Drug Repositioning via Multi-View Representation Learning With Heterogeneous Graph Neural Network.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning accelerates drug development by identifying new uses for existing drugs.
• Current computational methods for predicting drug-disease associations face accuracy challenges.
• High costs and long timelines of conventional drug development necessitate efficient computational approaches.

Purpose of the Study:

• To propose an advanced computational method, MRLHGNN, for accurate drug repositioning.
• To enhance the prediction of drug-disease associations using multi-view biological data.
• To overcome limitations of existing methods in drug repositioning.

Main Methods:

• Developed a Multi-view Representation Learning with Heterogeneous Graph Neural Network (MRLHGNN) model.
• Employed view-specific feature aggregation with variable-length meta-paths for expanded local receptive fields.
• Utilized a transformer-based semantic aggregation module and a multi-view fusion decoder with attention.

Main Results:

• MRLHGNN demonstrated superior effectiveness and interpretability compared to nine state-of-the-art methods in cross-validation experiments.
• The method accurately predicts potential drug-disease associations.
• Case studies confirmed MRLHGNN's utility as a powerful drug repositioning tool.

Conclusions:

• MRLHGNN offers a significant advancement in computational drug repositioning.
• The proposed method effectively leverages multi-view biological data for improved drug-disease association prediction.
• MRLHGNN provides a robust and interpretable platform for identifying novel therapeutic applications of existing drugs.