AI tool forecasts new COVID variants
Why it matters: The approach could prove more efficient than lab-based testing, because it doesn't rely on people becoming infected or getting vaccinated to develop antibodies.
- This could lead to better and quicker vaccines, including in the next pandemic.
How it works: Researchers developed a generative AI model that's trained on historical viral sequences to predict ways in which the organism could mutate.
- They then added structural details about the virus, like regions most easily targeted by the immune system.
- To test its predictive power, the researchers drew on the trove of data about COVID-19 from the pandemic, and how the stealthy virus kept evolving.
What they found: When presented with ancestral strains of coronavirus from before the pandemic, the tool, called EVEscape, predicted the most frequent mutations and dangerous variants of SARS-CoV-2, the researchers wrote Wednesday in Nature.
- Forecasting mutations could help public health officials develop more effective countermeasures, potentially minimizing the human and economic toll of a pandemic.
- EVEscape is already being used to make predictions about other viruses, including HIV and influenza.
What they're saying: "You can use these generative models to learn amazing things from evolutionary information — the data have hidden secrets that you can reveal," senior author Debora Marks, an associate professor of systems biology at Harvard Medical School, said in a statement.
- "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. "Viruses are flexible — it's almost like they've evolved to evolve."
What we're watching: EVEscape is being used in real time to make predictions about how COVID will evolve next.
- The team ranks new variants on their website and is sharing the information with entities including the World Health Organization.
- They're also testing EVEscape on understudied viruses such as Lassa and Nipah, which have pandemic potential.