Discussion

Here, we have used high-throughput sequencing-based neutralization assays to measure the titres of a diverse set of human sera against a large set of HAs representing H3N2 and H1N1 influenza strains circulating as of early 2026. For both of these subtypes, new subclades (K for H3N2 and D.3.1.1 for H1N1) have become dominant in the 6 months since the 2026 Southern Hemisphere vaccine strains were selected. In concordance with other studies (Kirsebom et al. 2025, Cheng et al. 2026, Dee et al. 2026, Ikonen et al. 2026, Liu et al. 2026, Separovic et al. 2026, Wang et al. 2026), our measurements show that subclade K is generally more poorly neutralized by human serum antibodies than other strains, likely explaining their rise. Importantly, we also find that among subclade K H3N2 strains, those with mutations in antigenic regions D and E have especially low titres (Liu et al. 2026), raising the possibility that such strains could increase in frequency over the coming year.

In addition to these major trends, our data also shows additional fine-grain variation in titres to strains within the same subclades, and extensive heterogeneity in titres across sera, some of which partially stratifies with age group. How to best account for this additional heterogeneity in forecasting evolution and choosing vaccine strains remains an open question. By making this large dataset immediately available for analysis, we therefore hope both to inform vaccine-strain selection for the 2026–7 season, as well as spur further studies of how to leverage large and near-real-time neutralization data (Kikawa et al. 2025) to advance public health by improving understanding and forecasting influenza evolution.

Our study has several limitations. First, although we analyse a large set of sera, these sera are taken from individuals from just five different geographic locations and the vast majority of children sera were from Seattle Children’s Hospital. While major seasonal influenza strains generally circulate globally, it is possible that population-level influenza immunity may differ somewhat from region to region (Bedford et al. 2015, Wen et al. 2016). Additionally, most sera were collected before K subclade strains became dominant in late-November 2025, and so, our measurements are more reflective of population influenza immunity at the start rather than the end of the 2025–6 influenza season. Our experiments also only study strain-to-strain variation in neutralization of HA, but cellular immunity and antigenic variation in neuraminidase can also affect influenza evolution (Sandbulte et al. 2011, Machkovech et al. 2015, Rosu et al. 2026). Finally, our study does not analyse influenza B, which antigenically evolves slower than influenza A but still causes appreciable disease and is an important component of the annual influenza vaccine (Rosu et al. 2022, Akin et al. 2025).

Despite these caveats, the subclades of both H3N2 (subclade K) and H1N1 (subclade D.3.1.1) that our experiments measured to have the lowest neutralization titres spread to become dominant during the 2025–6 Northern Hemisphere influenza season. This fact suggests that rapid large-scale measurements of neutralization titres provide substantial information relevant to forecasting viral evolution. Going forward, it will be important to assess if the more fine-grained data our study provided on which strains within subclades K and D.3.1.1 have the lowest neutralization titres prove useful for understanding which descendants of these subclades become dominant over the coming influenza season.