A team of researchers has created a novel machine learning tool that's cracking open one of biology's trickiest puzzles: finding Earth’s rarest microbes. It’s similar to looking for needles in a haystack, except the needles are microscopic and might hold the key to how our ecosystems work.

The machine learning tool, called ulrb—short for Unsupervised Learning Based Definition of Microbial Rare Biosphere—uses artificial intelligence (AI) to spot these elusive microorganisms, which pack a serious punch when it comes to keeping our planet's ecosystems healthy. It's like having a super-smart detective that can pick out the few rare gems from billions of ordinary ones.

This pioneering open-source software—the result of a collaboration between the University of Ottawa, Dalhousie University, the Interdisciplinary Centre for Marine and Environmental Research (CIIMAR), the Institute for Bioengineering and Biosciences of Instituto Superior Técnico, and the University of Porto—addresses long-standing challenges in microbial ecology and opens new doors for ecological research.

“This tool solves a major issue in microbial ecology: how do we define rare microorganisms?” says co-author Paula Branco, Associate Professor at the University of Ottawa’s School of Electrical Engineering and Computer Science. “We’ve created a method that is precise, adaptable and capable of improving biodiversity assessments. Before, we were basically guessing at what counted as 'rare' in the microbial world. Now we have a precise way to figure it out.”

Francisco Pascoal, a PhD candidate at CIIMAR, led the development of the ulrb R package as part of his doctoral research. “Our findings show that ulrb not only identifies rare microorganisms, but also works with non-microbial data, such as tree census datasets,” he says. “This versatility makes it a powerful tool for ecological applications.”

Conducted entirely computationally, the study tested ulrb against various microbiome datasets. The software demonstrated that it has statistical robustness and can support practical applications, such as characterizing coral microbiomes.

Available as open-source software on CRAN and GitHub, ulrb includes tutorials to assist users worldwide. Its impacts extend beyond academia, enhancing biodiversity assessments and aiding evaluations of the effects of climate change on microbial communities.

Initiated in 2022, the project marks a new chapter in how we study the hidden world of microorganisms. To find out more, read the article recently published in the scientific journal Communications Biology under the title “Definition of the microbial rare biosphere through unsupervised machine learning”.

This article was adapted and published with permission from the University of Ottawa.