Interactome-Based Machine Learning Predicts Potential Therapeutics for COVID-19.

Area of Science:

• Computational biology
• Virology
• Drug discovery

Background:

• COVID-19, caused by SARS-CoV-2, poses a significant global health challenge.
• The virus's spike protein is crucial for host cell entry and is a target for therapeutic intervention.
• Mutations in the spike protein of SARS-CoV-2 variants necessitate continuous efforts to identify effective treatments.

Purpose of the Study:

• To develop a computational pipeline for predicting human-SARS-CoV-2 protein-protein interactions (PPIs).
• To predict drug-protein interactions for potential COVID-19 therapeutics targeting human proteins.
• To identify novel drug candidates against SARS-CoV-2 and its mutated variants.

Main Methods:

• Collected interacting and non-interacting datasets for human-SARS-CoV-2 and drug-protein interactions from public databases.
• Employed Moran autocorrelation for coding protein sequences and PaDEL descriptors for drug coding.
• Utilized a Random Forest model for predicting human-spike PPIs and drug-protein interactions.

Main Results:

• Achieved 90.53% accuracy for human-spike PPI prediction and 96.15% accuracy for drug-protein interaction prediction.
• Identified 40 unique drugs, including eicosapentaenoic acid and dexamethasone.
• Identified 32 human proteins, such as ACACA and DST, as potential drug targets.

Conclusions:

• The developed pipeline effectively predicts human-SARS-CoV-2 PPIs and drug-protein interactions.
• The identified drugs and targets offer promising avenues for developing new COVID-19 therapies.
• This approach aids in combating the original SARS-CoV-2 strain and its evolving variants.