中央研究院 資訊科學研究所

In this talk, I will discuss the identification of the drug response genes of tyrosine kinase inhibitors Erlotinib and Sorafenib for lung cancer. We showed that these genes can perfectly classify a small set of patients from the BATTLE clinical trial who were without biomarker mutations into responding and non-responding using machine learning and regression approaches. Using pathway and protein-protein interaction network analysis, we further showed that these genes are associated with kinase signaling pathways. Specifically, Erlotinib-associated genes are related to protein ubiquitination and DNA damage, and Sorafenib-associated genes are related to autophagy, apoptosis and proliferation, shedding light on causes for drug resistance. Our analysis also suggests that ADORA3 could be is a good biomarker and target to counter Erlotinib resistance. Interestingly, our study suggests that some of these Erolotinib/Sorafenib-associated genes not only are drug-response biomarkers, but also are prognostic biomarkers for EGFR wild-type lung cancer patients. Our ML (machine learning)-based drug response model may contribute significantly to precision medicine treatment of lung cancer patients.

【黃明經 博士 實驗室】