Sigmoid microbial clusters. Heat maps were created that were sorted based on hierarchical clustering of both bacteria and metabolites (A). The microbial dendrograms were then cut such that six or seven microbial clusters were generated based on k-means prediction strength. The number and horizontal span of each cluster is indicated at the bottom of the dendrogram. The proportion of the phylotypic clades comprising each cluster is depicted for each cluster (B). To determine the differences in imputed metagenomic enrichments/depletions of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways between clusters, the metagenomic composition of each operational taxonomic unit (OTU) was then computed using reconstruction of unobserved states (PICRUSt) and quantitatively analyzed for relative genomic pathway abundance using Unified Metabolic Analysis Network (HUMAnN) and Kruskal-wallis one way analysis of variance. Only pathways that significantly varied in at least one cluster (Bonferroni, P <0.05) are shown (C). Beanplot colors differentiate cluster numbers; *Bonferroni, P <0.05.