A new AI tool called EVEscape developed by researchers at the Harvard Medical School (HMS) and the University of Oxford can predict new viral variants before they even emerge. Details of their study, published today in Nature, show that the tool has two elements. One models evolutionary sequences to predict changes that can occur in the virus, while the other provides detailed biological and structural details of the virus.
The team who developed EVEscape note in their paper that had it been in use from the beginning of the COVID-19 pandemic, it would have predicted the most frequent mutations occurring within the SAR-CoV-2 virus, as well as the most concerning variants that arose. Similarly, the tool made accurate predictions of other viruses including HIV and influenza.
Now, after showing the capabilities of EVEscape, the investigators are using it to look ahead to predict SARS-CoV-2 variants of concern and releasing rankings of the variants every two weeks.
“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 Debora Marks, associate professor of systems biology in the Blavatnik Institute at HMS and the senior author of the paper.
EVEscape was built by leveraging the earlier development of EVE (evolutionary model of variant effect), which was initially built as a generative model that learns to predict the functionality of proteins based on large-scale evolutionary data across species. This helped provide detailed information about genetic mutations that cause human diseases by discerning benign mutations from those linked with various health conditions such as cancer or cardiac disorders.
As the COVID-19 pandemic unfolded and multiple new variants emerged to continue its spread, Mark and colleagues realized that EVE could be re-tasked to predict viral variants. To do so, they used the generative model in EVE—which can predict mutations in viral proteins that won’t interfere with the virus’s function—and added biological and structural details about the virus, including information about regions most easily targeted by the immune system.
As explained by co-lead author Nicole Thadani, a former research fellow in the Marks lab: “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.”
To test their new model, the team went back to January 2020 and used the data known about SARS-CoV-2 from that time to predict what would happen with the virus over time.
The researchers data show that EVEscape predicted which SARS-CoV-2 mutations would occur during the pandemic with accuracy similar to experimental approaches that test the virus’s ability to bind to antibodies made by the immune system. It outperformed experimental approaches in predicting which of those mutations would be most prevalent. More importantly, EVEscape could make its predictions more quickly and efficiently than lab-based testing since it didn’t need to wait for relevant antibodies to arise in the population and become available for testing. The tool also made predictions about which anti-based therapies would eventually lose their efficacy in fighting the virus as new variants emerged.
The team then showed that the performance of the model was not exclusive to SARS-CoV-2, demonstrating that it could also be used as a predictive tool for HIV and influenza.
The code for EVEscape is freely available to researchers online via GitHub, as are the bi-weekly rankings of the SARS-C0V-2 variants of most concern on their website. The team also shares their rankings with public health entities such as the World Health Organization.
