Ligand-based pharmacophore modeling and Bayesian approaches to identify c-Src inhibitors.

Abstract Cellular Src (c-Src) kinases play a critical role in cell adhesion, proliferation, angiogenesis and cancer. Ligand-based pharmacophore models, used to identify the critical chemical features of c-Src inhibitors, were generated and validated by training, test and decoy sets, respectively. Best pharmacophore model, Hypo1, consists of four features such as HBA, HBD, Hy-Ar and RA. Hypo1 was used in virtual screening of the chemical databases such as Maybridge, Chembridge and NCI. The sorted compounds by Hypo1 were further reduced by applying drug-like properties and ADMET. Totally, 85 compounds which showed the good drug-like properties were selected from three databases and subjected to molecular docking for refinement of the retrieved hits by analysing the suitable orientation of the compounds in the active site of c-Src. Finally, 18 compounds were selected based on consensus scoring and hydrogen bond interactions with critical amino acids such as Met341, Thr338, Glu339 or Asp404. In addition, the Bayesian model was generated from the training set to find suitable fragments for inhibition of the c-Src function. Based on the above finding, we suggested that the Hypo1 and the good fragments from the Bayesian model will be helpful to select the compounds from various databases to identify the novel and potent c-Src inhibitor.