Artificial intelligence for drug discovery and clinical trials

Phase 3 outcomes clinical trials to evaluate new drugs, therapies and vaccines are among the most complex experiments performed in medicine. Around 50% of Phase 3 trials fail. US Food and Drug Administration’s recently published white paper, ‘22 case studies where Phase 2 and Phase 3 trials had divergent results’, a common theme is the difficulty of predicting clinical results in a wide patient base, even with the backing of solid data. More importantly, these barriers are costing healthcare industries, government and academic research hospitals billions of dollars, drive up costs, and delay life-saving treatments to patients and in some cases lead to adverse events. The crucial roadblock is the limited knowledge of key parameters, which need to be considered to test candidate molecules, eliminates adverse events and select optimal IC50’s. We are developing new machine learning architectures with automated learning and predictions that were gleaned from past experimental successes and failures of drugs leading to designing of faster, safer and more efficacious clinical trials. This work has significant impact on ethical and regulatory decisions facing patients, pharmaceutical industry and FDA.

AI for drug discovery sketch


  • 22 Case Studies Where Phase 2 and Phase 3 Trials had Divergent Results [PDF]


Other MIT Research Areas