A Data-Mining Strategy That Identifies Drugs and Genes Associated With Anti-Cancer Drug Sensitivity

Abstract

The success of cancer therapy for patients often hinges on the responsiveness of the cancer cells to therapeutics. Drug resistance to anti-cancer therapeutics, both intrinsic and acquired, has important clinical and scientific significance. Identification of drug resistance genes using traditional methodologies and translation of those findings to the clinic has proven challenging. We developed a predictive data mining-based bioinformatic framework using public patient data and high-throughput cancer cell drug screening data. This information was used for genome-wide rankings of putative drug resistance genes. Prominent drug resistance genes (e.g. ABCB1, EGFR, and AXL) were successfully identified by the pipeline, additional genes hypothesized to be novel drug resistance genes were then investigated. Experimental confirmation of the novel genes using knockdown technologies indicated a propensity for of decreased proliferation/viability of cancer cells and increased sensitivity for anticancer compounds after knockdown much like known drug resistance genes. We then assessed the potential of each gene as an anti-cancer therapeutic target by exploring how gene knockdown behaved with clinical anticancer compounds. A second arm of the data-mining pharmaco-genomic strategy involved identification of candidate compounds that decrease expression of drug resistance genes. Using the drug resistance gene AXL as a proof-of-concept, three compounds were identified that decreased AXL expression at sub-micromolar concentrations. These compounds were characterized using microarray and cell signaling studies and found to decrease cell cycle signaling as well as activity of the Akt, mTOR, and ERK pathways. This study illustrates a novel approach for rapid and efficient identification of drug sensitivity genes or gene expression altering compounds utilizing bioinformatic data-mining.