Nicotinamide phosphoribosyltransferase (NAMPT), an enzyme taking part in main NAD biosynthetic pathway, is an attractive target for anticancer therapy. The purpose of our study is to find novel NAMPT inhibitors based on in silico drug discovery means including the generation of 3D-QSAR models, and virtual screening techniques. Firstly, ten pharmacophore models were generated by Catalyst/HypoGen algorithm. Hypo1 with high correl value (0.96), large cost (77.77), and low root mean square deviation (0.81), featured by four chemical features was selected as the best one. Subsequently, Hypo1 was validated through test set prediction and Fischer's randomization methodologies. Then we screened some public compound libraries (Asinex, Ibscreen and Natural products database) using Hypo1 for a 3D query. The screened hits were further refined by Lipinski's rule of five, ADMET properties as well as molecular docking studies. Finally, six molecules with diverse scaffolds exhibited the right pharmacophore features and good binding modes between the receptor and ligands, and were selected as possible candidates against NAMPT for further study.