S$^2$Drug: Bridging Protein Sequence and 3D Structure in Contrastive Representation Learning for Virtual Screening

2025-11-10

AI Summary

This work addresses two key challenges in virtual screening: the disconnection between protein sequence and 3D structural information, and the prevalence of redundancy and noise in large-scale protein–ligand datasets. To this end, we propose S²Drug, a two-stage contrastive learning framework. S²Drug is the first contrastive method to explicitly integrate protein sequences—encoded via ESM2 pretraining—with 3D structural representations. It introduces a residue-level gating module and a binding-site prediction auxiliary task to enhance functional site awareness, and employs a customized negative sampling strategy to mitigate data noise. Evaluated on multiple virtual screening benchmarks, S²Drug achieves significant improvements in retrieval accuracy. Moreover, it attains state-of-the-art performance on binding-site prediction, demonstrating the effectiveness and generalizability of sequence–structure co-modeling.

Technology Category

• Machine Learning: Structured Learning
• Computer Vision: 3D Computer Vision
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Virtual screening (VS) is an essential task in drug discovery, focusing on the identification of small-molecule ligands that bind to specific protein pockets. Existing deep learning methods, from early regression models to recent contrastive learning approaches, primarily rely on structural data while overlooking protein sequences, which are more accessible and can enhance generalizability. However, directly integrating protein sequences poses challenges due to the redundancy and noise in large-scale protein-ligand datasets. To address these limitations, we propose extbf{S$^2$Drug}, a two-stage framework that explicitly incorporates protein extbf{S}equence information and 3D extbf{S}tructure context in protein-ligand contrastive representation learning. In the first stage, we perform protein sequence pretraining on ChemBL using an ESM2-based backbone, combined with a tailored data sampling strategy to reduce redundancy and noise on both protein and ligand sides. In the second stage, we fine-tune on PDBBind by fusing sequence and structure information through a residue-level gating module, while introducing an auxiliary binding site prediction task. This auxiliary task guides the model to accurately localize binding residues within the protein sequence and capture their 3D spatial arrangement, thereby refining protein-ligand matching. Across multiple benchmarks, S$^2$Drug consistently improves virtual screening performance and achieves strong results on binding site prediction, demonstrating the value of bridging sequence and structure in contrastive learning.

Problem

• Integrating protein sequences with 3D structures for virtual screening
• Reducing redundancy and noise in protein-ligand datasets
• Improving binding site prediction through sequence-structure fusion

Innovation

• Integrates protein sequences and 3D structures contrastively
• Uses two-stage pretraining and fine-tuning with ESM2 backbone
• Employs binding site prediction to refine protein-ligand matching