Title:	ANALYSIS OF CLUSTERS PERFORMANCE USING SEPARATION ACTIVES/INACTIVES AND MEAN INTERCLUSTER DISSIMILARITY
Author(s):	M. Rosmayati , A. B. Zuriana , C. A. Arifah , C. M. Noor Azliza
ISBN:	972-8924-16-X
Editors:	Pedro Isaías, Maggie McPherson and Frank Bannister
Year:	2006
Edition:	1
Keywords:	Ward’s clustering, genetic algorithm, chemoinformatic.
Type:	Full Paper
First Page:	441
Last Page:	448
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	Lead identification in drug discovery is a long and complicated process. The increasing numbers of molecules in chemical database make it harder to be screened. One of the solutions is by using compound selection methods where only a small portion of compounds is selected to represent the whole dataset. Cluster-based selection is one of the widely used methods in compound selection where the chemical compounds are located into dedicated clusters. In this paper, Ward’s clustering algorithm is chosen to cluster 2D fragment bit-strings. Genetic algorithm (GA) is applied on each cluster produced by Ward’s in order to optimise the performance of cluster in terms of separating actives molecules from inactives which is the main prerequisite in compound selection method. Another optimisation value used is mean intercluster dissimilarity. The result from optimisation of Ward’s clusters using GA is compared to the clusters produced by Ward’s alone. The optimisation of Ward’s clusters using GA shows better result based on the inter-cluster dissimilarity whilst Ward’s method is best at separating actives molecules from inactives with a small difference. These imply that the possibility of using GA on top of Ward’s clustering has potential to produce diverse dataset of compounds.
	Go Back