Molecular properties

Blood-Brain Barrier Penetration (Martins)

The Martins dataset labels compounds by whether they cross the blood-brain barrier. The benchmark uses molecular fingerprints to distinguish compounds labelled as penetrating the barrier from those labelled as non-penetrating.

Binary classificationLicence: CC BY 4.0 via TDC; underlying source terms require verification before re-hostingTDC-BBB-MartinsTarget: 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
LimiX2	0.7348	0.5046	0.7249	0.7959
GeneICL	0.7329	0.5337	0.7005	0.7954
RealMLP	0.7275	0.4751	0.7048	0.7641
Kumo Tabular (M)	0.7241	0.5063	0.696	0.7908
RealTabPFN 2.5	0.7159	0.5098	0.6822	0.7669
TabPFN 3.5	0.714	0.5102	0.6826	0.7941
TabDPT	0.7128	0.461	0.6904	0.7679
Logistic Regression (tuned)	0.71	0.4706	0.6909	0.7799
TabFM	0.7029	0.4397	0.6842	0.7816
AutoGluon	0.6999	0.4686	0.6748	0.7877
Causilo	0.6985	0.4084	0.6899	0.7707
MITRA	0.696	0.4881	0.6651	0.7706
Logistic Regression	0.6889	0.4688	0.661	0.7793
MLP	0.6883	0.433	0.6709	0.7379
Extra Trees (tuned)	0.6862	0.4588	0.6608	0.7618
MLP (tuned)	0.6797	0.4008	0.6696	0.7308
RealTabPFN v2	0.6788	0.4596	0.651	0.7747
TabM	0.676	0.4548	0.6492	0.7885
TabPFN Wide 5k (ne3)	0.6742	0.4587	0.6469	0.7833
Random Forest	0.669	0.4563	0.6425	0.7789
Extra Trees	0.663	0.4477	0.6382	0.7831
XGBoost	0.6576	0.4031	0.6395	0.735
Random Forest (tuned)	0.6474	0.4269	0.6266	0.7759
XGBoost (tuned)	0.645	0.4067	0.6257	0.7313