The rapid evolution of seasonal in?uenza requires the development of a new in?uenza vaccine by the World Health Organization (WHO) every one to two years. This evolution occurs through a process of antigenic drift where amino acid mutations in the hemagglutinin (HA) surface protein allow currently circulating viruses to evade adaptive immunity against previous vaccine viruses. Therefore, globally successful seasonal in?uenza viruses are often antigenically distinct from previous lineages. High-quality experimental assays for antigenic drift are laborious and low-throughput, leading researchers to develop computational models that can predict the success of in?uenza viruses from HA sequence data alone. Since the publication of these original sequence- only models in 2014, there have been signi?cant advances in in?uenza virology and computational methods that could bene?t in?uenza predictive models. Speci?cally, there are now computational methods to measure antigenic drift by accurately inferring missing measurements in HI assays, high-throughput mutagenesis assays to measure functional constraints on mutations in HA, research supporting the importance of proteins other than HA for in?uenza's ?tness, and detailed analysis of in?uenza's variable geographic circulation. I propose to create a new predictive model of in?uenza evolution that integrates these modern, biologically-informed ?tness metrics into a single framework. These new metrics will build on dense, high-quality HI assays from collaborators at the Centers for Disease Control and Prevention (CDC), deep mutational scanning assays of seasonal in?uenza from collaborators in Dr. Jesse Bloom's lab, a curated database of whole genome sequences for in?uenza, and empirical estimates of in?uenza's global migration rates. This new predictive model will improve the accuracy of predictions about which viruses are most likely to succeed in future in?uenza seasons. These improved predictions will inform recommendations by Dr. Bedford to the WHO at annual vaccine design meetings and, thereby, effect improvements in vaccine ef?cacy and reduce in?uenza-related morbidity and mortality in human populations.
Accurate prediction of seasonal in?uenza evolution depends on a comprehensive integration of biologically rel- evant viral characteristics. This study will integrate in?uenza genome sequences, experimental assays, and patterns of geographic distribution into a novel uni?ed predictive model. The predictions from this model will directly inform vaccine strain recommendations made to the CDC and WHO for all four seasonal in?uenza sub- types.
