Skip to main content
Fig. 1 | BMC Bioinformatics

Fig. 1

From: Evolutionary chemical binding similarity approach integrated with 3D-QSAR method for effective virtual screening

Fig. 1

Validation of different virtual screening methods for the kinase test set. The AUC values in the precision-recall curve calculated for data of each kinase are shown for each method tested. All methods were validated on the 51 selected kinases using their known inhibitors along with unknown molecules chosen based on random selection. a The LIGSIFT, ligand similarity using clique algorithm (LiSiCA) 2D, LiSiCA 3D, shape-it, and Tanimoto ligand-based structural similarity methods were used for comparison with the TS-ensECBS method. b The pharmacophore model (PharmaDB) incorporated in Discovery studio 2018, AutoDock VINA with both rigid and flexible side chains, the structure-based methods employed for comparison with the TS-ensECBS method. Two different test sets were used for ligand-based and structure-based methods. AUC values were calculated for each kinase and their distributions are shown in Figs. 1 and SI Fig. 6. The TS-ensECBS model clearly outperformed the ligand-based 2D and 3D chemical similarity and structure-based methods, although the pharmacophore model performed reasonably. However, molecular docking performed poorly, regardless of the sidechain flexibility in the binding site. Therefore, we decided to test the TS-ensECBS and receptor-based pharmacophore models further for unseen chemical databases

Back to article page