Computational Subgroups Stratification and Precision Drug Repositioning Using Explainable Artificial Intelligence

Enabling precision medicine requires developing robust patient stratification methods and identifying drugs tailored to homogeneous subgroups of patients from a heterogeneous population. Once discovering these subgroups, we can align patients and medications more specifically to achieve precision-based therapy. Developing de novo drugs is expensive and time-consuming with an ultimately low FDA approval rate. These limitations make developing new drugs for a small portion of a disease population unfeasible. Therefore, drug repositioning (DR) is an essential alternative for developing new drugs for a disease subpopulation. DR reduces the time, cost, and risk associated with a new drug. It is essential to develop data-driven approaches that find druggable homogeneous subgroups within the disease population and repurpose the candidate FDA-approved drugs for these subgroups. Different machine learning and data mining approaches have been developed for patient stratification, and DR. Their strategies mainly focused on subgroups with commonly known genotypic characteristics. Still, they may miss the importance of phenotypic characteristics during the stratification process. Additionally, the explainability of their results can be improved by including the ability to provide insight into the underlying biological mechanism unique to a subgroup and how the perturbation of biological entities contributes to drug selection for a given subgroup. My dissertation aims at providing an exploratory stratification and drug repositioning method by implementing the Explainable AI (XAI) concept using Data Mining and Network Analysis. The development and testing of this data-driven approach for patients’ stratification and drug repositioning can be done by accomplishing the following three aims: (1) Developing an explainable AI approach to identify druggable homogeneous subgroups; (2) Repositioning drugs for each subgroup; (3) Implementing the approach on a wide scale of diseases using pan-cancer analysis. Developing this method will provide the flexibility and explainability needed to give medical practitioners the ability to consider alternative treatments that remain specific for each patient.

Please contact Robert Sanders ( for Zoom information.