Rational Multi-Modal Transformers for TCR-pMHC Prediction

2025-09-21

AI Summary

Current TCR–pMHC binding prediction models suffer from insufficient architectural systematicity and limited interpretability. Method: We propose an interpretability-driven multimodal Transformer design paradigm. Our approach employs an encoder–decoder architecture with cross-modal attention and auxiliary training objectives. We introduce, for the first time, an explanation-aware early-stopping criterion based on post-hoc explanation quality—e.g., feature importance consistency—to guide neural architecture search and optimal input modality selection. Additionally, we perform sequence-level interpretability analysis to uncover biologically meaningful binding patterns. Results: The proposed model achieves state-of-the-art performance across multiple benchmarks while yielding mechanistic biological insights. It significantly outperforms existing methods in interpretability (via faithful attribution), robustness to input perturbations, and cross-dataset generalization—demonstrating that interpretability can be effectively integrated as a core inductive bias rather than a post-hoc add-on.

Technology Category

• Machine Learning: Transparent, Interpretable, Explainable ML
• Computer Vision: Interpretability, Explainability, and Transparency
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

T cell receptor (TCR) recognition of peptide-MHC (pMHC) complexes is fundamental to adaptive immunity and central to the development of T cell-based immunotherapies. While transformer-based models have shown promise in predicting TCR-pMHC interactions, most lack a systematic and explainable approach to architecture design. We present an approach that uses a new post-hoc explainability method to inform the construction of a novel encoder-decoder transformer model. By identifying the most informative combinations of TCR and epitope sequence inputs, we optimize cross-attention strategies, incorporate auxiliary training objectives, and introduce a novel early-stopping criterion based on explanation quality. Our framework achieves state-of-the-art predictive performance while simultaneously improving explainability, robustness, and generalization. This work establishes a principled, explanation-driven strategy for modeling TCR-pMHC binding and offers mechanistic insights into sequence-level binding behavior through the lens of deep learning.

Problem

• Lack systematic explainable architecture for TCR-pMHC prediction
• Need to optimize cross-attention strategies for sequence inputs
• Improve robustness generalization in transformer binding models

Innovation

• Post-hoc explainability method guides transformer design
• Optimized cross-attention with auxiliary training objectives
• Novel early-stopping criterion based on explanation quality