Graph Neural Network-Based Approaches to Drug Repurposing: A Comprehensive Survey

Abstract

Drug development is time-consuming and costly, with many efforts failing. About 90% of new molecules synthesized never reach the market, most of which have already undergone laboratory and animal testing phases. Drug repurposing, the use of drugs that successfully passed clinical trials for non-original purposes, reduces drug development time and cost. It is crucial for rare diseases and emerging diseases like COVID-19. Rare diseases lack sufficient incentives due to limited patient populations, while emerging diseases require urgent drug development.

Various approaches exist for drug repurposing, among which researchers highly favor computational methods due to their systematic approach and the elimination of laboratory costs. One of the leading approaches within computational methods is the application of artificial intelligence (AI). In these methods, AI models propose drugs for new therapeutic targets. Subsequently, these proposed drugs undergo evaluation through in-silico, in-vitro, in-vivo studies, and ultimately clinical trials.

Among AI-based methods, Graph Neural Networks (GNNs) have gained significant attention from researchers due to their ability to model problems in graph structures, which is highly suitable for drug repurposing tasks. This survey focuses on reviewing GNN-based approaches for drug repurposing. It provides a taxonomy of these models and examines various aspects such as model evaluation methods, implementation strategies, and datasets utilized in the field.