Harvard Medical School and Oxford Researchers Develop AI Tool to Predict Virus Evolution

Researchers at Harvard Medical School and the University of Oxford have developed an artificial intelligence tool called EVEscape that can predict how a virus will evolve to become a new variant, according to an HMS press release Wednesday.

In a Nature study published on Oct. 11, researchers demonstrated that EVEscape could have accurately foretold the ways Covid-19 would mutate, identifying variants before they evolved, and also determining which antibody-based therapies would become ineffective as the virus mutated. This capability could also accurately predict mutation trajectories of other viruses such as HIV and influenza.

“We want to know if we can anticipate the variation in viruses and forecast new variants — because if we can, that’s going to be extremely important for designing vaccines and therapies,” said study senior author Debora Marks in the release.

The researchers originally created a tool called “EVE,” which stands for “evolutionary model of variant effect.” EVE used extensive evolutionary data across species to envision the effects of disease-causing genetic mutations to protein function in humans. This tool afforded researchers the ability to determine if gene mutations were benign or malignant in several conditions, like cancers and heart rhythm disorders.

As the Covid-19 pandemic unfolded and the virus evolved, Marks and her team saw the opportunity to repurpose EVE and create a new tool — EVEscape — that would make predictions about the most likely ways that a virus could mutate.

“We underestimate the ability of things to mutate when they’re under pressure and have a large population in which to do so,” Marks said in the release. “Viruses are flexible — it’s almost like they’ve evolved to evolve.”

There are two components to EVEscape. One part is a generative model of evolutionary sequences that lends insight into viral mutations that may occur — the same model used in EVE. Researchers then added detailed biological and structural information about the virus. In combination, the two components can predict viral variants before they occur.

“We’re taking biological information about how the immune system works and layering it on our learnings from the broader evolutionary history of the virus,” said co-lead author Nicole Thadani in the press release.

Marks added that this meant EVEscape had a “flexible framework” that could be generalized to predict variants of other viruses.

The researchers are now using EVEscape to examine the Covid-19 virus in real time and identify future concerning variants, publishing a list of most likely new variants on their website every two weeks. The team is also working to widen EVEscape’s capabilities to include more viruses, such as Lassa and Nipah — two relatively unstudied pathogens that could potentially cause a pandemic, according to the release.

EVEscape could also test vaccines and therapies, seeing how effective they are against current and future viral mutants. This can help scientists design treatments that will remain effective as viruses evolve.

“We want to figure out how we can actually design vaccines and therapies that are future-proof,” Noor A. Youssef, a research fellow in Marks’ laboratory, said in the press release.

—Staff writer Ammy M. Yuan can be reached at ammy.yuan@thecrimson.com.