Matrix Factorization-based Technique for Drug Repurposing Predictions.

Area of Science:

• Computational Biology
• Pharmacology
• Drug Discovery

Background:

• Traditional drug design is expensive and inefficient, with high failure rates in clinical trials.
• Repurposing existing drugs computationally offers a cost-effective alternative for identifying new therapeutic indications.

Purpose of the Study:

• To enhance Non-negative Matrix Tri-Factorization (NMTF) for drug repurposing by incorporating protein-protein interaction networks.
• To develop a method for inferring novel drug indications, protein targets, and associated diseases.

Main Methods:

• Implemented a shortest-path evaluation for drug-protein pairs within a protein-to-protein interaction network.
• Integrated this enhancement into the NMTF framework to expand the data for machine learning.
• Applied the enhanced NMTF to an outdated dataset and compared predictions with current data.

Main Results:

• The enhanced method significantly increased the number of putative protein targets from 3,691 to 15,295.
• Achieved a high predictive performance with an Average Precision Score of 0.82.
• Validated top predictions through literature review, confirming drug-target and drug-disease associations.

Conclusions:

• The enhanced NMTF method effectively identifies novel drug targets and indications.
• This approach accelerates drug discovery by leveraging computational integration of diverse biological data.
• The methodology shows promise for drug-centric predictions, identifying therapeutic classes, protein targets, and diseases simultaneously.