Geometry of antigenic evolution improves influenza vaccine selection

Accurately predicting the antigenic evolution of influenza A/H3N2 viruses is crucial for choosing optimal seasonal vaccine strains. A team of researchers has developed a new method to forecast vaccine effectiveness by mapping the viruses' antigenic relationships. This Bayesian antigenic map, constructed using hemagglutination-inhibition and neutralization titers from 2002 to 2025, reveals twelve distinct antigenic clusters that progress in a stepwise manner. In most seasons, several clusters co-circulate.

Surprisingly, the WHO's recommended vaccine in 15 out of 21 seasons was from an earlier cluster compared to the dominant circulating cluster. By analyzing the direction of each vaccine update relative to recent viral mutations, the researchers found that this prediction could accurately forecast vaccine effectiveness one season ahead. The conventional measure of vaccine-virus match, known as antigenic distance, showed only a weak association with vaccine effectiveness until the direction of updates was considered.

By retrospectively ranking candidate vaccine strains based on their predicted effectiveness, the researchers identified a strain that would have outperformed the WHO recommendation in every season. This hypothetical strain would have led to a 10 percentage points higher mean predicted effectiveness. The findings suggest that incorporating the dynamics of antigenic evolution into vaccine selection could significantly improve vaccine effectiveness.