Influenza virus evolves rapidly to escape immunity elicited by prior infections and vaccinations. To combat this
evolution, the strains in the influenza vaccine are updated every season to keep pace with viral evolution.
However, it is currently difficult to accurately forecast which viral strain will dominate the coming season, and
vaccines are less effective when the wrong strain is chosen for the vaccine. Here we propose a combined
experimental / computational approach that will overcome some of the factors that currently make vaccine strain
selection so difficult. First, we will use a new deep mutational scanning approach to directly map the genotypes
of the viral surface proteins to their antigenic phenotype with respect to human antibody immunity. Second, we
will use this new experimental approach to reconstruct the heterogeneous human immune landscape over which
influenza virus evolves. Finally, we will build computational models that forecast how influenza evolves year-to-
year in response to human immunity. All of our data and forecasts will be integrated into easy-to-use interactive
visualizations. Overall, this work will improve the capacity to accurately identify which viral strains should be
selected for the seasonal influenza vaccine.