Computational Drug Repositioning for Triple-Negative Breast Cancers

(1) Triple-negative breast cancer (TNBC) is a severe clinical challenge in need of new therapies, but traditional drug development pipelines are time-consuming and expensive. More wholistic methods are needed to efficiently evaluate multiple drug targets in the context of TNBC. Drug response models aim to translate in vitro drug response measurements to in vivo drug efficacy predictions. While commonly used in retrospective analyses, my goal was to investigate the use of drug response modeling methods for the generation of novel drug discovery hypotheses in TNBC. (2) First, I review the current state of pan-cancer cell line screening datasets as these screening datasets are necessary for building drug response models. (3) Using one of these screening datasets, I generate models of drug response, which are then used to obtain imputed sensitivity scores for hundreds of drugs in over 1000 breast cancer patients. After examining the data for relationships between drugs and patient subtypes, I identified the Wee1 inhibitor AZD-1775 and an XPO1 inhibitor as compounds predicted to have preferential activity in TNBC. For AZD-1775, the imputed drug response formed significant associations with meaningful markers of drug response as well as the compound’s mechanism of action. AZD-1775 also efficiently inhibited the growth of preclinical TNBC models. (4) XPO1 in vitro inhibition also associated with the TNBC subtype. RNA-Seq analysis implicated two distinct mechanisms for XPO1 inhibition-mediated cell death, with the TNBC-based mechanism being consistent with the pan-cancer gene set associations. (5) Overall, the work here develops a framework to turn any cancer transcriptomic dataset into a dataset for drug discovery and shows the framework’s utility to quickly generate meaningful drug discovery hypotheses for a cancer population of interest