Structure-Based Prediction and Design of Adaptive Immune Receptors Targeting Peptide–MHC Complexes

Abstract

Adaptive immune recognition depends on the specific interaction between peptides presented by major histocompatibility complexes (MHCs) and the receptors that survey them. Advances in protein structure prediction and design now allow us to computationally model these interactions with unprecedented fidelity and engineer them with remarkably higher success rates. In this dissertation, we develop two complementary structure-based deep learning approaches: one for predicting peptide–MHC specificity by fine-tuning structure prediction networks directly on binding data, and another for designing de novo T cell receptors and TCR-mimic antibodies that recognize peptide–MHC targets with high accuracy. Together, these methods illustrate how integrating structurally-informed deep learning protein design and structure prediction frameworks enable both robust generalization and precise molecular engineering. These tools lay the foundation for programmable, therapeutic recognition of diseased cells and expand our understanding of adaptive immune specificity.