Protein Thoughts: Interpretable Reasoning with Tree of Thoughts and Embedding-Space Flow Matching for Protein-Protein Interaction Discovery

2026-05-19

AI Summary

This study addresses the limited interpretability of existing protein–protein interaction (PPI) prediction methods, which often conflate genuine biochemical mechanisms with spurious correlations. To bridge this gap, the work introduces— for the first time—an interpretable reasoning framework for PPI prediction, grounded in hypothesis-guided tree search. The approach decomposes biological signals into four modalities: sequence, structure, interface, and chemical properties, and integrates entropy-regularized Tree-of-Thoughts search with hypothesis-conditioned flow matching in embedding space to enable efficient exploration and validation. Evaluated on the SHS148k benchmark, the method substantially outperforms current models, improving the mean rank of the top true binding partner from 76 to 11.2 and achieving a Micro-F1 score of 91.08 ± 0.19, thereby offering both high predictive performance and strong interpretability.

Abstract

Protein-protein interactions (PPIs) govern nearly all cellular processes, yet computational methods for identifying binding partners typically produce ranked predictions without mechanistic justification. This creates a fundamental barrier to adoption because biologists cannot assess whether predictions reflect genuine biochemical insight or spurious correlations. We present \textbf{Protein Thoughts}, a framework that reformulates PPI discovery as an interpretable search problem with explicit reasoning. The system decomposes binding evidence into four biologically meaningful signals: sequence similarity reflecting evolutionary relationships, structural complementarity capturing geometric fit, interface balance, and chemical compatibility encoding residue-level interactions. Rather than collapsing these signals into an opaque score, we preserve their individual contributions through a transparent value function that enables both ranking and auditing. To navigate large candidate spaces efficiently, we introduce hypothesis-guided entropy-regularized Tree-of-Thoughts search. A fine-tuned language model generates search directives from embedding-derived features, classifying candidates as high-priority, exploratory, or skippable. These directives condition a Boltzmann policy that balances exploitation with entropy-driven exploration, while hypothesis-aware pruning prevents premature abandonment of promising candidates. For candidates exhibiting score disagreement, hypothesis-conditioned embedding-space flow matching transports protein embeddings toward the binder manifold. On the SHS148k benchmark, Protein Thoughts achieves mean best-binder rank of 11.2 versus 47.7 for an entropic tree search baseline, a 76% improvement, and for binding prediction the trained value function achieves $91.08 \pm 0.19$ Micro-F1, outperforming existing PPI methods on the same dataset.

Problem

• Protein-Protein Interaction
• Interpretability
• Mechanistic Justification
• Binding Prediction
• Computational Biology

Innovation

• interpretable reasoning
• Tree of Thoughts
• embedding-space flow matching
• protein-protein interaction
• hypothesis-guided search