Figure 2:

2D molecular structures of reference compound with ChEMBL ID a) CHEMBL113621 from DAT-Inhibitors dataset, b) CHEMBL1275709 from NET-Inhibitors dataset, and c) CHEMBL173344 from SERT-Inhibitors dataset. 2D molecular structures are rendered by an online software SmilesDrawer 2.0 [25]. d) Distribution of experimental binding affinities for the four training datasets (DAT-Inhibitors, NET-Inhibitors, SERT-Inhibitors , and hERG-Inhibitors). e) Distribution of predicted binding affinities derived from the four deep neural network predictors. f) Distribution of predicted binding affinities for newly generated inhibitors targeting DAT, NET, SERT, and hERG. g) Screening of 330 preliminary multi-target drug candidates. The color of each point represents the predicted binding affinities to DAT (purple, g)), NET (green, h)), and SERT (blue, i)). The light purple, green, and blue frames outline the medium ranges of 10 ADMET, physicochemical, and medicinal chemistry properties, respectively, while the dark purple, dark green, and dark blue frames outline the excellent ranges of these properties.