Dhananjay Bhaskar, PhD

Topic: Explainable machine learning models for indication expansion

Project Summary: Bringing a novel therapeutic to market is a time-consuming and expensive endeavor. Many promising candidates identified through virtual screening and preclinical studies fail in clinical trials due to poor efficacy or lack of improvement in the standard of care. Instead, a target-centric approach, based on the repurposing of safe compounds with known mode of action (MoA), offers many advantages. First, alternative indications (diseases) with high medical need and market potential are identified for a given target, by linking information about the MoA, disease state, and patient populations obtained from large public and proprietary datasets. Subsequently, suitable disease models are chosen (e.g. by literature mining), and therapeutics are fast-tracked for approval, thus limiting the risk of failure. The aim of this project is to develop and integrate novel machine learning methods, with emphasis on explainability of the predicted outcome, to improve the overall performance of Boehringer Ingelheim’s drug repurposing pipeline.