinen","year":"2023","journal-title":"Bioinformatics"},{"key":"2025071509024915000_btaf224-B15","doi-asserted-by":"crossref","first-page":"467","DOI":"10.1038\/s41577-018-0007-5","article-title":"Understanding the drivers of MHC restriction of T cell receptors","volume":"18","author":"La Gruta","year":"2018","journal-title":"Nat Rev Immunol"},{"key":"2025071509024915000_btaf224-B16","doi-asserted-by":"crossref","first-page":"214","DOI":"10.1016\/j.jtbi.2015.10.016","article-title":"How many TCR clonotypes does a body maintain?","volume":"389","author":"Lythe","year":"2016","journal-title":"J Theor Biol"},{"key":"2025071509024915000_btaf224-B17","doi-asserted-by":"crossref","first-page":"1060","DOI":"10.1038\/s42003-021-02610-3","article-title":"NetTCR-2.0 enables accurate prediction of TCR-peptide binding by using paired TCR\u03b1 and \u03b2 sequence data","volume":"4","author":"Montemurro","year":"2021","journal-title":"Commun Biol"},{"key":"2025071509024915000_btaf224-B18","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":"Moris","year":"2021","journal-title":"Brief Bioinform"},{"first-page":"1488851","volume-title":"Front Immunol","year":"2025","key":"2025071509024915000_btaf224-B19"},{"key":"2025071509024915000_btaf224-B20","first-page":"27730","article-title":"Training language models to follow instructions with human feedback","volume":"35","author":"Ouyang","year":"2022","journal-title":"Adv Neural Inf Process Syst"},{"key":"2025071509024915000_btaf224-B21","doi-asserted-by":"crossref","first-page":"114704","DOI":"10.1016\/j.celrep.2024.114704","article-title":"The observed T cell receptor space database enables paired-chain repertoire mining, coherence analysis, and language modeling","volume":"43","author":"Raybould","year":"2024","journal-title":"Cell Rep"},{"key":"2025071509024915000_btaf224-B22","doi-asserted-by":"crossref","first-page":"53","DOI":"10.1016\/0377-0427(87)90125-7","article-title":"Silhouettes: a graphical aid to the interpretation and validation of cluster analysis","volume":"20","author":"Rousseeuw","year":"1987","journal-title":"J Comput Appl Math"},{"key":"2025071509024915000_btaf224-B23","doi-asserted-by":"crossref","first-page":"303","DOI":"10.2174\/1568005014605955","article-title":"Use of T cell epitopes for vaccine development","volume":"1","author":"Sbai","year":"2001","journal-title":"Curr Drug Targets Infect Disord"},{"key":"2025071509024915000_btaf224-B24","doi-asserted-by":"crossref","first-page":"173","DOI":"10.1146\/annurev-immunol-042617-053402","article-title":"Cancer neoantigens","volume":"37","author":"Schumacher","year":"2019","journal-title":"Annu Rev Immunol"},{"key":"2025071509024915000_btaf224-B25","doi-asserted-by":"crossref","first-page":"D419","DOI":"10.1093\/nar\/gkx760","article-title":"VDJdb: a curated database of T-cell receptor sequences with known antigen specificity","volume":"46","author":"Shugay","year":"2018","journal-title":"Nucleic Acids Res"},{"key":"2025071509024915000_btaf224-B26","doi-asserted-by":"crossref","first-page":"664514","DOI":"10.3389\/fimmu.2021.664514","article-title":"Contribution of T cell receptor alpha and beta CDR3, MHC typing, V and J genes to peptide binding prediction","volume":"12","author":"Springer","year":"2021","journal-title":"Front Immunol"},{"key":"2025071509024915000_btaf224-B27","first-page":"3008","article-title":"Learning to summarize with human feedback","volume":"33","author":"Stiennon","year":"2020","journal-title":"Adv Neural Inf Process Syst"},{"key":"2025071509024915000_btaf224-B28","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":"Tickotsky","year":"2017","journal-title":"Bioinformatics"},{"key":"2025071509024915000_btaf224-B29","doi-asserted-by":"crossref","first-page":"D339","DOI":"10.1093\/nar\/gky1006","article-title":"The immune epitope database (IEDB): 2018 update","volume":"47","author":"Vita","year":"2019","journal-title":"Nucleic Acids Res"},{"key":"2025071509024915000_btaf224-B30","doi-asserted-by":"crossref","first-page":"i237","DOI":"10.1093\/bioinformatics\/btab294","article-title":"TITAN: T-cell receptor specificity prediction with bimodal attention networks","volume":"37","author":"Weber","year":"2021","journal-title":"Bioinformatics"},{"author":"Wu","key":"2025071509024915000_btaf224-B31","first-page":"194"},{"first-page":"347","volume-title":"Pac Symp Biocomput","author":"Zhang","key":"2025071509024915000_btaf224-B32"},{"key":"2025071509024915000_btaf224-B33","doi-asserted-by":"publisher","DOI":"10.1101\/2023.04.12.536635","article-title":"Context-aware amino acid embedding advances analysis of TCR-epitope interactions","volume":"12","author":"Zhang","year":"2023","journal-title":"eLife"},{"first-page":"988","year":"2023","author":"Zhang","key":"2025071509024915000_btaf224-B34"},{"year":"2025","author":"Zhang","key":"2025071509024915000_btaf224-B35"},{"first-page":"167","year":"2018","author":"Zheng","key":"2025071509024915000_btaf224-B36"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/41\/Supplement_1\/i429\/63745339\/btaf224.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/41\/Supplement_1\/i429\/63745339\/btaf224.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,15]],"date-time":"2025-07-15T13:02:54Z","timestamp":1752584574000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/41\/Supplement_1\/i429\/8199358"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,7,1]]},"references-count":36,"journal-issue":{"issue":"Supplement_1","published-print":{"date-parts":[[2025,7,1]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btaf224","relation":{},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"type":"print","value":"1367-4803"},{"type":"electronic","value":"1367-4811"}],"subject":[],"published-other":{"date-parts":[[2025,7]]},"published":{"date-parts":[[2025,7,1]]}}}