Benchmarking Oxford Nanopore Long-Read Metagenomics for Microbial Characterization in Laboratory Rodent Health Monitoring

Abstract

Microbiome changes can affect laboratory mouse health, experimental outcomes, and interpretation of disease phenotypes, but routine colony health monitoring may miss complex or unexpected microbial shifts. This is especially important in immunodeficient and genetically engineered mouse colonies, where opportunistic organisms, pathobionts, or altered gut communities may contribute to enteric disease. Although 16S rRNA gene sequencing is useful for describing broad bacterial community structure, it has limited ability to resolve closely related taxa or provide genome-level context. This thesis evaluated whether Oxford Nanopore Technologies (ONT) long-read metagenomic sequencing could reproduce the broad community structure detected by 16S profiling while improving genus-level, species-level, and genome-resolved candidate-organism discovery in mouse fecal microbiome investigations. Defined mock communities were first used to benchmark ONT taxonomic classification under conditions where the expected organisms were known. Mouse fecal samples were then profiled using both 16S rRNA gene sequencing and ONT shotgun metagenomics to compare community structure across taxonomic ranks. Host-depleted ONT reads were assembled into metagenome-assembled genomes (MAGs), which were evaluated using genome quality, taxonomic placement, abundance, reference similarity, virulence-factor annotation, and comparative genome alignment. Mock-community benchmarking showed that classifier choice strongly affected ONT taxonomic results, and that a high classified-read fraction did not necessarily correspond to greater accuracy. Alignment-based approaches generally produced stronger classification performance than k-mer/lowest-common-ancestor (LCA)-based screening approaches, although with greater computational cost. In mouse fecal samples, 16S and ONT showed the strongest agreement at broad taxonomic ranks but diverged at genus and species levels. ONT detected greater fine-scale richness, while 16S provided a stable overview of dominant community structure. MAG recovery produced 174 dereplicated candidate genomes, including several high-quality MAGs that were relatively distant from available mouse gut references, suggesting under-characterized taxa or strain-level diversity. Virulence-factor annotation identified host-interaction and persistence-associated genes across multiple candidate MAGs, but no single VFDB-rich organism was consistently dominant across all samples. These results show that ONT long-read metagenomics can recapitulate the broad community patterns captured by 16S profiling while providing greater fine-scale taxonomic resolution, MAG-based genome evidence, coverage context, and functional annotation. This expanded resolution supports movement from broad community description toward genome-level candidate pathobiont prioritization in laboratory animal microbiome investigations.