Drug repositioning based on weighted local information augmented graph neural network.

Area of Science:

• Computational biology
• Pharmacology
• Artificial intelligence in drug discovery

Background:

• Drug repositioning accelerates drug discovery by repurposing existing drugs for new therapeutic uses.
• Existing drug-disease association models often neglect crucial relationships between node embeddings.
• Developing advanced computational models is essential for efficient drug discovery pipelines.

Purpose of the Study:

• To propose a novel graph neural network model, DRAGNN, for enhanced drug repositioning.
• To improve the accuracy of drug-disease association predictions by incorporating relevant node embeddings.
• To accelerate the identification of new therapeutic applications for existing drugs.

Main Methods:

• Developed DRAGNN, a weighted local information augmented graph neural network.
• Incorporated a graph attention mechanism for dynamic node information collection.
• Emphasized heterogeneous and homogeneous information aggregation while omitting self-node aggregation.
• Utilized average pooling for neighbor information aggregation and a multi-layer perceptron for prediction.

Main Results:

• DRAGNN demonstrated effectiveness in drug repositioning through 10x10-fold cross-validation on three benchmark datasets.
• The model successfully predicted drug-disease associations, validated by authoritative data sources.
• Further validation included molecular docking and drug-disease network analysis, confirming model robustness.

Conclusions:

• DRAGNN offers a powerful computational approach for drug repositioning, improving prediction accuracy.
• The model's focus on node embeddings and local information enhances the identification of novel drug-disease relationships.
• This work provides a strong foundation for accelerating future drug discovery efforts through computational methods.