Molecular properties

Ames Mutagenicity

The Ames dataset records whether compounds cause mutations in a bacterial test. Using molecular fingerprints, the benchmark predicts the mutagenic or non-mutagenic label assigned to each compound.

Binary classificationLicence: CC BY 4.0 via TDC; underlying source terms require verification before re-hostingTDC-AMESTarget: Ytdc ↗

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
Kumo Tabular (M)	0.6437	0.2921	0.6439	0.7111
TabPFN 3.5	0.6387	0.288	0.64	0.7113
MITRA	0.6341	0.2729	0.6352	0.6846
MLP	0.6306	0.2677	0.6318	0.6973
Extra Trees (tuned)	0.6291	0.2676	0.6304	0.6998
Causilo	0.6238	0.2541	0.6246	0.6878
Logistic Regression	0.6216	0.2476	0.622	0.679
Random Forest	0.6215	0.2609	0.6263	0.7011
GeneICL	0.6214	0.2457	0.6231	0.6823
Extra Trees	0.6197	0.2591	0.6249	0.7024
RealMLP	0.6191	0.2511	0.623	0.6731
TabFM	0.6187	0.2385	0.6185	0.6736
Random Forest (tuned)	0.6183	0.249	0.6212	0.6937
AutoGluon	0.6178	0.2638	0.6258	0.6889
XGBoost	0.6168	0.2356	0.6168	0.6645
RealTabPFN 2.5	0.6152	0.2486	0.6206	0.6882
TabM	0.6068	0.2476	0.6253	0.6946
Logistic Regression (tuned)	0.6055	0.2512	0.6151	0.685
TabDPT	0.5949	0.2134	0.6054	0.6344
LimiX2	0.5844	0.2444	0.6094	0.7007
RealTabPFN v2	0.5782	0.2235	0.5982	0.6914
MLP (tuned)	0.5751	0.2189	0.6057	0.6965
XGBoost (tuned)	0.5694	0.2038	0.5999	0.6517
TabPFN Wide 5k (ne3)	0.4706	0.1075	0.559	0.6854