PoseX (Protein-Ligand Docking Benchmark)
Open-source benchmark for protein-ligand docking evaluating 23 methods. 718 self-docking and 1,312 cross-docking entries from PDB structures 2022-2025 (prevents data leakage). Compares physics-based, AI docking, and AI co-folding approaches.
Composite
91.4
Experimental validation
Retrospective
Stages
Hit IDLead ID / ADMET
Modalities
protein_structuresmall-molecule
Task types
dockingpose_prediction
Size
self-dock_+_1: 718
cross-dock_entries: 312
methods: 23
cross-dock_entries: 312
methods: 23
License
MIT
First release
2025-05
Last updated
2025-12
Official site
Leaderboard
Dataset
Code / GitHub
HuggingFace
→ HF
Paper
PoseX: A Comprehensive Benchmark for Protein-Ligand Docking · · 2025 · paper · 10 citations
Flags
leakage_preventioncross_dockingleaderboard
Experts
—
Groups
—
Hosted by
—
Related benchmarks
Rubric (7-criterion)
rigor
5
coverage
4
maintenance
4
adoption
4
quality
5
accessibility
5
industry_relevance
5
Notes
Key findings: AI surpasses physics-based docking overall; relaxation crucial for AI-generated poses; pocket specification boosts performance; some co-folding methods struggle with chirality. Uses only 2022-2025 PDB structures to prevent train/test contamination.