Skip to main content
. 2020 Feb 4;8:46. doi: 10.3389/fchem.2020.00046

Figure 1.

Figure 1

(A) Generation process of GDBChEMBL. (B) CLscore distributions for GDB17, its subsets FDB17 and GDBMedChem, and public databases ChEMBL, ZINC, and DrugBank. (C) Frequency distribution of molecular shingles up to a diameter of 6 bonds in ChEMBL. (D) SAscore vs. CLscore in various databases. A lower SAscore indicates higher synthetic accessibility, and a higher CLscore indicates higher similarity to ChEMBL molecules. (E) Occupancy of triplet value bins (HAC, stereocenters, heteroatoms) in all GDB17 cpds with CLscore ≥3.3 (black line) and after uniform sampling forming GDBChEMBL (red line).