Enhancing Drug Repurposing through Graph Neural Networks and Knowledge Graphs
Authors: Y. Zhang, L. Wang, H. Chen, R. Liu
Drug repurposing is a promising strategy to accelerate drug development by identifying new uses for existing drugs. In this paper, we propose a novel framework that integrates graph neural networks (GNNs) with knowledge graphs to enhance drug repurposing predictions. Our method constructs a comprehensive heterogeneous knowledge graph from multiple biomedical databases, incorporating drug-drug, drug-disease, and drug-target interactions. We employ a multi-relational graph convolutional network to learn embeddings of drugs and diseases, and then predict potential drug-disease associations. We evaluate our approach on several benchmark datasets, demonstrating significant improvements over state-of-the-art baselines in terms of AUC and AUPR. Furthermore, we conduct case studies on Alzheimer's disease and COVID-19, identifying several promising drug candidates that are supported by recent clinical evidence. Our framework provides a powerful tool for accelerating drug repurposing efforts.
