Drug repurposing for cocaine use disorder (CUD) using a combined strategy of artificial intelligence (AI)-based prediction and retrospective clinical corroboration PROJECT SUMMARY/ABSTRACT Aim 1: Identify repurposed anti-CUD drug candidates using an AI-powered drug discovery approach Leveraging the unique and large-scale drug and disease phenotypic relationship knowledge bases that we have built and vast amounts of publicly available genetics and genomics data, we propose to develop an AI- powered drug repurposing system to identify anti-CUD drug candidates from all approved drugs. The output from Aim 1 is a list of promising repurposed anti-CUD candidates with interpretable mechanisms of action. Aim 2: Fine tune repurposed candidates by predicting their blood-brain barrier (BBB) permeability We will determine the BBB permeability of repurposed anti-CUD candidates identified in Aim 1 using a novel machine learning predictive model that we built, which applies to both small and macro-molecules that penetrate the human BBB through various biological mechanisms. The output from Aim 2 is a refined list of promising repurposed anti-CUD candidates with interpretable mechanisms of action and high BBB permeability in humans. Aim 3: Evaluate repurposed candidates using patient electronic health records (EHRs) We will evaluate repurposed anti-CUD candidates for their efficacy in ‘real world’ patients using patient electronic health record (EHR) data. Currently we have access to EHR data of 73.9 million unique patients including 223,460 patients diagnosed with CUD and 66,050 patients with a cocaine-positive urine drug screen. We will perform large-scale case-control studies to evaluate the efficacy of repurposed candidates in reducing risk, mortality, relapse, ER visits or other adverse effects of CUD patients. The output from Aim 3 is a further refined list of promising repurposed anti-CUD candidates with interpretable mechanisms of action, high BBB permeability in humans, and potential clinical efficacy in ‘real-world’ population. We will closely work with CTN and delineate the most expedient pathway to FDA approval for the identified candidates. We anticipate that these findings can be expeditiously translated into clinical trials in the CTN to benefit CUD patients.