Table 2:

Dataset summary. Four datasets are utilized, each containing SMILES strings of inhibitors targeting DAT, NET, SERT, and hERG respectively. Alongside each SMILES string, the respective binding affinity in the unit of kcal/mol is also included as the label for each sample. Additionally, the final two columns represent the 10-fold cross validation Pearson correlation coefficient (R) and root-mean-square rrror (RMSE) for each binding affinity predictor across the four datasets.

Dataset name	Sample size	Binding affinity range (kcal/mol)	10-fold R	10-fold RMSE
DAT-Inhibitors	2662	[−14.18, −2.90]	0.8212	0.8979
NET-Inhibitors	2981	[−14.63, −5.47]	0.7732	0.9683
SERT-Inhibitors	4341	[−15.00, −5.64]	0.8022	0.9448
hERG-Inhibitors	6298	[−13.84, −3.27]	0.8092	0.7981