Figure 3.

Principal component, Procrustes and coinertia analysis. The first column of plots contains microbiome data (red spots) and the second column contains metabolome data (green spots). The first row contains cecal data and the second row contains sigmoid data. Principal component analysis was performed on the cecum microbiome (A), cecum metabolome (B), sigmoid microbiome (D) and sigmoid metabolome (E). Inter-omic (C and F) Procrustes analysis was then performed. Longer lines on Procrustes plots indicate more within-subject dissimilarity of the microbiome and metabolome. Significance values shown were calculated using the protest function from the vegan R package, which performs repeated symmetric Procrustes analysis to estimate significance. (G) and (H) show operational taxonomic unit (OTU)-level coinertia analysis. Individual OTUs are plotted based on their cointeria-predicted covariance with the metabolome from the cecum (A) and sigmoid (B). To reduce noise in this visualization, the data were thresholded such that only OTUs measured above background in ≥18% of samples are shown. Distance from the center is indicative of the strength of covariance.