A Machine Learning approach for assessing drug development risk

ABSTRACT

Identifying new drug targets and developing safe and effective drugs is both challenging and risky. Furthermore, characterizing drug development risk - the probability that a drug will eventually receive regulatory approval - has been notoriously hard given the complexities of drug biology and clinical trials. This inherent risk is often misunderstood and mischaracterized, leading to inefficient allocation of resources, and, as a result, an overall reduction in R&D productivity. We propose a Machine Learning (ML) approach that provides a more accurate and unbiased estimate of drug development risk than traditional models.