Application of fuzzy clustering analysis to compound datasets for drug lead identification

Recently, the increasing number of chemical compound datasets to be screened has been growing rapidly due to the fast developments of high-throughput screening in drug discovery. These compound datasets requires compound selection methods which have become one of the main technique in drug discovery especially in drug lead identification process. Thus, finding the best method in compound selection is needed to the pharmaceutical industry to ensure the accurate results of this process. One of most used compound selection method is cluster-based compound selection, which involves subdividing a set of compounds into clusters and choosing one compound or a small number of compounds from each cluster. In this cluster-based compound selection, non-overlapping methods such as Ward's, Group Average, Jarvis Patrick's and K-means are preferred methods to cluster the diverse set of compounds. However, there are little study on overlapping method such as fuzzy c-mean (FCM) and fuzzy c-varieties (FCV) clustering algorithms. Therefore, these two clustering algorithms are applied and their performance is compared based on the effectiveness of the clustering results in terms of separation between actives and inactives (Pa) into different clusters and mean intercluster molecular dissimilarity (MIMDS). The analysis shows FCM gives the best results compare to FCV in terms of Pa indicating that FCM has a promising use in compound selection algorithms. But, FCV is perform better than the FCM in term of MIMDS when a higher number of compounds and higher fuzziness index value are concerned.