Lanosterol 14alpha-demethylase (CYP51) is an important target for antifungal drugs. An improved three-dimensional model of CYP51 from Candida albicans (CACYP51) was constructed by ligand-supported homology modeling and molecular dynamics simulations. The accuracy of the constructed model was evaluated by its performance in a small-scale virtual screen. The results show that known CYP51 inhibitors were efficiently discriminated by the model, and it performed better than our previous CACYP51 model. The active site of CACYP51 was characterized by multiple copy simultaneous search (MCSS) calculations. On the basis of the MCSS results, a series of novel azoles were designed and synthesized, and they showed good in vitro antifungal activity with a broad spectrum. The MIC(80) value of four of these compounds against C. albicans is 0.001 microg mL(-1), indicating that they are promising leads for the discovery of novel antifungal agents.