A comparative benchmarking and evaluation framework for heterogeneous network-based drug repositioning methods.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Computational drug repositioning accelerates drug discovery by identifying new uses for existing drugs, reducing costs and timelines.
• Heterogeneous network-based methods are promising but lack systematic evaluation of performance, scalability, and usability.
• Existing comparative studies are limited and yield conflicting results.

Purpose of the Study:

• To conduct a comprehensive, systematic benchmarking of 28 heterogeneous network-based drug repositioning methods.
• To establish a standardized framework for evaluating method performance, scalability, and usability.
• To provide researchers with reliable data for selecting appropriate drug repositioning tools.

Main Methods:

• Benchmarking 28 heterogeneous network-based drug repositioning methods across 11 diverse datasets.
• Developing a comprehensive evaluation framework assessing performance, scalability, and usability.
• Utilizing matrix completion and factorization techniques for performance analysis.

Main Results:

• HGIMC, ITRPCA, and BNNR demonstrated superior overall performance, often leveraging matrix completion/factorization.
• HINGRL, MLMC, ITRPCA, and HGIMC showed the best predictive performance.
• NMFDR, GROBMC, and SCPMF exhibited the highest scalability.
• HGIMC, DRHGCN, and BNNR were identified as the most user-friendly methods.

Conclusions:

• The study provides a robust evaluation of heterogeneous network-based drug repositioning methods.
• HGIMC, ITRPCA, BNNR, HINGRL, MLMC, NMFDR, GROBMC, SCPMF, and DRHGCN are highlighted for their strengths.
• The developed online tool (HN-DREP) and workflow (HN-DRES) facilitate method selection and future research.