OPTIMIZING STRUCTURE-BASED VIRTUAL SCREENING PROTOCOL TO IDENTIFY PHYTOCHEMICALS AS CYCLOOXYGENASE-2 INHIBITORS

By employing Databases of Useful Decoys (DUD) and its enhanced version (DUD-E), several attempts to construct validated Structure-based Virtual Screening (SBVS) protocols to identify cyclooxygenase-2 (COX-2) inhibitors have been performed. Both databases tagged active COX-2 inhibitors for compounds with IC50 values < 1mM. In the search for phytochemicals as natural COX-2 inhibitors, however, most of their IC50 values are in the micromolar range, which will likely be identified as non-inhibitors for COX-2 by the available SBVS protocols. In this article, validation of an SBVS protocol by adding marginal active COX-2 inhibitors from DUD-E as active compounds is presented. Binary quantitative-structure activity relationship analysis by using recursive partition and regression tree method was performed subsequently to optimize the predictive ability of the protocol. The enrichment factor and the F-measure values of the optimized protocol could reach 44.78 and 0.47, respectively. The optimized protocol could identify 1 out of 9 phytochemicals as COX-2 inhibitors