Posted October 12, 2023
By Ray Blanco
AI Could Have Crushed COVID
Thanks to artificial intelligence we may never be as confounded by a virus as we were with COVID-19.
Seemingly always with another trick up its sleeve, SARS-CoV-2 was so difficult to track and treat because it was always evolving. After we finally seemed to get a handle on it, that’s when we got the Delta variant.
After Delta it was Omicron…
Always changing, always adapting, leaving medical experts always playing catch up.
Debora Marks of the Blavatnik Institute at Harvard Medical School characterizes the ability of viruses to adapt by saying…
“We underestimate the ability of things to mutate when they’re under pressure and have a large population in which to do so. Viruses are flexible — it’s almost like they’ve evolved to evolve.”
But now with generative artificial intelligence, Marks’ team at HMS along with researchers at the University of Oxford have given us a chance to be one step ahead of these rapidly evolving viruses.
These researchers have developed an AI tool called EVEscape that is able to use massive datasets in order to predict the most likely and most dangerous possible mutations of viruses. Being able to identify these new variants will massively improve our ability to develop new treatments, even before the variants present themselves.
EVEscape’s previous version, called simply EVE (Evolutionary-model of Variant Effect), was successfully used in order to identify mutations that could cause disease and which were benign.
As COVID worsened throughout the world, the team developing EVE saw an opportunity to use their technology to predict viral variants.
The generative model behind EVE was used, along with massive amounts of data, to create EVEscape.
Putting It To The Test
Researchers figured the best way to see how EVEscape would have performed at identifying COVID mutations was to simply roll back the clock to January 2020 and do what none of us would ever want to do ourselves…
Live through the COVID years again.
Using only data available at the time, EVEscape was used to run simulations of all the most likely and most dangerous variants that SARS-CoV-2 could possibly take.
EVEscape was able to outperform other experimental models in determining which variants would become the most prevalent. This is especially significant because unlike those lab-based tests, EVEscape identified the mutations using a fraction of the time and resources.
If this wasn’t impressive enough, the team also showed how EVEscape could be used similarly to predict the behavior of and help treat HIV and influenza.
In what is possibly the biggest takeaway of the test, the model was also able to predict that antibody-based therapies would become less effective throughout the pandemic and that the virus would mutate to avoid these treatments.
The questions surrounding COVID treatments, especially the vaccines, became a huge issue that this kind of future-facing technology could directly address.
If we can see how both our immune systems and the target virus would respond to a vaccine, both individually and across a population, it could go a long way towards improving the effectiveness of treatments and building trust in the public.
According to Noor Youssef, a researcher for EVEscape…
“We want to figure out how we can actually design vaccines and therapies that are future-proof.”
With that, we’d like to hear your thoughts. Do you trust this kind of “retro testing”? Would you be more likely to trust a vaccine or other treatment if it used something like EVEscape to test possible responses? Share your thoughts at email@example.com.