GuacaMol

Goal-directed + distribution-learning benchmarks for molecular generative models.

Composite
80.5
Experimental validation
Retrospective
Stages
Lead ID / ADMETDevelopmental Candidate
Modalities
small-molecule
Task types
molecule-generation
Size
tasks: 20
train_set: 1,600,000
License
MIT
First release
2019-03
Last updated
2022-07
Official site
→ project page
Leaderboard
→ leaderboard
Dataset
→ dataset
Code / GitHub
→ repository
HuggingFace
→ HF
Paper
GuacaMol: Benchmarking Models for de Novo Molecular Design · Brown N, Fiscato M, Segler MHS, Vaucher AC · 2019 · paper · doi:10.1021/acs.jcim.8b00839 · 820 citations
Flags
none
Experts
Marwin Segler, Nathan Brown
Groups
BenevolentAI
Hosted by
MoleculeNet, Therapeutics Data Commons (TDC)
Related benchmarks
MOSES, Practical Molecular Optimization (PMO)

Rubric (7-criterion)

rigor
4
coverage
4
maintenance
2
adoption
5
quality
4
accessibility
5
industry_relevance
4

Notes

First-generation generative benchmark; largely superseded by PMO for goal-directed.

← Back to all benchmarks

Compare:
Open comparison →