In brief
BBERT, an AI tool developed by Israeli scientists, can detect bacterial DNA in tiny fragments from soil with high accuracy and predict coding regions. Its application to nearly 2,000 public datasets uncovered many previously unclassified fragments, promising new insight into the soil microbiome.
Israel’s Hebrew University announced that a new AI model, called BBERT, can recognize bacterial DNA in extremely fragmented soil samples, as reported by Xinhua on October 29, 2026.
BBERT was built to process DNA/RNA sequences as short as 100 base pairs, learning patterns that distinguish bacterial fragments from those of other organisms. In laboratory tests, the model identified bacterial DNA with roughly 90 percent accuracy, pinpointed protein‑making sequences with 88 percent accuracy, and found the correct reading frame in 96 percent of cases.
The team applied BBERT to 1,969 publicly available soil DNA datasets, a large effort that revealed many fragments that previous tools had left unclassified. The AI also predicts coding regions, helping scientists discover genetic material that cannot be matched to existing reference databases.
According to the researchers, the ability to identify bacterial DNA in these tiny fragments could deepen our understanding of the soil microbiome -- the diverse community of microorganisms that live beneath our feet and drive essential ecological processes.
Beyond cataloguing known bacteria, BBERT can uncover previously overlooked genetic information, offering clues about new species and novel functions that may have practical applications in agriculture, medicine and biotechnology.
While the study focused on soil samples, the technology could be adapted to other environments where DNA is highly fragmented, such as deep‑sea sediments or ancient archaeological sites.
Overall, Israeli scientists have produced a powerful tool that turns raw, broken genetic data into actionable insight, opening new doors for research into the hidden life hidden in the soil.
