MoleculeNet

Multi-task molecular ML benchmark covering quantum, physical, biophysical, physiological properties.

Composite

78.0

Experimental validation

Retrospective

Stages

Lead ID / ADMETHit ID

Modalities

small-molecule

Task types

regressionclassificationquantum

Size

tasks: 17
molecules: 700,000

License

MIT

First release

2018-03

Last updated

2023-11

Official site

→ project page

Leaderboard

→ leaderboard

Dataset

→ dataset

Code / GitHub

→ repository

HuggingFace

→ HF

Paper

MoleculeNet: A Benchmark for Molecular Machine Learning · Wu Z, Ramsundar B, Feinberg EN, et al. · 2018 · paper · doi:10.1039/C7SC02664A · 3,600 citations

Flags

data-leakage-known

Experts

Bharath Ramsundar, Vijay Pande

Groups

Pande Lab (Stanford, alumni network), DeepChem Community

Hosted by

MoleculeNet, DeepChem, Papers With Code — Drug Discovery

Related benchmarks

TDC ADMET Group, Polaris ADMET

Rubric (7-criterion)

rigor

coverage

maintenance

adoption

quality

accessibility

industry_relevance

Notes

Widely cited (3600+); aging splits with known scaffold leakage.

← Back to all benchmarks

Compare:

Open comparison →