Drug-disease networks and drug repurposing

Area of Science:

• Computational biology
• Pharmacology
• Network science

Background:

• Drug repurposing offers a cost-effective approach to developing treatments for new diseases.
• Millions of potential drug-disease combinations exist, necessitating efficient methods for identifying viable candidates.
• In silico predictions are crucial for narrowing down the search space in drug discovery.

Purpose of the Study:

• To develop and analyze a novel network of drugs and their associated diseases.
• To identify potential drug-disease combinations using network-based link prediction.
• To evaluate the efficacy of these prediction methods.

Main Methods:

• Compiled a comprehensive drug-disease network using databases, NLP tools, and manual curation.
• Applied network-based link prediction algorithms, including graph embedding and network model fitting.
• Validated prediction performance using cross-validation tests.

Main Results:

• Achieved high prediction performance, with Area Under the ROC Curve (AUC) exceeding 0.95.
• Demonstrated average precision nearly a thousand times better than random chance.
• Identified several network-based methods, particularly graph embedding, as highly effective.

Conclusions:

• Novel network analysis and link prediction methods significantly improve the identification of drug-disease associations.
• The developed approach offers a powerful tool for accelerating drug repurposing efforts.
• The findings suggest a promising direction for computational drug discovery and development.