MIT researchers have developed an AI system named VaxSeer that aims to improve the selection process for seasonal flu vaccines, according to a report published on MIT News. The tool predicts which influenza strains will be most prevalent and identifies potential vaccine candidates well in advance of flu season, using data from viral sequences and lab tests.
Developed by a team from the Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Abdul Latif Jameel Clinic for Machine Learning in Health at MIT, VaxSeer is designed to address an inherent problem in flu vaccine preparation: the unpredictability of the virus's evolution. Typically, health officials have had to make educated guesses months before the actual flu season, a process fraught with uncertainty and risk of mismatches that can significantly decrease vaccine effectiveness. The AI system's predictions have reportedly outperformed those made by the World Health Organization for A/H3N2 in nine out of 10 seasons and matched or outperformed predictions for A/H1N1 in six out of 10 seasons.
As stated by Wenxian Shi, VaxSeer relies on a "large protein language model" to understand the relationship between strain dominance and mutations. This approach differs from traditional evolution models by considering the combined effects of mutations rather than analyzing single amino acid changes in isolation.
Anticipating viruses' next moves is a complex challenge, especially given the pace at which they can evolve. Regina Barzilay, an MIT professor and principal investigator at CSAIL involved with VaxSeer, explained that the tool is MIT's attempt to "catch up" with fast-evolving viruses. The current focus of VaxSeer is on the influenza virus's HA protein, the main antigen in flu viruses. However, the team is also working on methods to predict viral evolution in situations where there is limited data and considering to expand the system to include other factors such as immune history and manufacturing constraints.
The implications of this technology extend beyond just predicting flu strains. Assistant Professor Jon Stokes, from the Department of Biochemistry and Biomedical Sciences at McMaster University, underscored the potential of such predictive modeling to transform our approach to other rapidly adapting organisms, including antibiotic-resistant bacteria and drug-resistant cancers.
VaxSeer's study and development were featured in an open-access report in Nature Medicine. It was supported in part by the U.S. Defense Threat Reduction Agency and the MIT Jameel Clinic. This work is supported by researchers Wenxian Shi, Jeremy Wohlwend, Menghua Wu, and Regina Barzilay, marking a significant milestone in our bid to outmaneuver infectious diseases with cutting-edge technology.
