Determination of drug compound response is dependent on many factors that affect both therapeutic benefit and side effects of compounds to cancer patients. One major challenge of cancer therapeutics is the development of drug resistance. An increased understanding of the mechanisms of drug efficacy will help us to elucidate and predict the bases of drug resistance. The project here seeks to develop an integrative approach to correctly predict novel drug-gene interactions that influence the response to drug compounds. In this project, the molecular mechanisms of action (MOAs) of drug compounds will be examined through a method that combines drug compound structural features, compound activity and genomic profiles found in large molecular and pharmacological datasets with prior information in the form of literature curated pathway information and approved drug properties. This methodology will be useful for drug discovery purposes when used in conjunction with drug-screening libraries as well as for the prediction of off-target interactions for the re- purposing of approved drugs. This project seeks to determine omic and structural predictors of drug response for a publicly available 20,000 compound drug screening library of the NCI-60 cancer cell lines made available by the National Cancer Institute as part of the Development Therapeutics Program (DTP). Both the omic profiles and drug screening data will be obtained through the NCI CellMiner website. The NCI-60 cancer cell lines on which these compounds have been tested are well-characterized in terms of various genomic features that will serve as potential predictive features, including: gene expression and variant information, including copy number variants (CNV) and single nucleotide polymorphisms (SNP). The first aim of this project will be to perform a regression analysis to determine which features are predictive of drug response in the NCI-60 cell lines and examine how these features aggregate into higher order groups of interacting gene products (e.g. biological pathways). The second aim is to develop a methodology to determine predictors that identify direct drug targets from the sets of predictive features by examining compounds for drug-like characteristics and potentially druggable targets. The third aim of this project will be to experimentally validate a number of theresulting potentially direct drug-gene interactions with a focus towards novel interactions that involve cancer-relevant genes and potentially novel anti-cancer candidates.