A Target-Specific Evidence Function for Indication Expansion Queries in the Open Targets Platform

Drug indication expansion and repositioning leverage investments in drug discovery and development research to include greater numbers of target-associated diseases in scientific and regulatory approval workflows. Yet time and resources are required to identify and evaluate each potential opportunity. Open Targets is a large-scale public-private partnership which provides platform access to integrated public domain information for target, disease, and drug associations. Open Targets was designed to support expanded target identification, prioritisation, and validation using aggregate evidence association networks. Association scores for target-disease pairs in Open Targets are available for literature, RNA expression, genetic, somatic mutation, animal model, affected pathway, and known drug evidence. However, association score distributions are highly variable, overlap for target-indication pairings with and without associated drugs, and are complex due to the diverse underlying target-disease biological relationships. An ongoing challenge for users of Open Targets remains how to efficiently identify meaningful opportunities for indication expansion for a given target from association evidence scores. In this project, we propose and evaluate a target-specific evidence function for therapeutic indication status prediction and apply this in indication expansion queries. We find that our approach outperforms a harmonic sum benchmark and further demonstrate how association heterogeneities revealed by this method may be used to enhance indication expansion workflows for Open Targets.