VaxSeer Predicts Flu Strains With AI to Improve Vaccine Selection and Outpace Viral Evolution

VaxSeer Uses Machine Learning to Predict Virus Evolution and Antigenicity

VaxSeer is an AI system developed at MIT that applies machine learning to predict how viruses evolve and how vaccines will perform. Its goal is to improve the accuracy of vaccine strain selection, reducing reliance on guesswork and improving protective outcomes.

The Challenge of Vaccine Selection

Each year, health authorities must decide months in advance which influenza strains to include in the seasonal vaccine. This decision is critical because a mismatch between vaccine strains and circulating viruses can significantly reduce vaccine effectiveness, leading to preventable illness and increased pressure on healthcare systems.

Influenza viruses mutate constantly and unpredictably, making it difficult to anticipate dominant strains. The Covid-19 pandemic highlighted how quickly viral evolution can challenge vaccine strategies, underscoring the need for more precise forecasting tools.

To address this, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Abdul Latif Jameel Clinic for Machine Learning in Health developed VaxSeer. This AI tool aims to predict dominant flu strains and identify the most protective vaccine candidates well in advance.

How VaxSeer Works

VaxSeer operates with two main prediction engines:

• Dominance Predictor: Estimates the likelihood that each viral strain will spread widely.
• Antigenicity Predictor: Assesses how effectively a vaccine strain can neutralize the targeted virus.

These predictions combine into a coverage score that forecasts how well a vaccine will perform against future viral strains. The score ranges from negative infinity to zero, with values closer to zero indicating better antigenic matches.

When tested retrospectively over a decade, VaxSeer outperformed the World Health Organization’s (WHO) vaccine strain selections in most cases. For the A/H3N2 subtype, VaxSeer’s choices surpassed WHO recommendations in nine out of ten seasons. For A/H1N1, it matched or outperformed WHO choices in six out of ten seasons. Notably, VaxSeer identified a relevant strain for the 2016 flu season a full year before the WHO included it in its vaccine.

Future Directions

Currently, VaxSeer focuses on the hemagglutinin (HA) protein, the primary antigen on the influenza virus. Future iterations may incorporate additional viral proteins like neuraminidase (NA) and consider factors such as immune history, manufacturing capabilities, and dosage levels.

Expanding this approach to other viruses will require extensive, high-quality datasets that capture viral evolution alongside immune responses—data which are not always publicly accessible.

By modeling viral evolution and vaccine interactions, AI tools like VaxSeer offer a pathway for health officials to make better-informed and faster decisions, maintaining an advantage in the ongoing challenge between infection and immunity.

Given how quickly viruses evolve, therapeutic development often struggles to keep pace. VaxSeer represents an effort to close this gap, with implications that extend beyond influenza. Predictive modeling could also help anticipate changes in antibiotic-resistant bacteria or drug-resistant cancers, enabling proactive clinical interventions before resistance becomes widespread.