EKGDR: An End-to-End Knowledge Graph-Based Method for Computational Drug Repurposing.

Area of Science:

• Computational drug discovery
• Pharmacology
• Bioinformatics

Background:

• Traditional de novo drug development is lengthy, costly, and has a high failure rate.
• Drug repurposing offers a more efficient and cost-effective alternative by identifying new uses for existing drugs.
• Challenges in computational drug design include data heterogeneity and limited known drug-disease interactions.

Purpose of the Study:

• To introduce EKGDR, an end-to-end knowledge graph-based computational drug repurposing approach.
• To address data heterogeneity and limited interaction data challenges in drug repurposing.
• To improve the prediction of drug-disease interactions.

Main Methods:

• Utilized a drug knowledge graph integrating drug interactions, categorization, and molecular descriptors.
• Employed graph neural networks for end-to-end embedding of the knowledge graph.
• Learned drug-disease interaction intents by aggregating relational messages along multihop paths.

Main Results:

• EKGDR achieved superior performance in predicting drug-disease interactions.
• Achieved AUROC of 0.9475, AUPRC of 0.9490, and Recall@200 of 0.8315.
• Demonstrated effectiveness by identifying candidate drugs for Alzheimer's and Parkinson's diseases.

Conclusions:

• EKGDR represents a significant advancement in computational drug repurposing.
• The knowledge graph and graph neural network approach effectively predicts drug-disease interactions.
• EKGDR shows promise for accelerating the discovery of novel therapeutic applications for existing drugs.