Fig. 10
From: SMILES-based deep generative scaffold decorator for de-novo drug design
Histograms of different descriptors calculated in three sets of molecules obtained from the ChEMBL multi-step decorator model: Generated molecules from non-dataset scaffolds (blue), generated molecules from validation set scaffolds (orange) and training set molecules (green). a Molecular weight (Da); b ClogP; c Synthetic Accessibility Score; d Quantitative Estimate of Drug Likeness (QED). Notice that one of the filtering conditions of the ChEMBL subset was that the molecules had \(QED > 0.5\)