A three-dimensional (3D) quantitative pharmacophore model was developed from a training set of structurally diverse substituted ureas against Raf-1 kinase, which was well validated to be highly predictive by two methods, namely, test set prediction and Cat-Scramble method. Then a virtual database searching was performed with the pharmacophore model as a 3D query, and the bioactivities of the retrieved hits were predicted by the pharmacophore. Subsequently, molecular docking was carried out on the selected hits whose estimated IC(50) was less than 1 microM. Finally, 29 hits were identified as potential leads against Raf-1 kinase, which exhibited good estimated activities, high docking scores, similar binding mode to experimentally proven compounds and favorable drug-like properties. The study may facilitate the discovery and rational design of novel leads with potent inhibitory activity targeting Raf-1 kinase.