Scellmate brings contamination-aware curation to thousands of single-cell microbial genomes

Single-cell genomics has long promised microbiologists something that bulk sequencing cannot deliver: a view of microbial communities in which every cell’s genome is read individually, preserving the fine-grained variation that averaging destroys. In practice, however, that promise has been constrained by stubborn technical problems. Single amplified genomes, or SAGs, produced by amplifying the DNA of one isolated cell, tend to be fragmented, riddled with amplification artifacts, taxonomically ambiguous, and vulnerable to contamination from stray DNA. A team at The University of Hong Kong has now introduced a computational framework designed to tackle these weaknesses at scale, and its application to nearly twelve thousand sewage-derived SAGs reveals layers of microbial diversity and mobile-gene traffic that bulk methods had concealed.

The framework, called Scellmate, is described in a study published in Nature Communications by Yanren Wang, Xiaoqing Xu, Xi Chen, Yulin Zhang, Wenshan Zheng and Tong Zhang. It integrates three functions that researchers previously had to stitch together from separate tools: whole-genome curation of SAGs, species-level taxonomic anchoring, and linkage analysis that connects SAGs to the mobile genetic elements they carry. The software is semi-supervised, meaning it learns from labeled reference data while still accommodating the messy, incomplete genomes that single-cell amplification produces, and it is built to scale to datasets containing tens of thousands of individual cell genomes.

Before trusting any pipeline on real environmental data, the authors benchmarked its components against controlled tests. The curation module was evaluated using in silico datasets, where the true composition of each simulated genome was known and the software’s ability to detect contamination and discard low-quality regions could be measured directly. The SAG-to-mobilome linkage module was tested on a four-strain mock community, a synthetic mixture of known bacteria in which genuine host associations between cells and mobile elements were established in advance. Performance on these benchmarks gave the team grounds to apply the full workflow to a far more complex and uncontrolled sample: raw sewage.

Sewage is an ideal proving ground for this kind of analysis because it concentrates an enormous diversity of microbes from human guts, skin, soil and water, along with the plasmids, phages and other mobile elements that shuttle genes among them. From the sewage sample, the researchers generated 11,976 SAGs. After Scellmate’s whole-SAG curation, 4,198 genomes passed quality thresholds and were retained for downstream analysis. That curation step matters enormously: without it, fragmented and contaminated genomes can masquerade as novel organisms or corrupt downstream statistics. With it, the curated collection captured pronounced cell-to-cell heterogeneity that bulk metagenomic methods, which average signal across millions of cells, simply could not resolve.

One of the study’s most striking findings emerged at the species level. Within abundant taxa such as Aliarcobacter, a genus of curved bacteria commonly found in wastewater and animal-associated environments, Scellmate resolved extensive hidden diversity that matched metagenome-assembled genomes, or MAGs, had failed to separate. MAGs are reconstructed by binning bulk sequencing data, and while they have transformed microbiome research, they blur distinctions between closely related strains whose DNA co-occurs in the same sample. The single-cell approach sidesteps that problem because each genome comes from one physical cell, allowing the researchers to distinguish subpopulations that bulk binning collapses into a single composite genome.

The strain-level analysis yielded equally telling insights. The team pinpointed a distinct cluster of Agathobacter rectalis, a common gut bacterium, carrying specialized carbon-utilization pathways not shared across the wider population. This kind of metabolic differentiation within a named species is exactly what ecologists mean by niche differentiation: genetically distinct subpopulations exploiting different resources in the same environment. Detecting it required the cell-resolved resolution that SAGs provide, and it suggests that even well-characterized human gut symbionts harbor functional diversity with consequences for how communities process nutrients in the gut and, by extension, in sewage systems that receive gut contents.

The third pillar of the study is arguably its most consequential for microbiology at large: linking mobile genetic elements to their host cells. Plasmids and other mobile elements spread antibiotic resistance genes, virulence factors and metabolic capabilities across microbial communities, yet in bulk sequencing data a plasmid sequence is detached from the organism that carries it. Scellmate’s linkage module reconnects the two, because in a SAG the plasmid DNA and the host chromosome were amplified from the same cell. Applying this to the curated sewage dataset, the researchers identified 6,474 host-mobilome linkages spanning 2,378 SAGs, a dataset-scale map of who carries what in a real environmental community.

Crucially, these computationally inferred linkages were not left unvalidated. The team compared them against host-association data curated within the Integrated Microbial Genomes system, a widely used public database maintained by the Joint Genome Institute, and found high concordance, meaning the single-cell-derived assignments agreed with independently established host associations. Beyond confirming known pairings, the analysis significantly expanded species-level host assignments for mobile elements, attaching plasmids and other elements to host species that had not previously been documented as their carriers. For researchers tracking the movement of antibiotic resistance genes through wastewater, that expansion is directly relevant, since sewage is a recognized conduit for gene exchange between environmental and human-associated bacteria.

The linkage data also exposed social behavior among the mobile elements themselves. The researchers detected frequent inter-species sharing patterns of diverse cryptic plasmids within the order Bacteroidales, a large group of bacteria that dominates the mammalian gut. Cryptic plasmids are those whose functions are not apparent from their sequence, yet their widespread distribution across multiple host species in the same community indicates active horizontal transfer. Mapping which plasmids appear in which hosts, cell by cell, provides the kind of transfer network that epidemiologists and microbial ecologists have long attempted to infer indirectly, and doing so in sewage offers a snapshot of gene-flow hotspots in a densely mixed habitat.

Taken together, the study demonstrates that high-throughput single-cell genomics, when paired with a robust curation and linkage framework, can resolve fine-scale heterogeneity and host-mobilome dynamics that bulk approaches average away. The work arrives at a moment when culture-independent microbiome research is generating data faster than methods can interpret it, and Scellmate addresses a genuine bottleneck: turning noisy, fragmented single-cell genomes into reliable biological observations at population scale. For environmental microbiology, wastewater surveillance and the study of mobile genetic elements, the framework offers a path from thousands of raw SAGs to curated genomes, resolved species boundaries, strain-level functional insight and validated host-mobilome linkages, all within a single reproducible workflow. The Hong Kong team’s results suggest that the cell-resolved view of microbial communities is finally becoming practical at the scale the field has been waiting for.

Subject of Research: A computational framework for contamination-aware curation of single amplified genomes and host-mobilome linkage analysis in microbiome research

Article Title: Scellmate enables contamination aware curation and host mobilome linkage analysis of large SAG datasets

Article References: Wang, Y., Xu, X., Chen, X., Zhang, Y., Zheng, W., & Zhang, T. (2026). Scellmate enables contamination aware curation and host mobilome linkage analysis of large SAG datasets. Nature Communications. https://doi.org/10.1038/s41467-026-78527-4

Image Credits: AI Generated

DOI: 10.1038/s41467-026-78527-4

Keywords: single-cell genomics, single amplified genomes, Scellmate, microbiome, mobile genetic elements, plasmids, wastewater, metagenomics, Bacteroidales, horizontal gene transfer, bioinformatics, microbial ecology