Pan-Peptide Meta Learning for T-cell receptor–antigen binding recognition

The identification of the mechanisms by which T-cell receptors (TCRs) interact with human antigens provides a crucial opportunity to develop new vaccines, diagnostics and immunotherapies. However, accurate prediction recognition TCR–antigen pairing represents substantial computational challenge in immunology. Existing tools only learn binding patterns from many known TCR repertoires fail recognize that have never been presented immune system or for few are known. specificity neoantigens exogenous peptides is studies immunotherapy. Therefore, we developed Pan-Peptide Meta Learning (PanPep), general robust framework binding, combining concepts meta-learning neural Turing machine. machine adds external memory avoid forgetting previously learned tasks, used here accurately predict any peptide, particularly unseen ones. We applied PanPep various challenging clinical including (1) qualitatively measuring clonal expansion T cells; (2) efficiently sorting responsive cells tumour neoantigen therapy; (3) identifying immune-responsive TCRs large cohort COVID-19 study. Our comprehensive tests show outperforms existing tools. also offers interpretability, revealing nature peptide interactions 3D crystal structures. believe can be useful tool decipher it has broad applications. Machine learning methods between antigens, but they struggle no little data exist regarding system. A method called based on quickly tasks predicts antigens.