An Interpretable Framework Applying Protein Words to Predict Protein-Small Molecule Complementary Pairing Rules

arXiv cs.LG / 4/21/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces the PWRules framework to make protein–small molecule binding predictions more interpretable by deriving fragment-level “pairing rules” from binding affinity data.
PWRules identifies privileged small-molecule fragments, maps them to protein “words” (semantic sequence units), and ranks predicted active compounds using the PWScore function.
On benchmark datasets, PWScore performs competitively with established physics-based (Glide) and deep learning (PSICHIC) approaches, and generalizes to protein targets beyond the training set, including SARS-CoV-2 main protease.
The method’s interpretability is supported by structural analysis showing that learned word–fragment rules concentrate near ligand-binding pockets even without explicit structural supervision.
When combined with physics-based or deep learning methods, integrating PWRules yields better enrichment performance, suggesting it contributes complementary interaction information rather than replacing existing models.

Abstract

Despite the high accuracy of 'black box' deep learning models, drug discovery still relies on protein-ligand interaction principles and heuristics. To improve interpretability of protein-small molecule binding predictions, we developed the PWRules framework, which applies binding affinity data to identify privileged small molecule fragments and subsequently defines complementary pairing rules between these fragments and protein words (semantic sequence units) through an interpretability module. The resulting word-fragment rules are then ranked by the PWScore function to prioritize active compounds. Evaluations on benchmark datasets show that PWScore achieves competitive performance comparable to the physics-based model (Glide) and the deep learning model (PSICHIC) and shows broad applicability for protein targets outside the training dataset, e.g., SARS-CoV-2 main protease. Notably, PWScore captures complementary interaction information, yielding superior enrichment performance when integrated with these established methods. Structural analysis of protein-ligand complexes indicates that learned word-fragment rules are significantly enriched near ligand-binding pockets, despite training without explicit structural guidance. By extracting and applying complementary pairing rules, PWRules provides an interpretable framework for drug discovery.