Employing Protein Language Models to Predict TCR-Peptide-MHC Binding: A Comprehensive Methods Guide

Predicting the binding specificity between T-cell receptors (TCRs) and peptide-major histocompatibility complexes (pMHCs) represents one of the most challenging problems in computational immunology. Accurate prediction of these interactions is essential for advancing immunotherapy development, neoantigen vaccine design, and understanding adaptive immune responses. This chapter provides a comprehensive methodological guide to computational approaches for TCR-pMHC binding prediction, with particular emphasis on protein language model-based methods that have transformed the field since 2022. We describe the theoretical foundations, input requirements, output specifications, and practical implementation details for major prediction tools, including TCR-ESM, NetTCR-2.2, ERGO-II, pMTnet, ATM-TCR, and emerging approaches utilizing ESM-2, tcrLM, and graph neural networks. Additionally, we provide detailed protocols for data preparation, model execution, and result interpretation, along with essential notes addressing common challenges and best practices for method selection based on specific research objectives.