MINING THE NCI ANTICANCER DRUG DISCOVERY DATABASES - GENETIC FUNCTIONAPPROXIMATION FOR THE QSAR STUDY OF ANTICANCER ELLIPTICINE ANALOGS

The U.S. National Cancer Institute (NCI) conducts a drug discovery pro gram in which similar to 10 000 compounds are screened every year in v itro against a panel of 60 human cancer cell lines from different orga ns of origin. Since 1990, similar to 63 000 compounds have been tested , and their patterns of activity profiled. Recently, we analyzed the a ntitumor activity patterns of 112 ellipticine analogues using a hierar chical clustering algorithm. Dramatic coherence between molecular stru ctures and activity patterns was observed qualitatively from the clust er tree. In the present study, we further investigate the quantitative structure-activity relationships (QSAR) of these compounds, in partic ular with respect to the influence of p53-status and the CNS cell sele ctivity of the activity patterns. Independent variables (i.e., chemica l structural descriptors of the ellipticine analogues) were calculated from the Cerius(2) molecular modeling package. Important structural d escriptors, including partial atomic charges on the ellipticine ring-f orming atoms, were identified by the recently developed genetic functi on approximation (GFA) method. For our data set, the GFA method gave b etter correlation and cross-validation results (R-2 and CVR2 were usua lly similar to 0.3 higher) than did classical stepwise linear regressi on. A procedure for improving the performance of GFA is proposed, and the relative advantages and disadvantages of using GFA for QSAR studie s are discussed.