DRTerHGAT: A drug repurposing method based on the ternary heterogeneous graph attention network

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Drug repurposing accelerates development by identifying new uses for existing drugs.
• Computational methods efficiently screen potential drug-disease associations.
• Integrating drugs, proteins, and diseases into a unified model for drug repurposing is challenging.

Purpose of the Study:

• To develop an advanced computational drug repurposing method.
• To create a comprehensive model integrating drugs, proteins, and diseases.
• To improve the accuracy and efficiency of identifying potential drug candidates.

Main Methods:

• Proposed a ternary heterogeneous graph attention network (DRTerHGAT).
• Developed a novel protein feature extraction using a protein language model and multi-task autoencoder.
• Constructed a drug-protein-disease ternary heterogeneous graph with multiple relationship types.
• Utilized graph convolutional networks (GCN) and heterogeneous graph node attention networks (HGNA) for feature extraction.

Main Results:

• DRTerHGAT demonstrated superior performance compared to existing advanced methods and its variants.
• The method effectively extracted deep features from drugs, proteins, and diseases.
• Validated the model's efficacy in drug repurposing, specifically highlighting its application in Alzheimer's disease.

Conclusions:

• DRTerHGAT offers a powerful and effective approach for computational drug repurposing.
• The integration of protein features and complex relationships significantly enhances drug discovery.
• The model shows promise for identifying novel therapeutic strategies, particularly for diseases like Alzheimer's.