MERS-Mpro Predictor: A Machine Learning-Based Tool for Rapid Screening of Potential MERS-CoV Main Protease Inhibitors

Area of Science:

• Computational chemistry and cheminformatics
• Machine learning in drug discovery
• Virology and infectious diseases

Background:

• Middle East Respiratory Syndrome coronavirus (MERS-CoV) poses a global health threat.
• Lack of approved antiviral treatments necessitates novel therapeutic strategies.
• Targeting the MERS-CoV main protease (Mpro) is a key strategy for antiviral development.

Purpose of the Study:

• To develop and validate a ligand-based machine learning framework for identifying MERS-CoV Mpro inhibitors.
• To leverage molecular representations from SMILES strings for predictive modeling.
• To create a user-friendly tool for virtual screening of potential antiviral compounds.

Main Methods:

• Utilized machine learning algorithms including logistic regression, support vector machines, random forests, and XGBoost.
• Employed SMILES-based molecular representations for model training.
• Implemented rigorous validation including internal/external datasets, data partitioning, Y-scrambling, and applicability domain assessment.
• Developed an interactive web application for virtual screening.

Main Results:

• The Random Forest classifier demonstrated superior generalization and predictive performance on an external dataset.
• Comprehensive validation confirmed the reliability and robustness of the developed models.
• The web application allows for rapid screening of compounds with activity predictions and probability scores.

Conclusions:

• A reproducible ligand-based machine learning approach was successfully established for MERS-CoV Mpro inhibitor identification.
• The developed framework and tool can significantly accelerate early-stage antiviral drug discovery.
• This study provides a valuable resource for identifying potential therapeutic agents against MERS-CoV.