The influence of ligands on AlphaFold3 prediction of cryptic pockets

2026-07-07 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

AlphaFold3 (AF3) demonstrates significant capability in predicting cryptic pockets, which are crucial druggable binding sites formed by conformational changes. Researchers found that AF3 can reproduce the necessary scale of conformational change for these sites. Critically, the presence of a cryptic-site ligand during prediction leads AF3 to predominantly generate conformations competent for ligand binding, whereas its absence results in conformations lacking the pocket. The choice of ligand also substantially influences predictions, with AF3 capable of correctly positioning ligands within models. While some memorization bias exists, its detrimental impact appears limited. The study emphasizes that generating ensembles of co-folded predictions is essential to account for ligand position variability and enhance the likelihood of obtaining accurate binding modes. These AF3-generated protein-ligand structural ensembles hold promise for cryptic-site drug discovery and identifying potential binding ligands.

Key takeaway

For research scientists focused on drug discovery targeting cryptic pockets, you should integrate AlphaFold3 predictions by supplying known or candidate ligands to guide conformational sampling. Generating structural ensembles of co-folded protein-ligand complexes is crucial to account for ligand position variability and increase the probability of identifying correct binding modes. This approach can significantly enhance the discovery of novel ligands and druggable sites.

Key insights

AlphaFold3 predicts cryptic pockets and ligand binding modes more effectively when a ligand is included in the prediction.

Principles

• Ligand presence guides cryptic pocket formation.
• Ligand choice impacts prediction outcomes.
• Ensembles improve binding mode accuracy.

Method

AlphaFold3 is used to generate protein-ligand structural ensembles, exploring conformational changes and ligand positioning for cryptic site formation.

In practice

• Supply cryptic-site ligands to AF3 for pocket prediction.
• Generate multiple co-folded predictions for binding modes.
• Evaluate various ligands for optimal pocket induction.

Topics

• AlphaFold3
• Cryptic Pockets
• Ligand Binding
• Protein Conformation
• Drug Discovery
• Structural Biology

Best for: AI Scientist, Research Scientist

Related on AIssential

• 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.
• AF2BIND: predicting small-molecule binding sites using the pair representation of AlphaFold2 — AlphaFold2's internal representations can be repurposed to accurately predict de novo small-molecule binding sites.
• 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.
• Contrastive Geometric Learning Unlocks Unified Structure- and Ligand-Based Drug Design — ConGLUDe unifies structure- and ligand-based drug design via contrastive geometric learning, enabling diverse drug discovery tasks.
• AlphaFold: Grand Challenge to Nobel Prize with John Jumper — AlphaFold's success demonstrates AI's profound utility in accelerating scientific discovery, even without full interpretability.
• AlphaFold hits ‘next level’: the AI database now includes protein pairing — The AlphaFold protein-structure database has been significantly expanded to include predictions of protein complexes, specifically 1.7 million "homodimers," which are crucial for understanding protein function.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.