ADMET-AI

Graph-based ADMET prediction + benchmark on 41 endpoints leveraging ChEMBL + TDC.

Composite
79.5
Experimental validation
Retrospective
Stages
Lead ID / ADMET
Modalities
small-molecule
Task types
regressionclassification
Size
endpoints: 41
molecules: 90,000
License
MIT
First release
2023-11
Last updated
2024-12
Official site
→ project page
Leaderboard
→ leaderboard
Dataset
→ dataset
Code / GitHub
→ repository
HuggingFace
→ HF
Paper
ADMET-AI: a machine learning ADMET platform for evaluation of large-scale chemical libraries · Swanson K, Walther P, Leitz J, et al. · 2023 · paper · doi:10.1093/bioinformatics/btae416 · 85 citations
Flags
none
Experts
Regina Barzilay
Groups
MIT CSAIL / Jameel Clinic / Coley Lab, Barzilay Group (MIT)
Hosted by
Related benchmarks
TDC ADMET Group, MoleculeNet

Rubric (7-criterion)

rigor
4
coverage
4
maintenance
4
adoption
3
quality
4
accessibility
5
industry_relevance
4

Notes

Strong baselines + web tool; builds on TDC.

← Back to all benchmarks

Compare:
Open comparison →