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

Single-cell genomics enables cell-resolved insights into in-situ microbial communities, yet single amplified genomes (SAGs) are often fragmented, taxonomically ambiguous, and prone to contamination, limiting their biological interpretation. Here, we present Scellmate, a scalable, semi-supervised framework that integrates whole-SAG curation, species-level taxonomic anchoring and SAG–mobilome linkage analysis for large-scale SAG datasets. Following benchmarking of its SAG curation and SAG–eMGE linkage modules using in silico and four-strain mock SAG datasets, respectively, Scellmate was applied to 11,976 sewage-derived SAGs. This approach yielded 4,198 SAGs passing Scellmate’s whole-SAG curation, which captured pronounced heterogeneity masked by bulk methods. At the species level, we resolved extensive hidden diversity within abundant taxa such as Aliarcobacter that was unresolved by matched metagenome-assembled genomes; at the strain level, we pinpointed a distinct Agathobacter rectalis cluster carrying specialized carbon-utilization pathways, suggesting niche differentiation. Leveraging these curated SAGs, we identified 6,474 host–mobilome linkages across 2,378 SAGs; these linkages not only showed high concordance with host-association data within the Integrated Microbial Genomes (IMG) system but also significantly expanded species-level host assignments. Furthermore, we detected frequent inter-species sharing patterns of diverse cryptic plasmids within the order Bacteroidales. Together, this work demonstrates that high-throughput single-cell genomics, empowered by the robust framework, resolves fine-scale heterogeneity and host–mobilome dynamics, advancing culture-independent microbiome research. In this study, the authors develop a computational framework, Scellmate, that curates and analyzes large-scale microbial single-cell datasets from complex environments, and show that this helps to reveal microbial diversity and infer microbe–plasmid and microbe–phage associations in sewage.