Network mirroring for drug repositioning.

Area of Science:

• Computational biology
• Pharmacology
• Machine learning

Background:

• Traditional drug discovery is costly and time-consuming with low success rates.
• Drug repositioning offers a cost-effective alternative by repurposing approved drugs for new indications.
• Current methods for drug repositioning are often slow and expensive due to stepwise screening.

Purpose of the Study:

• To develop an efficient machine learning-based approach for identifying candidate diseases and drugs for repositioning.
• To leverage disease similarity to predict drug efficacy across different diseases.
• To improve the success rate and reduce the cost of drug repositioning.

Main Methods:

• Constructed two disease networks: one for disease-protein associations and another for disease-drug information.
• Utilized Kullback-Leibler divergence to measure dissimilarity between network edge distributions for disease-drug repositioning potential.
• Applied the method to identify top candidate diseases for drug repositioning, focusing on the top 20% ranked diseases.

Main Results:

• The proposed machine learning method achieved an F-measure of 0.75, significantly outperforming greedy searching (F-measure of 0.5).
• Application to dementia demonstrated 75% accuracy in identifying repositioned drugs for a disease with no known treatments.
• The approach effectively identifies high-potential drug repositioning candidates.

Conclusions:

• This research introduces a novel, quantitative method for discovering high-potential drug repositioning candidates using disease networks.
• The findings offer profound insights into the potential for discovering previously unrecognized drug repositioning opportunities.
• The developed machine learning approach enhances efficiency and accuracy in drug repositioning strategies.