Exploratory chemoinformatic analysis of cell type-selective anticancer drug targeting.

In pharmaceutical development, structure-activity relationship studies aim to identify characteristics of chemical structures associated with well-defined activity end points. While this goal-driven approach is ideally suited for lead development purposes, a more exploration-driven approach is needed to discover cell type-selective drug targeting mechanisms in complex data sets. Growth inhibition profiles across different cancer cell lines are potentially informative with respect to molecular mechanisms targeting the activity of anticancer agents to specific tumor cells, yet only a small number of mechanistic associations between chemical structure and growth inhibition profiles have been discovered to date. Here, we have applied an exhaustive statistical analysis strategy to more than 10000 compounds in the NCI's anticancer agent database to identify molecular substructures associated with specific cytotoxicity signatures against a panel of human tumor-derived cancer cell lines (the Developmental Therapeutics Program 60-cell line panel). Some of the most significant substructures conferring cell type-selective cytotoxic activity include a large family of delocalized lipophilic cations; chloropurines, chloropyrimidines, and thiazoles; organosulfur chelators and organometallic complexes; and an unexpectedly related family of alkyl-lysophospholipids and phosphate prodrugs. Information from cell-based assays and gene expression measurements have been related to substructures represented in the chemical space covered by the library, yielding several candidate targeting mechanisms.