DRONet: effectiveness-driven drug repositioning framework using network embedding and ranking learning.

Area of Science:

• Pharmacology and Drug Discovery
• Bioinformatics and Computational Biology

Background:

• Drug repositioning identifies new uses for existing drugs, crucial for efficient drug development.
• Current methods lack prediction accuracy and fail to incorporate drug effectiveness information, limiting reliable results.

Purpose of the Study:

• To propose DRONet, a novel framework for drug repositioning that leverages effectiveness comparative relationships (ECR).
• To enhance the prediction performance of identifying new drug indications.

Main Methods:

• Utilized network embedding to learn deep drug features from a heterogeneous drug-disease network.
• Constructed a drug-indication dataset incorporating effectiveness-based drug contrast relationships.
• Combined embedding features and ECR using a ranking learning model to prioritize candidate drugs.

Main Results:

• DRONet demonstrated significantly higher prediction accuracy compared to state-of-the-art methods.
• Achieved an 87.4% improvement in Hit@1 and a 37.9% improvement in mean reciprocal rank.
• Case analysis confirmed the high reliability of DRONet's predicted results.

Conclusions:

• DRONet offers a robust framework for drug repositioning by effectively integrating network embedding and ECR.
• The method shows strong potential to guide clinical drug development by identifying reliable new drug indications.