Computational approach to the identification of novel Aurora-A inhibitors

Aurora kinase A has been emerging as a key therapeutic target for the design of anticancer drugs. For the purpose of finding biologically active and novel compounds and providing new ideas for drug-design, we performed virtual screening using commercially available databases. A three-dimensional common feature pharmacophore model was developed with the HipHop program provided in the Catalyst software package, and this model was used as a query for screening the databases. A recursive partitioning (RP) model was developed as a filtering system, which was able to classify active and inactive compounds. Eventually, a step-wise virtual screening procedure was conducted by applying the common feature pharmacophore and the RP model in succession to discover novel potent Aurora-A inhibitors. A total of 68 compounds were selected for testing of their in vitro anticancer activities against various human cancer cell lines. Based on the activity data, we have identified fifteen compounds that warrant further investigation. Several compounds have a high inhibition rate (above 80% at 10 μM) and a GI 50 lower than 5 μM for the cell lines DU145 and HT29. Enzyme assay for these compounds identified hits with micro molar activity. Compound C11 has the highest activity (IC 50 = 5.09 μM). The hits obtained from this screening scheme could be potential drug candidates after further optimization.