Jerusalem, 2 November, 2026 (TPS-IL) — Israeli researchers have developed an artificial intelligence tool that can identify bacterial DNA even when it comes from microorganisms missing from existing genetic databases, potentially helping scientists uncover previously hidden forms of microbial life.
The technology could open new avenues for studying soil microorganisms, many of which remain poorly understood. Identifying these bacteria could help scientists better understand soil health, plant growth and the role microbes play in the environment.
Called BBERT, the tool recognizes patterns in bacterial DNA rather than relying on comparisons with existing genetic databases. It can identify fragments as short as 100 DNA building blocks, including sequences that conventional methods often cannot classify.
When scientists analyze soil DNA, they obtain millions of short fragments from bacteria, fungi, plants and other organisms. Identifying their origins can be difficult, particularly when the organisms are absent from genetic reference databases. Much of this unexplained genetic material is known as microbial “dark matter.” BBERT addresses this challenge using an artificial intelligence technique also employed in language technology. Rather than learning patterns in sentences, it learns patterns in bacterial DNA, allowing it to recognize unfamiliar sequences without requiring a match in an existing database.
BBERT was developed by researchers at the Hebrew University of Jerusalem, the University of Haifa and the Volcani Institute, the research arm of Israel’s Ministry of Agriculture. The findings were published in the peer-reviewed journal *Nucleic Acids Research*.
The study was led by Dr. Amir Erez of the Hebrew University’s Racah Institute of Physics and Dr. Omri M. Finkel of its Department of Plant and Environmental Sciences.
Exploring Microbial Dark Matter
In tests, BBERT distinguished bacterial from non-bacterial DNA with approximately 90% accuracy, outperforming other deep-learning tools evaluated. It identified fragments containing instructions for making proteins with approximately 88% accuracy and determined how those instructions should be read with approximately 96% accuracy.
The researchers tested the tool on 1,969 publicly available soil DNA datasets. BBERT identified a substantial proportion of fragments that a widely used comparison tool had been unable to classify.
The analysis also offered new insight into microbial dark matter. Unclassified fragments were more likely to contain non-bacterial DNA than fragments researchers could interpret, suggesting the unexplained material comes from several sources, not just unfamiliar bacteria.
BBERT can also help scientists assemble short DNA fragments into longer sequences to reconstruct bacterial genomes — the complete sets of genetic instructions carried by bacteria. In simulated tests, filtering DNA through the tool reduced average reconstruction time from approximately 66 hours to 25 hours per dataset. The resulting sequences contained fewer errors and gaps while retaining approximately 96% of the genetic coverage achieved without filtering.
Tests involving wild legume root nodules, structures in which plants house bacteria, also showed improvements. BBERT halved reconstruction time and reduced peak computer memory use from 188 gigabytes to 82 gigabytes, while reducing contamination in several reconstructed genomes.
The technology could eventually benefit agriculture and environmental research. Scientists may use it to find microorganisms that help plants absorb nutrients, protect crops against disease or withstand drought. It could also help researchers study how soil microbial communities respond to pollution and climate change. These applications will require further research and testing.
The tool has limitations. It can miss bacterial fragments, mistake other genetic material for bacterial DNA and struggle to distinguish bacteria from archaea, another major group of microorganisms.
BBERT can run on a personal computer with a suitable graphics card, potentially making advanced DNA analysis more accessible to smaller laboratories and allowing researchers to revisit existing environmental DNA collections for information that earlier methods could not identify.
