Speaker
Description
Viral fitness, the ability of a virus to spread in a host population, is measured by its effective reproduction number (Re). Understanding how a virus's genotype influences its fitness can lead to the early identification of highly transmissible variants and aid in predicting viral evolution. In this study, we developed a model that predicts the Re of SARS-CoV-2 variants from their spike protein sequences, by fine-tuning a protein language model. Notably, our model correctly predicted the elevated Re of the XBB variant compared to other variants, based on pre-emergence data of the Omicron XBB variant. Furthermore, the model correctly predicted that the acquisition of the F486P mutation enhances XBB's fitness. Moreover, utilizing the fitness prediction model, we have established a framework to infer the effect of mutations on fitness, considering epistatic effects. This study offers critical insights into the genetic determinants that contribute to the viral fitness escalation during the evolution of SARS-CoV-2.
