The Science Behind Predicting Virus Mutations
Effective vaccines changed the course of the COVID-19 pandemic by preventing illness, reducing severity, and saving millions of lives. Yet, five years on, SARS-CoV-2 continues to circulate and mutate into new variants. These variants require updated vaccines, but designing, manufacturing, and distributing them takes time. This raises a critical question: how can scientists prepare vaccines for viral versions that haven’t emerged yet?
An answer lies in an AI model called EVE-Vax, developed by scientists at Harvard Medical School (HMS), the Massachusetts Consortium on Pathogen Readiness (MassCPR), and other institutions. Published recently in Immunity, this model uses evolutionary, biological, and structural data to predict how a virus might mutate and to design viral surface proteins likely to appear in the future. Applied to SARS-CoV-2, EVE-Vax designed spike proteins that triggered immune responses similar to those caused by actual viral proteins observed during the pandemic.
This research suggests EVE-Vax can offer valuable foresight into viral evolution and aid the creation of "designer" proteins to assess vaccine protection before new variants arise. The goal is to help develop vaccines that stay ahead of mutating viruses.
The Evolution of EVE
More than a decade ago, Debora Marks, professor of systems biology at HMS, and her team explored whether evolutionary genetic data spanning millions of years could predict protein structure and function. Their 2021 AI model, called EVE (evolutionary model of variant effect), leveraged evolutionary data across species to assess protein functionality. When applied to humans, EVE distinguished benign gene variants from those causing disease.
During the COVID-19 pandemic, the team adapted EVE to viral proteins with a new model named EVEscape. This model accurately predicted the most common SARS-CoV-2 mutations and flagged variants likely to increase infections. This success led to the question: could the model forecast future viral evolution? This capability is crucial since vaccines for fast-mutating viruses like SARS-CoV-2 are updated yearly, relying on predictions made months in advance. Mismatches between vaccine design and circulating variants reduce vaccine effectiveness.
To address this, the researchers developed EVE-Vax, focusing on predicting and designing viral proteins to inform vaccine development proactively. The objective was to create novel proteins that are functional and elicit immune responses comparable to real viral proteins.
Predicting SARS-CoV-2’s Future Maneuvers
Using EVE-Vax, the team designed 83 new versions of the SARS-CoV-2 spike protein, each containing up to ten different mutations. Collaborative experiments with colleagues at UMass Chan Medical School, Massachusetts General Hospital, and Beth Israel Deaconess Medical Center tested these “designer” spike proteins. Safe, nonreplicating SARS-CoV-2 variants engineered for the study showed that these proteins could infect human cells and triggered immune responses closely matching those seen during five different stages of the pandemic.
The core insight is that evolutionary data reveals what mutations a virus can adopt and what might occur next. The researchers demonstrated that hundreds of new spike proteins could be engineered efficiently and at low cost, offering a new avenue for vaccine design beyond traditional methods.
For instance, EVE-Vax could have predicted the substantial immune escape observed with the Omicron-targeting COVID-19 booster, indicating that alternative booster designs would have been more effective. Unlike models that predict only viral mutations, EVE-Vax forecasts immune responses, which is more relevant for real-world vaccine development.
Looking Ahead: Expanding to Other Viruses
The team is now extending EVE-Vax to other viruses, including avian influenza, which poses increasing risks globally. A key advantage of EVE-Vax is its ability to work with limited data, making it applicable to understudied or emerging viruses like Lassa and Nipah. Ultimately, EVE-Vax aims to provide crucial information on how viruses might evolve, enabling vaccine designs that protect against future variants before they appear.
For those interested in advancing AI applications in biological research and vaccine design, resources like Complete AI Training’s latest AI courses offer valuable insights into the intersection of AI and life sciences.
Reference
Youssef N, Gurev S, Ghantous F, et al. Computationally designed proteins mimic antibody immune evasion in viral evolution. Immunity. 2025;0(0). doi: 10.1016/j.immuni.2025.04.015.
Your membership also unlocks:
