MGRL: Predicting Drug-Disease Associations Based on Multi-Graph Representation Learning.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug repositioning accelerates the discovery of new therapeutic applications for existing drugs.
• Traditional drug discovery is time-consuming and costly.
• Accurate computational models are crucial for efficient drug repositioning.

Purpose of the Study:

• To develop a novel computational method, MGRL, for predicting drug-disease associations.
• To enhance the efficiency and accuracy of drug repositioning strategies.
• To leverage multi-graph representation learning for drug-disease association prediction.

Main Methods:

• Utilized graph convolution networks for learning drug and disease representations from self-attributes.
• Employed graph embedding algorithms to model relationships between drugs and diseases.
• Integrated learned features into a random forest classifier for predictive modeling.

Main Results:

• MGRL achieved a high Area Under the Curve (AUC) of 0.8506 via five-fold cross-validation.
• Demonstrated superior performance compared to existing computational methods.
• Case studies confirmed the reliability and practical applicability of the MGRL method.

Conclusions:

• MGRL represents a pioneering approach in utilizing multi-graph learning for drug-disease association prediction.
• The method offers a significant advancement in computational drug discovery.
• MGRL provides a reliable and efficient tool for identifying new drug-disease links.