liuganghuggingface/moldata-muv

MUV maximum unbiased validation virtual screening

MUV maximum unbiased validation virtual screening. DeepChem/MoleculeNet release.

Part of the Molecular Datasets collection.

Get started

from datasets import load_dataset
dataset = load_dataset("liuganghuggingface/moldata-muv", "default")

Where the data comes from

• Original data: published with the original study
• Original study: Rohrer and Baumann. Maximum Unbiased Validation (MUV) Data Sets for Virtual Screening Based on PubChem Bioactivity Data (2009)
• Redistributed by: MoleculeNet (DeepChem) (dataset page)

Tables

Config	Splits	Rows	Targets
default	full	93,087	MUV-466 Dose Response Assays for S1P1 Agonists and Agonism Potentiators - Potentiator Assay 60K MLSMR (classification; binary), MUV-548 Dose-response biochemical assay for inhibitors of protein kinase A (PKA) activity (classification; binary), MUV-600 Dose-response cell-based assay for inhibitors of the nuclear receptor Steroidogenic Factor 1 (SF-1) (classification; binary), MUV-644 Dose-response biochemical assay of inhibitors of Rho kinase 2 (Rock2) (classification; binary), MUV-652 HIV-1 RT-RNase H MLSCN MH077605 Probe Assessment: Dose response Assay (classification; binary), MUV-689 Colorimetric assay for HTS discovery of chemical inhibitors of EphA4 receptor antagonists (classification; binary), MUV-692 Dose-response cell-based assay for activators of the nuclear receptor Steroidogenic Factor 1 (SF-1) (classification; binary), MUV-712 Tumor Hsp90 Inhibitors Dose Response Confirmation (classification; binary) and 9 more

data/ contains rows with readable molecular structures; identical observations are collapsed. observations/ contains all source rows.

Source label columns are text, including values such as <10. SMILES_original retains the input representation; SMILES_canonical provides a canonical representation. structure_key groups molecular components, and structure_status records parsing outcomes.

Data quality

• experiment: Labels are measured in a lab.
• all structures readable: All 93,087 SMILES were read with RDKit.
• 23.6% undefined stereo: 23.6% of molecules have a stereocenter whose configuration is not given, so mirror-image forms cannot be told apart.
• 39 stereoisomer sets agree: Found 39 groups of molecules that differ only in stereochemistry. Their values agree.
• 4.9% PAINS alerts: 4.9% of molecules contain a PAINS substructure, which often looks active in screening assays without real activity.
• 0.1% in rarer class: In a typical task, 0.1% of labeled molecules are in the rarer class. Accuracy is misleading here; use precision and recall.
• 15.7% measured per task: In a typical task, 15.7% of molecules have a value. A missing value means not measured, not inactive.
• 14,691 molecules per task: A typical task has 14,691 labeled molecules.
• no source split: Make your own split. A random split puts near-identical molecules in both training and test sets, so results look better than they will on new compounds. Splitting by scaffold, or by date when dates are given, gives a more realistic estimate.

Dataset notes

• Constructed benchmark with extremely few actives per assay. Decoy and selection procedures affect applicability; missing labels are not negatives.

License

Dataset-wide license: unspecified. Original source.

Citation

Cite the original study and data source:

• Rohrer and Baumann. Maximum Unbiased Validation (MUV) Data Sets for Virtual Screening Based on PubChem Bioactivity Data (2009)
• Wu et al. MoleculeNet: a benchmark for molecular machine learning (2018)
• Official DeepChem loader and task definitions