Improving computational drug repositioning through multi-source disease similarity networks

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Computational drug repositioning identifies new uses for existing drugs.
• Network-based methods integrate drug, disease, and target information.
• Single-disease similarity networks limit information diversity.

Purpose of the Study:

• To develop an advanced computational drug repositioning method.
• To integrate multiple disease similarity networks for improved predictions.
• To identify novel drug-disease associations with clinical relevance.

Main Methods:

• Constructed phenotypic, ontological, and molecular disease similarity networks.
• Integrated networks into disease multiplex and multiplex-heterogeneous networks.
• Applied a tailored Random Walk with Restart (RWR) algorithm for prediction.

Main Results:

• Multiplex and multiplex-heterogeneous networks outperformed single-layer networks.
• The MHDR method surpassed state-of-the-art methods in 10-fold cross-validation.
• Identified numerous novel drug-disease associations, including those supported by shared proteins, pathways, and protein complexes.

Conclusions:

• Integrating multiple disease similarity networks enhances drug repositioning accuracy.
• The MHDR approach offers a powerful tool for discovering new therapeutic applications.
• Predicted associations show practical impact, with many validated by clinical trials.