AI models outperform physics-based methods in protein-ligand docking

A new benchmark called PoseX has been developed to evaluate protein-ligand docking methods, comparing AI approaches against traditional physics-based techniques. Experiments using PoseX demonstrated that AI methods generally outperform physics-based approaches in docking success rates. The research also found that combining AI modeling with physics-based post-processing, particularly after addressing ligand chirality issues in AI co-folding methods, can significantly improve results. AI

IMPACT AI methods show superior performance in protein-ligand docking, suggesting a shift from physics-based approaches in drug discovery.

RANK_REASON The cluster describes a new benchmark and research findings published on arXiv.

Read on arXiv cs.LG →

• AlphaFold3
• Boltz-1x
• DiffDock
• Schrödinger Glide

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →