Revealing SARS-CoV-2 M^pro mutation cold and hot spots: Dynamic residue network analysis meets machine learning

Area of Science:

• Virology
• Computational Biology
• Drug Discovery

Background:

• Predicting viral evolution and mutations is critical for developing effective, long-lasting antiviral drugs.
• Understanding current mutations is feasible, but predicting future evolutionary changes remains a significant challenge.

Purpose of the Study:

• To evaluate the predictive power of dynamic residue network (DRN) analysis for identifying mutation 'cold' and 'hot' spots in the SARS-CoV-2 main protease (Mpro).
• To develop machine learning models integrating DRN metrics with other data to improve mutation prediction for guiding drug development.

Main Methods:

• Conducted molecular dynamics simulations on SARS-CoV-2 Mpro (Wuhan strain).
• Calculated eight dynamic residue network metrics to identify unique network features.
• Compared DRN-identified potential mutation spots with experimental data and mutation frequencies across multiple SARS-CoV-2 lineages.

Main Results:

• Individual DRN metrics showed limited ability to predict residue mutation frequency.
• Integrating eight DRN metrics with structural and sequence data significantly improved mutation prediction accuracy using machine learning models.
• The study identified potential mutation cold and hot spots within the Mpro.

Conclusions:

• A robust computational method was developed to understand pathogen evolution and predict mutation-prone sites.
• This approach can guide the design of long-lasting drugs by targeting less mutable functional residues in viral proteins like Mpro.