Evolving fitness and immune escape: a retrospective analysis of SARS-CoV-2 spike protein (2020-2024) using protein language model.

INTRODUCTION

The COVID-19 pandemic posed global health challenges. Understanding SARS-CoV-2's evolutionary dynamics, especially fitness and immune escape, is vital for public health. This study uses protein language models to assess how genetic variations affect viral adaptability and immunity.

METHODS

We applied the CoVFit model to predict Fitness and Immune Escape Index (IEI), validated by a null model based on neutral evolution. We analyzed 2,504,278 SARS-CoV-2 spike sequences, including 160,892 variants, tracking evolution from 2020 to May 2024, comparing real and random mutants' Fitness and IEI.

RESULTS

Our analysis revealed an increase in Fitness (mean rising from 0.227 in 2020 to 0.930 in 2024) and IEI (mean increasing from 0.171 to 0.555) for North American samples. Globally, the comparison of Fitness and IEI between real and random mutants (generated by the null model) revealed statistically significant differences (real mutant Fitness 0.3849 vs. random mutant 0.2046, p < 0.001, KS test; real mutant IEI 0.2894 vs. random mutant 0.1895, p < 0.001, KS test), indicating strong selective pressure; the JN.1 lineage dominated (94% of sequences by April 2024), underscoring its evolutionary advantage.

CONCLUSIONS

CoVFit offers key insights into SARS-CoV-2 evolution, aiding vaccine design. Persistent viral adaptation despite interventions highlights the need for surveillance and adaptive strategies using tools like CoVFit for preparedness.