Pharmacophore models of β-secretase inhibitors were developed by using Catalyst HypoGen program with a training set of 25 compounds (IC50 values from 0.002μmol·L-1 to 25 μmol•L－1) containing 4 different kinds of structures. A fitting pharmacophore hypothesis (Ccorrel＝0.969, Config＝16.32, Δcost＝62.422) was characterized by one aromatic ring center, one aliphatic hydrophobic core, one pos positively ionizable center and one hydrogen bond donor. The model was applied in to predicting the activity of other 209 inhibitors as a test set. The result indicates that the pharmacophore model exhibits good predictive ability and is able to provide clear guidelines for screening new lead compounds of β-secretase inhibitors.

Key words: pharmacophore model, β-secretase, nonpeptidomimetic inhibitor