DR-189 Validating the Utility of Meta-Analysis for Learning Translatable Predictive Models from in vitro Pharmacogenomics

Several large initiatives have profiled cancer cell lines – lab grown models of disease – for their response to panels of drugs. These studies aim to identify genomic features that differentiate responders from non-responders. However, translating predictions back to patients has been challenging due to the differences between cell lines and patients, as well as experimental artifacts. We build upon previous work in integrating across studies from different labs to evaluate whether cross lab analysis results leads to better predictions. The genomic and treatment data collected by the Hartwig Foundation (HMF) will allow us to test predictions learned from cell lines.

Benjamin Haibe-Kains University Health Network Canada