Machine Learning for Healthcare and Life Sciences

Drug Repurposing

Drug Repurposing

Finding new indications for existing drugs is a promising venue for pharma companies when it comes to reducing drug development costs. In recent years, 30% of regulatory approvals by the FDA have been for new uses of previously approved drugs and vaccines1. A repurposing approach to drug discovery and development can streamline the time- and cost-intensive process of bringing new therapies to market, which can take the industry up to 20 years and cost in excess of $2.5 billion.

Currently, the discovery of new therapeutic uses for existing medicines is largely the result of serendipitous findings or isolated research. Our collaborations with pharma companies work to design, build, and deploy a systematic process for drug repurposing, potentially becoming a blueprint for use across the industry.

Our systematic approach for drug repurposing combines human insight with unique machine-learning and causal inference algorithms. We apply the algorithms on vast amounts of observational real-world data accessed through IBM Watson Health, as well as on drug information from pharmacological knowledge bases, such as DrugBank, to test hundreds of candidates for repurposing in various disease domains.

1. See http://dx.doi.org/10.1016/j.drudis.2013.11.005