Searching for the next virus

The Covid-19 pandemic is not over yet, but some researchers are already worrying about mousepox.

Colin Carlson, a biologist at Georgetown University, has spent the last few years training computers to predict which dangerous viruses could jump from animals to humans, following in the footsteps of the coronavirus (which came from bats), H.I.V. (chimpanzees) and hundreds of other pathogens.
His team used machine learning to develop a short list of potentially dangerous viruses that could eventually make a leap. Mousepox — a virus that infects mice and is similar to smallpox but had not been considered a significant danger to humans — repeatedly came up “super high,” he told my colleague Carl Zimmer.

Digging through the scientific literature, the researchers came across documentation of a mysterious outbreak in 1987 in rural China. Schoolchildren came down with an infection that caused sore throats and inflammation in their hands and feet. When samples from that outbreak were analyzed decades later, scientists found mousepox DNA.
Mousepox is just one of many possible viruses that could cause a new pandemic that computers might be able to suss out beforehand. I asked Carl to explain the complex process experts use to look for potentially dangerous viruses. He said the work started in the field: “It’s not easy. You have to go and catch bats or rodents or tranquilize a lion with darts to take a sample. Not only that, but chances are that in one animal, you wouldn’t find a virus. So you have to catch a bunch.

“Let’s say you’re looking in raccoons. You have to swab them, get feces samples, identify the genetic material. You identify 10 new viruses. Now what? Should we worry about them? Do they pose a threat? What machine learning can do is say, ‘This virus looks a lot like other viruses we’re familiar with.’ You can go through thousands of known viruses. You can make predictions. Then you can test them on a virus you’ve never seen before.”
Could machine learning, in its still-early phase, have foretold Covid’s advent? No, Carl said, because the virus wasn’t known before 2019. But now that we’re sure it originated in bats, machine learning might help us identify types of bats that pose a threat. “Finding those bats should be really high-priority,” he said.

Mammals alone may carry up to 100,000 separate viruses, not even counting those in birds or reptiles. “We’re swimming in an ocean of virus diversity and we barely know about it,” said Carl, author of the book “A Planet of Viruses.” “That’s one reason scientists need to harvest powerful tools like machine learning.”
One thing seems certain: Opportunities for animal-to-human transmission will keep rising because of climate change. As animals seek cooler climes, species will bump against each other. Viruses will leap between them. “A virus that was very distant will become very close,” Carl said.