Structure-Regularized Interpretable TCR-Epitope Prediction

2026-06-29

AI Summary

Current TCR–epitope binding prediction models suffer from limited generalizability to unseen epitopes and insufficient interpretability, while the impact of generated structural data on model learning remains unclear. This work proposes TCR-SRIM, which for the first time incorporates structural information as a regularizer within an interpretable framework, integrating protein language model embeddings with residue-level contact prototypes to explicitly model TCR–epitope interactions. Experiments on the TCR-XAI benchmark demonstrate that the method achieves state-of-the-art predictive performance while delivering higher-quality explanations. Furthermore, the study reveals that although current generated structures can aid prediction, they tend to distort interaction patterns and reduce the diversity of predicted binding sites.

Abstract

T cell receptor (TCR)-epitope binding prediction is essential for understanding adaptive immunity and developing immunotherapies. Existing sequence- and structure-based models often generalize poorly to unseen epitopes and provide limited interpretability. Furthermore, the impact of generated structures on model learning remains unclear. We present TCR-SRIM, a structure-regularized interpretable-by-design model that combines protein language model embeddings with interpretable contact prototypes to capture residue-level TCR-epitope interactions. TCR-SRIM achieves state-of-the-art predictive performance and improved interpretation quality on the TCR-XAI benchmark. Using its inherent interpretability, we further evaluate the effect of generated structures on model learning. While structures predicted by AlphaFold3, TCRModel2, and tFold-TCR yield competitive performance, they lead to less accurate interaction patterns and reduced binding-site diversity than experimentally-resolved structures. Our results highlight limitations of current structure prediction models for TCR-epitope learning and demonstrate the value of interpretable-by-design models for studying generated biological structures.

Problem

• TCR-epitope binding
• generalization
• interpretability
• structure prediction
• adaptive immunity

Innovation

• structure-regularized
• interpretable-by-design
• TCR-epitope interaction
• contact prototypes
• protein language model