Predictive Modeling of Respiratory Virus Evolution: Current Capabilities and Limitations

Modeling of viral evolution trajectories is of primary practical importance when selecting strains to be included in influenza and SARS-CoV-2 vaccines. Regular updating of the vaccine composition has a pronounced effect on their efficacy, which depends, first of all, on the correspondence of the strains selected for inclusion in the vaccine to those that will circulate during the epidemic season. Given the complexity of the choice, which is based on a large amount of experimental data, and the need to make it in advance for early deployment of vaccine production, the development of predictive methods can significantly improve the correspondence of the selected vaccine strains to those circulating in the upcoming epidemic season. The article considers the main approaches to modeling viral evolution, describes their strengths and limitations. The optimal approach is selected for predicting the evolution of influenza viruses circulating in Russia. The results of using this approach with an assessment of its predictive power on retrospective genomic data of the A(H1N1)pdm09 subtype from the EpiFlu GISAID database are presented. A method for assessing the correspondence of the set of the most adapted strains obtained as a result of the predictive model to the strains selected by experts for inclusion in influenza vaccines in the corresponding epidemic seasons is proposed. This approach can be expanded to predict the evolution of other respiratory viruses, including SARS-CoV-2.