Machine learning helps scientists identify environmental preferences of microbes

Researchers have used machine learning to scan the genome of bacteria and find a way to predict their environmental pH settings. Led by experts at the University of Colorado Boulder, this new approach promises to help guide ecosystem restoration efforts, agriculture, and even the development of health-related probiotics.

“We know that bacteria with important ecological functions are abundant in any environment, but their environmental preferences remain unclear.” Evolutionary Biology at CU Boulder. “The idea is to use this technology to understand the basics of their natural history.”

Understanding whether a particular bacterium thrives best in an acidic, neutral, or basic environment is just the first step, says lead author Josep Ramoneda, CIRES Visiting Scientist. . “Using this approach, we can predict how microbes will adapt to almost any environmental change,” he said. For example, sea level rise is bringing more salt water to coastal wetlands. “We can predict how microbes will respond to these environmental changes,” Ramoneda said.

The new work was published in the magazine today, April 28. scientific progressCo-authors include CIRES and CU Boulder, as well as colleagues from Canada.

Microorganisms, including bacteria, are important to ecosystem function. They help plants grow, enable nutrient cycling in lakes, and even support human digestion. We often know very little about them, except for their genetic makeup. Recent decades of genetic “fishing” techniques have led to an exponential growth in the database of bacterial genomes.

So the researchers took what scientists knew about some bacterial groups that thrive at one pH or another, and used machine learning to map those groups’ environmental pH settings to their genetic makeup. Associated. This work involved sorting the genomes of over 250,000 bacterial species from approximately 1,500 soil, lake, and river samples.

“What we found is that we can make inferences about their pH preferences based solely on genomic data,” said Ramoneda. By being able to first guess what pH to use, it could help grow pesky colonies of bacteria that were previously unable to grow. Figuring out how to “cultivate” bacteria so they can be studied in the lab could take years, Fierer said, and machine learning methods could make the process much more efficient. .

Farmers and forestry professionals often add live bacteria to “inoculate” growing plants with beneficial bacteria, Ramoneda said. By ensuring that the inoculum is adapted to the local pH, they are now more rapidly identifying the types of bacteria that may help restore native grasslands and pine forests, or better grow maize and soybeans. may give you a better insight into

Next, the team plans to gain insight into the temperature preferences of bacteria. This is another complex system that likely involves a great many genes. This could, for example, help us better understand how warming affects soil bacterial communities.

“Another way is to try to grow them all in the lab, but that’s a pain,” Fierer said.