Affordable high-throughput genome sequencing technologies had been expected to usher an era of precision medicine, whereby each patient receives an individualized treatment based on their genetic profile. However, precision medicine is yet to realize this potential and the number of clinically applied drug biomarkers lags far behind the number proposed in the scientific literature. Lack of reproducibility of proposed biomarkers has been highly problematic, and in particular the lack of effectiveness of biomarkers identified in pre-clinical studies when applied in clinical trial. In cancer, the number of drugs with FDA approved biomarkers remains at less than 30, and almost all of these arose because a drug was designed to target a specific known driver gene. Overall, only a handful of clinically actionable biomarkers have been discovered for drugs already in use. The objective of this project is to develop computational methods that will improve on existing biomarker discovery, and ultimately improve patient care and survival.
My first aim i s to develop an approach that will allow me to impute drug sensitivity in very large clinical datasets, such as TCGA. This imputed data will then be compared to measured markers in these data (e.g. somatic mutations) in order to identify novel predictors of drug response. The statistical models used to impute drug sensitivity will be developed in a pre-clinical disease model, where drug response has been accurately measured. In my second aim I will quantify the contribution of germline genetic variation to drug response in cancer. ?Germline genetic variation? refers to the common genetic differences that exist between individuals, which are for the most part preserved in tumors. The contribution of this inter-individual genetic variation to variability in drug response between patients remains unquantified, although it is becoming clearer that it plays an important role. Finally, I will combine the information gathered in the first steps and build effective integrative models of drug response. Following the R00 phase (R01 and beyond), I envision that such models will be tested in patients; subsequently, measured drug response in these cohorts will be used to further refine future predictions. This will result in a framework where models actively improve their predictive performance as more data is gathered over time. To ensure the success of this project, I have assembled an expert mentorship team for the K99 phase, which includes a statistician, wet-lab biologists and clinicians, thus covering the entire scope of the precision medicine discovery and implementation pipeline. The University of Chicago, a world-class research institution, will provide the ideal environment and resources for such a cross-disciplinary, collaborative endeavor.
Despite enormous investment in precision medicine, most drug repurposing and biomarker discovery strategies have failed when applied in the clinic. This project will improve drug treatments by developing novel computational methods that will allow us to find new biomarkers and to better predict which patients will respond to what drugs, with application to cancer. The novel approach is based on integrating clinical sequencing and pre-clinical drug screening datasets.