Deep Learning-Based Drug Repurposing Using Knowledge Graph Embeddings and GraphRAG

Abstract

Drug repurposing has become a crucial strategy to accelerate drug discovery and reduce development costs. Conventional drug development is time consuming and expensive; it often takes more than a decade and billions of dollars to bring a new drug into the market. To address these challenges, this work puts forward a deep learning-based strategy for drug repurposing using knowledge graph embeddings and Graph-based Retrieval-Augmented Generation. The system leverages pretrained TransE model embeddings on the Drug Repurposing Knowledge Graph (DRKG) that includes over 97,000 biomedical entities and 4.4 million relationships. In the proposed framework, knowledge graph embeddings have been combined with LLMs to provide explainable predictions of drug repurposing. The system processes natural language queries about diseases, retrieves relevant drug candidates by deep learning based scoring, and generates interpretable explanations of drug disease relationships. This research combines PyTorch based embedding operations with NetworkX graph analysis and OpenAIs GPT models for an efficient, accurate, and explainable solution to automated drug repurposing, thus helping researchers and clinicians identify novel therapeutic applications for existing drugs.