Abstract
In this study we describe a method to identify important genes that appear to be involved in the cytotoxic mechanisms of key molecular fragments (biophores) contained within the structures of anticancer compounds. The anticancer biophores were mined by the MULTICASE program by analyzing 60 datasets containing 3271 compounds tested against the NCI-60 human cancer cell lines. For each identified fragment, statistically relevant genes were found by relating the activity profiles of the molecules containing the fragment and the gene expression profiles of the different cell lines. Microarray gene expression data of 13111 genes was used in conjunction with the LeFE algorithm to accomplish this task. We have demonstrated that regression analysis can then predict the cytotoxic activity of a compound in cell lines outside of those included in the regression model, even if it belongs to a different cancer type provided that the expression levels of identified genes are known for the cell lines. It is hoped that identifying key genes within the context of specific substructures responsible for the cytotoxic activity of anticancer agents could offer a better handle for designing specialized drugs targeting specific tumors based on their genetic profile.
Keywords: NCI-60, MULTICASE, anti-cancer, chemogenomics, cancer genes
3