Epistatic models predict mutable sites in SARS-CoV-2 proteins and epitopes

Area of Science:

• Virology
• Computational Biology
• Genomics

Background:

• Emerging SARS-CoV-2 variants pose significant threats to public health, impacting transmissibility, pathogenicity, and therapeutic efficacy.
• Predicting future variants is crucial for proactive pandemic response and vaccine development.

Purpose of the Study:

• To develop a predictive model for forecasting unseen SARS-CoV-2 variants by analyzing protein domain mutability.
• To assess the model's accuracy in predicting viral variability and its correlation with experimental data.

Main Methods:

• Statistical modeling using pre-existing coronavirus sequence data to capture amino acid conservation and epistasis.
• Analysis of mutability in SARS-CoV-2 protein domains, focusing on the spike protein's receptor binding domain.
• Correlation of predicted mutability with experimental measures of protein stability and observed viral variability over time.

Main Results:

• The statistical models significantly outperformed conservation profiles in estimating observed SARS-CoV-2 variability.
• Predicted mutability in the spike protein's receptor binding domain correlated well with protein stability (AUC ~0.8).
• The model demonstrated increasing accuracy with more available data, indicating predictive capacity.

Conclusions:

• The developed model accurately predicts SARS-CoV-2 variability and anticipates future variants.
• The approach identifies key mutation sites, especially when combined with immune response data, highlighting overrepresented positions in current variants of concern.
• This predictive framework can aid in understanding viral evolution and proactively addressing future viral outbreaks.