Mobile Menu

Deep learning for drug repurposing

Drug repurposing is an effective strategy to identify new uses for existing drugs. Ohio State University researchers have now created a framework that combines patient-related datasets with high-powered computation to arrive at repurposed drug candidates.

Drug repurposing

Drug repurposing (or drug repositioning) is a strategy to accelerate the drug discovery process. This approach identifies novel uses for already existing approved drugs. The primary advantage of this approach over existing traditional drug development is that it starts from compounds with well-characterised pharmacology and safety profiles. This significantly reduces the risk of adverse side effects and attrition in clinical phases.

Researchers have found many successful repurposed drugs serendipitously. However, in recent years, researchers have developed computation-based repurposing methods to leverage pre-clinical information, such as GWAS and gene expression data. Nonetheless, pre-clinical outcomes are not always consistent with clinical therapeutic effects in humans. As a result, attention has shifted towards the use of real-world data (RWD), such as EHRs and patient surveys. As RWDs are direct observations from human bodies, they represent a promising source for drug repurposing.

Deep learning framework

In this study, published in Nature Machine Intelligence, researchers created an efficient and easily customised framework for generating and also testing multiple candidates for drug repurposing using a retrospective analysis of RWD. Specifically, they applied deep learning and causal inference methodologies to control the confounders in RWD and systematically estimate the drug effects on various disease outcomes. The team’s framework emulates randomised clinical trials for drugs present in a large-scale medical claims database.

The team utilised their framework on insurance claims from nearly 1.2 million heart-disease patients. This provided information on their assigned treatment, disease outcomes and also other potential confounders. From this, the team were able to successfully identify drugs and drug combinations that substantially improved coronary artery disease outcomes but were not yet indicated for treating the disease.

Ping Zhang, senior author and Assistant Professor of Computer Science and Engineering and Biomedical Informatics at Ohio State, stated:

“My motivation is applying this, along with other experts, to find drugs for diseases without any current treatment. This is very flexible, and we can adjust case-by-case.

The general model could be applied to any disease if you can define the disease outcome.”

Image credit: By starline –

Share this article