Improved prediction of PARP inhibitor response and identification of synergizing agents through use of a novel gene expression signature generation algorithm.

Despite rapid advancement in generation of large-scale microarray gene expression datasets, robust multigene expression signatures that are capable of guiding the use of specific therapies have not been routinely implemented into clinical care. We have developed an iterative resampling analysis to predict sensitivity algorithm to generate gene expression sensitivity profiles that predict patient responses to specific therapies. The resultant signatures have a robust capacity to accurately predict drug sensitivity as well as the identification of synergistic combinations. Here, we apply this approach to predict response to PARP inhibitors, and show it can greatly outperforms current clinical biomarkers, including mutation status, accurately identifying PARP inhibitor-sensitive cancer cell lines, primary patient-derived tumor cells, and patient-derived xenografts. These signatures were also capable of predicting patient response, as shown by applying a cisplatin sensitivity signature to ovarian cancer patients. We additionally demonstrate how these drug-sensitivity signatures can be applied to identify novel synergizing agents to improve drug efficacy. Tailoring therapeutic interventions to improve patient prognosis is of utmost importance, and our drug sensitivity prediction signatures may prove highly beneficial for patient management.