Molecular properties

Tox21 Mitochondrial Membrane Potential

The Tox21 SR-MMP assay screens for compounds that decrease mitochondrial membrane potential in HepG2 cells. The benchmark uses molecular fingerprints to predict the assay's active or inactive label for each compound.

Binary classificationLicence: Tox21 / MoleculeNet distribution; underlying source terms applyMoleculeNet-Tox21-SR-MMPTarget: SR-MMPmoleculenet ↗

At this operating point

Model comparison

p=10,000, n=100 · Higher Macro-F1 is better

Random Forest is marked with a diamond. Models are ordered by the selected metric.

Across sample budgets

Learning curves

Models

All metric values at this operating point

Model	Macro-F1	MCC	Balanced accuracy	ROC-AUC
LimiX2	0.5727	0.1685	0.5731	0.7001
TabFM	0.5682	0.1517	0.5592	0.6561
Causilo	0.5676	0.1493	0.5673	0.6816
Kumo Tabular (M)	0.5576	0.1659	0.5548	0.6984
Extra Trees (tuned)	0.5542	0.1588	0.5566	0.6724
TabDPT	0.5515	0.1398	0.5493	0.614
RealTabPFN 2.5	0.5362	0.1475	0.5436	0.6921
MLP (tuned)	0.5332	0.115	0.5462	0.5509
AutoGluon	0.5313	0.08869	0.5327	0.6648
RealMLP	0.5292	0.1202	0.5361	0.6513
TabPFN 3.5	0.5257	0.145	0.5368	0.6836
GeneICL	0.5211	0.1296	0.5319	0.6635
Logistic Regression	0.511	0.1125	0.525	0.6552
RealTabPFN v2	0.5061	0.08495	0.5297	0.6897
XGBoost	0.5056	0.09688	0.5241	0.6468
XGBoost (tuned)	0.4977	0.08606	0.5197	0.6579
Logistic Regression (tuned)	0.4917	0.08562	0.5158	0.6574
TabM	0.4892	0.06031	0.5188	0.6469
Extra Trees	0.4855	0.09962	0.5137	0.6865
Random Forest	0.4851	0.09085	0.5135	0.6944
MITRA	0.4838	0.08918	0.5128	0.7017
TabPFN Wide 5k (ne3)	0.4837	0.06286	0.5127	0.6985
Random Forest (tuned)	0.471	0.05303	0.5065	0.6863
MLP	0.4637	0.02673	0.5031	0.5604