Journal of Pioneering Artificial Intelligence Research

Abstract

T cell receptor (TCR)–peptide-human leukocyte antigen (pHLA) interactions are central to adaptive immunity and represent a critical determinant of the efficacy and safety of cancer immunotherapies, including adoptive cell transfer, neoantigen vaccines, and engineered TCR-T cell therapies. Despite decades of structural and biochemical research, the accurate computational prediction of TCR–pHLA binding affinity remains one of the most challenging problems at the intersection of biophysics and immunology. This review surveys the landscape of free energy calculation methods including molecular mechanics Poisson-Boltzmann/Generalized Born Surface Area (MM-PBSA/GBSA), alchemical free energy perturbation (FEP), enhanced sampling molecular dynamics, and emerging machine-learning-augmented approaches as applied to the TCR–pHLA binding problem. We examine what structural and thermodynamic features make this protein–protein interface uniquely difficult to model, assess the accuracy and scalability of current methods against experimental benchmarks, and critically evaluate how TCR cross-reactivity compounds the prediction challenge. Finally, we discuss what advances in computational free energy prediction would unlock for the design of safer and more effective TCR-based therapies and highlight the most productive directions for future methodological development.