Machine Learning-Driven Simulations of the SARS-CoV-2 Fitness Landscape from Deep Mutational Scanning Experiments

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Not specified in abstract

• Primary Outcome: Superior performance in predicting combinatorial mutation effects compared to additive models

• Nnt Or Sensitivity: Strong extrapolative performance ranking Omicron variants when trained on near wild-type variants

• Confidence Interval: Not reported

Clinical Application

High feasibility given computational nature and open-source software availability Highly relevant for Australian Department of Health variant surveillance, CSIRO research, and university-based virology programs. Could inform TGA vaccine strain selection and public health response strategies Public health agencies, virologists, and infectious disease specialists involved in pandemic preparedness and variant surveillance

Abstract

Predicting protein variant effects is a key challenge in preparing for pathogenic viral strains, understanding mutation-linked diseases, and designing new proteins. Protein sequence-structure-function relationships are difficult to model due to complex allosteric and epistatic effects. To investigate efficient modeling strategies, we trained supervised machine learning (ML) models with deep mutational scanning (DMS) libraries of SARS-CoV-2 receptor binding domain (RBD) sequences labeled with angiotensin converting enzyme 2 (ACE2) binding affinity. These models demonstrate superior performance predicting combinatorial mutation effects compared to adding or averaging the effects of point mutations and exhibit strong extrapolative performance ranking omicron variants when training only near wild type (WT) variants. We characterize the RBD fitness landscape by combining ML with Markov Chain Monte Carlo simulations to predict evolutionary patterns from the WT sequence. These generate comparable sequence profiles to high-fitness sequences in DMS data and predict mutations in unseen omicron variants. These models provide insight into the relationship between RBD sequence elements and offer a new perspective on the use of DMS to predict emerging viral strains, which we anticipate will be applicable to other evolutionary prediction tasks. To facilitate application and future development of this strategy, we introduce Mavenets: https://github.com/SztainLab/mavenets.

References

1. 1.Durumeric, A. E. P., McCarty, S., Smith, J., Köhler, J., Elez, K., Raich, L., Suriana, P. A., & Sztain, T. (2026). Machine Learning-Driven Simulations of the SARS-CoV-2 Fitness Landscape from Deep Mutational Scanning Experiments. Journal of Chemical Information and Modeling. https://doi.org/10.1021/acs.jcim.6c00332