The 3D pharmacophore model of DPP-Ⅳ inhibitors was established using the Discovery Studio software with the training set of 20 DPP-Ⅳ inhibitors. The best pharmacophore hypothesis (Hypo 1) consists of one hydrogen-bond acceptor, one hydrophobic point, one positive ionizable group, one aromatic ring as well as five excluded volumes. Fischer's validation clearly shows that proposed Hypo 1 has highly predictive ability and can be efficiently used as a 3D query for virtual screening to retrieve potential inhibitors from ZINC databases. The hit compounds subsequently were docked into the DPP-Ⅳ active site and 21 compounds were obtained based on PLP2 scoring function. Therefore, this study could provide scientific basis for denovo design of DPP-Ⅳ inhibitors.