An Ensemble Matrix Completion Model for Predicting Potential Drugs Against SARS-CoV-2

Area of Science:

• Computational biology
• Pharmacology
• Infectious disease research

Background:

• The COVID-19 pandemic necessitates rapid identification of effective treatments.
• Existing drug repurposing offers a faster alternative to novel drug development.
• Vaccine efficacy requires ongoing observation, highlighting the need for immediate therapeutic strategies.

Purpose of the Study:

• To develop and evaluate a computational model for predicting virus-drug associations.
• To identify existing drugs that can be repurposed for treating viral infections, including COVID-19.
• To address the challenge of sparse and unbalanced data in virus-drug association prediction.

Main Methods:

• Manual collection of experimentally confirmed virus-drug associations (175 drugs, 95 viruses, 933 associations).
• Development of EMC-Voting, a semi-supervised ensemble model using matrix completion and weighted soft voting.
• Performance evaluation through fivefold cross-validation and comparison with baseline models.

Main Results:

• The EMC-Voting model demonstrated high predictive performance.
• A case study on SARS-CoV-2 showed an outstanding AUPR value of 0.934 for virus-drug association prediction.
• The model effectively handles sparse and unbalanced datasets common in this field.

Conclusions:

• The EMC-Voting model is a promising tool for computational drug repurposing.
• This approach can significantly accelerate the identification of potential treatments for viral diseases.
• The findings support the use of computational methods for rapid clinical decision-making during outbreaks.