Table 6.

The selection of datasets on which models were trained, and details pertaining to these sets

Dataset	Tasks	Type	Compounds	Split	Metric
MUV	17	Classification	93,127	Random	PRC-AUC
HIV	1	Classification	41,913	Scaffold	ROC-AUC
BBBP	1	Classification	2053	Scaffold	ROC-AUC
Tox21	12	Classification	8014	Random	ROC-AUC
SIDER	27	Classification	1427	Random	ROC-AUC
QM8	12	Regression	21,786	Random	MAE
ESOL	1	Regression	1128	Random	RMSE
LIPO	1	Regression	4200	Random	RMSE