A deconfounding adversarial classifier to encode gene expression profiles for drug repurposing

Despite the enormous increase of financial investments in pharmaceutical R&D, the number of newly approved drugs has greatly diminished during the past decades. Finding new uses for approved drugs, i.e. drug repurposing, has consequently become a major alternative strategy for the pharma industry seeking to speed up the development process, hence reducing costs while providing new treatments for unmet medical needs. 
This project aims at producing a next-generation gene expression based drug repurposing tool by automatically extracting features through a deconfounding adversarial deep learning framework. By building on our previous complementary approaches, we will develop an innovative, unifying tool that is able to both create an efficient network-like representation of the chemical space and directly predict new therapeutic applications.

Investigator: