Every year, global health experts face the critical task of selecting influenza strains for the upcoming seasonal vaccine. This decision must be made months before flu season begins, often feeling like a race against time. When the chosen strains match those in circulation, the vaccine is highly effective. However, if predictions miss the mark, the vaccine’s protection can drop significantly, leading to potentially preventable illnesses and strain on healthcare systems.
This challenge became even more pronounced during the COVID-19 pandemic, as new variants emerged just as vaccines were being deployed. Influenza behaves similarly, constantly mutating and presenting a moving target for vaccine design. To address this uncertainty, scientists at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the MIT Abdul Latif Jameel Clinic for Machine Learning in Health have developed an AI system called VaxSeer. This tool aims to enhance the accuracy of vaccine strain selection, reducing reliance on guesswork.
VaxSeer’s Innovative Approach
VaxSeer employs deep learning models trained on decades of viral sequences and lab test results to predict dominant flu strains and identify the most protective vaccine candidates well in advance. Unlike traditional models that analyze single amino acid mutations independently, VaxSeer uses a large protein language model to understand the combinatorial effects of mutations and dynamic dominance shifts. This makes it better suited for rapidly evolving viruses like influenza.
Wenxian Shi, a PhD student in MIT’s Department of Electrical Engineering and Computer Science and lead author of a new paper on the work, explains,
“VaxSeer adopts a large protein language model to learn the relationship between dominance and the combinatorial effects of mutations.”
An open-access report on the study was published in Nature Medicine.
The Future of Flu Vaccination
VaxSeer features two core prediction engines: one estimates how likely each viral strain is to spread (dominance), and the other assesses how effectively a vaccine will neutralize that strain (antigenicity). Together, they produce a predicted coverage score, a forward-looking measure of how well a given vaccine is likely to perform against future viruses.
In a 10-year retrospective study, VaxSeer’s recommendations were evaluated against those made by the World Health Organization (WHO) for two major flu subtypes: A/H3N2 and A/H1N1. For A/H3N2, VaxSeer’s choices outperformed the WHO’s in nine out of ten seasons. For A/H1N1, it matched or outperformed the WHO in six out of ten seasons. Notably, for the 2016 flu season, VaxSeer identified a strain not chosen by the WHO until the following year.
Understanding VaxSeer’s Predictions
VaxSeer estimates how rapidly a viral strain spreads using a protein language model and determines its dominance by accounting for competition among different strains. These insights are plugged into a mathematical framework based on ordinary differential equations to simulate viral spread over time. For antigenicity, the system estimates how well a given vaccine strain will perform in a lab test called the hemagglutination inhibition assay, which measures how effectively antibodies can inhibit the virus from binding to human red blood cells.
Outpacing Viral Evolution
According to Shi,
“By modeling how viruses evolve and how vaccines interact with them, AI tools like VaxSeer could help health officials make better, faster decisions – and stay one step ahead in the race between infection and immunity.”
Currently, VaxSeer focuses on the flu virus’s HA (hemagglutinin) protein, but future versions could incorporate other proteins and factors such as immune history and manufacturing constraints.
Regina Barzilay, the School of Engineering Distinguished Professor for AI and Health at MIT, highlights the broader implications of this work.
“Given the speed of viral evolution, current therapeutic development often lags behind. VaxSeer is our attempt to catch up,”
she says. The team is also exploring methods to predict viral evolution in low-data settings, which could have implications beyond influenza, such as anticipating antibiotic-resistant bacteria or drug-resistant cancers.
Shi and Barzilay collaborated with MIT CSAIL postdoc Jeremy Wohlwend and recent CSAIL affiliate Menghua Wu on this groundbreaking research, supported by the U.S. Defense Threat Reduction Agency and MIT Jameel Clinic.
As VaxSeer continues to evolve, it promises to revolutionize how health officials approach flu vaccination, potentially transforming strategies for other rapidly evolving pathogens.
