New AI tool predicts how microbes survive in Earth’s saltiest habitats

From the Dead Sea to solar salt pans, some of Earth's harshest environments are teeming with microscopic life. Yet more than 81% of the world's microorganisms cannot be grown in the laboratory, leaving scientists unable to study how they survive such extreme saltiness.

Now, researchers at Nanjing Tech University in China have developed SuSha, a computational tool that predicts a microorganism's salt tolerance directly from its genetic blueprint. Reported in the KeAi journal Engineering Microbiology, the multi-model ensemble learning framework offers a new way to infer the lifestyle of microbes that have never been cultured.

Previous approaches relied on single “salt-tolerance” genes, but these often fail to reflect an organism's true behaviour in complex natural habitats. Instead, the SuSha team looked at the whole genome. By analyzing the amino acid composition of 123 bacterial and archaeal species with well-defined salinity preferences, they built a 24-dimensional “fingerprint” capturing both individual amino acids and their broader chemical properties.

Subsequently, by combining several machine learning algorithms — random forest, bagging, and extra trees — into a single ensemble model, SuSha achieved a global accuracy of 0.765 and an AUC of 0.941, outperforming any individual model.

The team then tested SuSha on 2,678 genomes reconstructed from metagenomes across six regions worldwide, spanning freshwater, marine, and hypersaline habitats. In crystallizer ponds with salinity above 35%, the tool correctly identified over 90% of the microbes as extreme halophiles. In the freshwater sediments of Lake Baikal, it accurately flagged salt-sensitive organisms and avoided false alarms.

“Salinity adaptation is a systems-level trait shaped by the entire genome, not by a single gene,” explains Jun Zhou, the study's corresponding author. “By reading the amino acid signature of a whole genome, SuSha lets us infer how a microbe we have never cultured would cope with salt — across the full range from near-freshwater to saturation.”

Notably, salt-loving microbes are a rich source of salt-tolerant enzymes for industry and saline wastewater treatment. “Rather than relying on blind trial-and-error, we can screen enormous genomic datasets and pinpoint the most promising organisms first,” adds Zhou.

To maximize reliability, SuSha switches algorithms depending on the environment — using a robust tree-based ensemble for oceans and freshwater, and a linear ensemble for near-saturated habitats. “Crucially, its predictions are transparent and traceable back to specific genomic features, grounding each result in real evolutionary biology,” Zhou says.

Contact author details:

Jun Zhou, College of Biotechnology and Pharmaceutical Engineering, Nanjing Tech University, Nanjing, 211816, China, zhoujun@njtech.edu.cn

Funder:

This work was supported by the National Natural Science Foundation of China (U24A20543, 32371538), the Jiangsu Agriculture Science and Technology Innovation Fund (JASTIF, CX[23]1038), and the Jiangsu Synergetic Innovation Center for Advanced Bio-Manufacture (XTSW4C01).

Conflict of interest:

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

See the article:

Ren, S., Ren, S., Chen, H., Zhang, W., Zhang, T., Chong, H., Wang, Z., Cao, W., Yong, X., Zhou, J. SuSha: A multi-model ensemble learning framework for predicting microbial salinity adaptation. Engineering Microbiology, Volume 6, 2026, 100292. https://doi.org/10.1016/j.engmic.2026.100292