JanusDDG: A Thermodynamics-Compliant Model for Sequence-Based Protein Stability via Two-Fronts Multi-Head Attention

2025-04-04

AI Summary

This work addresses the problem of accurately predicting the change in protein stability (ΔΔG) induced by single- and multi-point mutations from sequence-only input, while strictly enforcing fundamental thermodynamic constraints—antisymmetry and transitivity. We propose the first thermodynamically constrained bidirectional cross-attention mechanism: leveraging protein language model (PLM) embeddings as input, the query and value vectors encode the wild-type–mutant representation difference, while the key vector alternates to jointly enable perturbation-awareness and contextual preservation. Additionally, we introduce a physics-informed regularization term that explicitly enforces thermodynamically consistent ΔΔG relationships. Our method achieves state-of-the-art performance across multiple benchmark datasets, matching or surpassing structure-based approaches in accuracy—despite using no 3D structural information. To our knowledge, this is the first method enabling high-fidelity, structure-free prediction of multi-mutation ΔΔG while satisfying rigorous thermodynamic principles.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Diffusion Models for Vision
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Understanding how residue variations affect protein stability is crucial for designing functional proteins and deciphering the molecular mechanisms underlying disease-related mutations. Recent advances in protein language models (PLMs) have revolutionized computational protein analysis, enabling, among other things, more accurate predictions of mutational effects. In this work, we introduce JanusDDG, a deep learning framework that leverages PLM-derived embeddings and a bidirectional cross-attention transformer architecture to predict $Delta Delta G$ of single and multiple-residue mutations while simultaneously being constrained to respect fundamental thermodynamic properties, such as antisymmetry and transitivity. Unlike conventional self-attention, JanusDDG computes queries (Q) and values (V) as the difference between wild-type and mutant embeddings, while keys (K) alternate between the two. This cross-interleaved attention mechanism enables the model to capture mutation-induced perturbations while preserving essential contextual information. Experimental results show that JanusDDG achieves state-of-the-art performance in predicting $Delta Delta G$ from sequence alone, matching or exceeding the accuracy of structure-based methods for both single and multiple mutations.

Problem

• Predicts protein stability changes from residue mutations
• Ensures thermodynamic consistency in stability predictions
• Uses transformer architecture for mutation effect analysis

Innovation

• Bidirectional cross-attention transformer architecture
• PLM-derived embeddings for stability prediction
• Cross-interleaved attention mechanism