AI‑driven metagenomic framework promises to culture the planet’s hidden microbes
Researchers at King Abdullah University of Science and Technology (KAUST), in collaboration with an international consortium, have unveiled a predictive framework that marries metagenomic sequencing with artificial‑intelligence models to guide the cultivation of previously uncultivable bacteria and archaea. The work, described on Phys.org on May 11, addresses a long‑standing bottleneck in microbiology: over 99 % of microbial species are known only from DNA fragments.
The framework scans large‑scale environmental DNA databases, extracts genomic signatures indicative of growth requirements, and feeds them into a deep‑learning model trained on successful laboratory cultures. The AI then proposes optimal media compositions, temperature ranges, and co‑culture partners for each target organism.
Lead author Dr. Fatima Al‑Mansour explained: “Our algorithm predicts, with 85 % accuracy, the minimal nutrient cocktail needed to coax a given microbe out of the ‘unculturable’ state.” Co‑author Prof. Daniel Klein added that the method is already yielding isolates of novel methane‑oxidizing archaea from deep‑sea sediment samples.
Microbial ecologists note that unlocking these hidden taxa could revolutionise biotechnology, from new antibiotics to bio‑fuel pathways. Professor Elena Gomez of the Max Planck Institute cautioned that “while predictive, the approach still requires iterative lab validation, and some metabolic dependencies may be too complex for current models.”
The team will test the framework on a worldwide collection of soil samples in a pilot program slated for late 2026, aiming to bring at least 100 novel strains into culture. Success could feed into global efforts such as the Earth Microbiome Project and inform climate‑modeling studies that rely on accurate microbial functional data.