Microbiome studies often focus on taxonomic composition, but functional profiling provides deeper insight into microbial biochemistry and host-microbiome interactions. To investigate how gut microbiomes functionally adapt across species with distinct gastrointestinal physiology and dietary needs, we compared the functional potential of human, dog, and cat gut microbiomes.
We profiled over 2,900 gut metagenomes from dogs (n=2,200), cats (n=360), and humans (n=350), using both reference-based mapping (HUMAnN 4) and a novel multi-omic function prediction method (FUGAsseM). FUGAsseM predicts functions for uncharacterized proteins using gene co-expression, genomic proximity, and protein domain interactions. We then applied linear mixed models (MaAsLin 3) to 1) detect host-specific functional enrichments contributed by particular microbial species and 2) assess the extent of novel predicted microbial function in these under-characterized environments.
HUMAnN 4 achieved high mapping rates (75–80%) across all hosts and annotated ~27% of detected gene families with Gene Ontology (GO) terms. Incorporating FUGAsseM increased the annotation rate to over 45%, notably improving protein function characterization. This improvement was pronounced for under-characterized microbial lineages in companion animals dominated by uncharacterized species genome bins. Comparative functional analysis revealed host-specific metabolic profiles: glutamate catabolism and 3-phenylpropionate degradation were enriched in dogs, whereas peptide secretion and chaperone-mediated protein folding were more abundant in humans. In contrast, other processes, such as lactose catabolism and cysteine biosynthesis, showed similar abundances across hosts but were encoded by distinct taxa: Clostridium and Lactobacillus in dogs, Bifidobacterium and Enterococcus in cats, and Bacteroides and Escherichia in humans. These functions appear differentially partitioned among host-specific microbial guilds.
Together, these findings demonstrate the functional consequences of host-microbiome co-evolution at the molecular level and emphasize the improved visibility into microbiome activity provided by de novo putative protein function prediction. Since most microbial environments contain >65% unannotated proteins, this approach offers expanded functional annotations and better characterization of host-specific microbial adaptation, underscoring the value of a One Health perspective.