Contrastive Geometric Learning Unlocks Unified Structure- and Ligand-Based Drug Design

2026-01-14 · Source: stat.ML updates on arXiv.org · Field: Science & Research — Mathematics & Computational Sciences, Life Sciences & Biology · Depth: Expert, quick

Summary

A new model, Contrastive Geometric Learning for Unified Computational Drug Design (ConGLUDe), unifies traditional structure-based and ligand-based computational drug design, which previously relied on disjoint data and modeling. Introduced by Lisa Schneckenreiter and colleagues, ConGLUDe employs a geometric protein encoder to generate whole-protein representations and implicit binding site embeddings, coupled with a fast ligand encoder. This approach eliminates the need for predefined binding pockets. Through contrastive learning, ConGLUDe aligns ligands with both global protein representations and multiple candidate binding sites, enabling ligand-conditioned pocket prediction, virtual screening, and target fishing. The model is trained jointly on protein-ligand complexes and extensive bioactivity data. Benchmarking demonstrates competitive zero-shot virtual screening performance, significant improvement over existing methods in target fishing, and leading performance in ligand-conditioned pocket selection. This work, submitted on 14 Jan 2026 and last revised 11 Jun 2026, advances towards general-purpose foundation models for drug discovery.

Key takeaway

For AI Scientists and Research Scientists developing computational drug design pipelines, ConGLUDe offers a unified approach to overcome traditional data and modeling silos. You should consider integrating contrastive geometric learning to simultaneously improve virtual screening, target fishing, and ligand-conditioned pocket prediction. This method removes the need for predefined binding pockets, streamlining your drug discovery workflows and potentially accelerating lead optimization.

Key insights

ConGLUDe unifies structure- and ligand-based drug design via contrastive geometric learning, enabling diverse drug discovery tasks.

Principles

• Unify disjoint data sources.
• Implicitly embed binding sites.
• Align global and local representations.

Method

ConGLUDe couples a geometric protein encoder with a fast ligand encoder, aligning them via contrastive learning on protein-ligand complexes and bioactivity data to predict pockets, screen, and fish targets.

In practice

• Perform zero-shot virtual screening.
• Improve target fishing accuracy.
• Select ligand-conditioned pockets.

Topics

• Contrastive Geometric Learning
• Drug Design
• Virtual Screening
• Target Fishing
• Protein-Ligand Interaction
• Foundation Models

Best for: AI Scientist, Research Scientist

Related on AIssential

• A Triple-Modal Contrastive Learning Framework with Sequence, Graph, and 3D Features for Drug-Target Interaction Prediction — TriMod-DTI uses triple-modal contrastive learning with 1D, 2D, and 3D features for superior drug-target interaction prediction.
• Probe Before You Edit: Probing-Guided Molecular Optimization for LLM Agents in Structure-Based Drug Design — Probing molecular edit responses guides LLM agents to jointly optimize binding affinity and druggability in drug design.
• How a Google DeepMind Spinoff Hunts Hidden Drug Targets — IsoDDE enables discovery of novel, cryptic protein binding pockets, expanding druggable targets beyond AlphaFold's capabilities.
• Bridging the phenotype-target gap for molecular generation via multi-objective reinforcement learning — ExMolRL combines phenotype-guided generation with target-aware reinforcement learning for de novo drug discovery.
• A physics-informed graph neural network to approximate docking-based binding affinity for DYRK2 in Alzheimer’s drug repurposing — Integrating physical interaction terms into GNNs can create efficient, interpretable surrogates for computationally intensive molecular docking.
• Co-Diffusion: An Affinity-Aware Two-Stage Latent Diffusion Framework for Generalizable Drug-Target Affinity Prediction — Co-Diffusion enhances drug-target affinity prediction in cold-start scenarios using a two-stage, affinity-aware latent diffusion framework.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by stat.ML updates on arXiv.org.