Scientists are using AI to forecast the future of COVID

A new artificial intelligence system called EVEscape can predict how viruses may change as they evolve—potentially predicting the next changes in variants and mutations of COVID-19, as well as changes in other viruses such as influenza and HIV.

The system, developed in 2021 by researchers at Harvard Medical School and the University of Oxford, details Article of October 11 nature. Its creators claim that if it had been available at the start of the pandemic, EVEscape would have been able to predict the most common mutations and most worrisome coronavirus variants.

More than two years after the model was developed, its creators say they have amassed a wealth of data showing that such a system “can make surprisingly accurate predictions,” said the study’s co-lead author, a doctoral student at MIT Sarah Gurev said. ; Nikki Thadani, postdoctoral fellow at Harvard Medical School; Pascal Notin, PhD student at the University of Oxford wealth.

A new digital tool for both offense and defense

Researchers first developed EVE (short for Variation Effect Evolution Model) in 2020 to predict whether genetic mutations in humans are pathogenic or benign. They soon realized that the same technology could be applied to viruses like COVID-19, and it became known as the spin-off EVEscape.

EVEscape is currently being used to assess emerging coronavirus variants, assessing which ones may be the most dangerous. The tool scores each variant reported to the global coronavirus sequence repository. The higher the score, the greater the chance that the variant has evaded immunity from previous infection and vaccination.

This is no small task. Thousands of new coronavirus strains are emerging every month, the trio said. The organization produces a report every two weeks, publishes it online and shares it with agencies such as the World Health Organization to help with pandemic planning.

In its latest reportEVEscape flagged a number of variants as most concerning, including:

• BA.2.86.1 – A sub-variant of the so-called “Pirola” strain that has much more mutations than the normal strain.
• XBB.1.5 – The so-called “Kraken” variant that proliferated around the world earlier this year.
• DV.7.1 – Distant relative of the original Omicron. (Most Omicron derivatives today are descendants of BA.2, so-called “stealth Omicron”.)

Ultimately, this data could help customize vaccines and create therapies such as monoclonal antibodies (administered in hospital settings to high-risk patients) and antiviral drugs like Paxlovid.

Even with tools like EVEscape, it’s impossible to be completely certain which exact variant of the coronavirus will be dominant in six months’ time — a fact that doesn’t bode well for the vaccines that currently exist.

But a system like it “has great potential to be used by vaccine manufacturers to design mutation-proof vaccines” because it can pinpoint target areas of the virus that are less likely to mutate in the future – something that goes hand in hand with the coronavirus’s ever-morphing spike protein different.

EVEscape can also be used to protect against existing viruses with pandemic potential, such as Lassa fever and Nipah virus, which are on the World Health Organization’s list of pathogens with pandemic potential. Both start as non-specific illnesses, unlike COVID-19 or the flu, but can easily progress to more severe symptoms and death. For example, Nipah virus is fatal in 40 to 75 percent of cases. The case fatality rate of COVID-19 is 1% to 4%.

Both viruses have not been fully studied, the three said. EVEscape will allow researchers to predict immune escape “for every mutation in these viruses,” allowing them to develop a “watch list” of changes that could jeopardize the few existing therapies.