Dynamic pharmacophore model optimization: identification of novel HIV-1 integrase inhibitors

We extended the previously described dynamic pharmacophore model studies of HIV-1 integrase (IN) by considering more key residues in the active site, including Mg2+. First, we applied a Monte Carlo sampling method to map the complementary features of the IN binding surface. Two types of dynamic pharmacophore models were generated. One considers Mg2+ as part of the IN and therefore as an excluded volume, and the other treats Mg2+ as a positively charged feature, representing a new type of pharmacophore model aimed to identify compounds potentially preventing Mg2+ binding. Second, we validated the models with 385 known active (IC50 < 20 microM) and 235 (IC50 > 100 microM) inactive IN inhibitors. Third, we used the derived models to screen our small molecule database. Twenty-two structurally novel compounds were tested in an in vitro assay specific for IN, and two of them showed IC50 < or = 10 microM for strand transfer reaction.