News

Influenza, known for its ability to mutate, presents an ongoing challenge to public health. These mutations can affect the antigenic properties of the virus, influencing how well the immune system can recognise and respond to the virus, and also impact the effectiveness of influenza vaccines. The ongoing tracking of these changes and updating vaccines accordingly are a major challenge.

The World Health Organization (WHO) Global Influenza Surveillance and Response System (GISRS) system relies on genetic sequencing and extensive laboratory tests conducted by National Influenza Centres and Influenza Collaborating Centres worldwide to assess circulating influenza viruses. These test results inform decisions on vaccine updates for the upcoming season. However, this process is resource-intensive, time-consuming and requires extensive coordination among laboratories across the globe.

To address these challenges, a team of researchers, led by University of Melbourne Professor Matthew McKay, Lab Head at the Doherty Institute and the Department of Electrical and Electronic Engineering, has developed a machine learning model that can accurately predict antigenic changes in circulating influenza viruses and their capacity to evade immunity from prior infections or vaccinations.

Published in Nature Communications, the model uses genetic sequence data of the viral strain and information from past influenza seasons to predict the outcome of antigenic lab tests that are crucial for determining how a viral strain will react to the current influenza vaccine, whether a vaccine update is required, and if so, what the best vaccine virus would be.

University of Melbourne Dr Ahmed Abdul Quadeer, Senior Research Fellow in the McKay lab and co-leader of the research, highlighted the model’s sophistication in predicting viral strain changes.

“The mapping between genetic and antigenic changes in influenza virus is complex. Our data-driven machine learning model does a very good job in learning this linkage and provides accurate predictions of the virus's antigenic properties, as we found across the 14 seasons we tested the model on,” said Dr Quadeer.

Currently, only a portion of circulating influenza viruses is tested every season, due to practical constraints, such as resources, cost and time. Thanks to its capacity to swiftly analyse vast amount of data, the machine learning model has the potential to predict antigenic properties for all sequenced circulating viruses in that season.

By providing a more complete understanding of the influenza antigenic landscape, the data-driven model can not only provide complementary input to existing surveillance protocols and guide vaccine strain selection, but also offers valuable insights into virus evolution.

The Royal Melbourne Hospital’s Professor Ian Barr, Deputy Director of the WHO Collaborating Centre for Reference and Research on Influenza at the Doherty Institute, underscored the transformative potential of the machine learning model.

“Machine learning and AI tools present scientific advances that have the potential to transform our approach to managing seasonal influenza and potentially other infectious diseases,” said Professor Barr.

Researchers have packaged the model for seasonal influenza antigenic prediction into an easy-to-use web application.

Peer review: Shah S, et al. Seasonal antigenic prediction of influenza A H3N2 using machine learning. Nature Communications (2024). https://doi.org/10.1038/s41467-024-47862-9

Collaboration: This multidisciplinary research, combining genetic sequence data analysis, machine learning, influenza virology and immunology, was an international collaboration with researchers from the Doherty Institute (including the WHO Collaborating Centre for Research and Reference on Influenza, one of the key laboratories globally that conducts antigenic surveillance of seasonal influenza viruses and informs vaccine selection at the WHO), the Hong Kong University of Science and Technology, and The University of Hong Kong.

Funding: Australian Research Council, Hong Kong Research Grants Council.