Extended Data Figure 6: Inter-individual microbial community variation according to diet and sequencing technique.

a, b, To measure the degree to which diet influences inter-individual differences in gut microbial gene expression, we clustered RNA-seq profiles from baseline (a) and diet (b) periods. Dots indicate pairs of samples that cluster by subject. The potential for diet to partition samples was measured by splitting trees at the arrowed branches and testing the significance of the resulting 2 × 2 contingency table (diet versus partition; Fisher’s exact test). To avoid skewed significance values caused by non-independent samples, we only clustered a single sample per subject, per diet period. In the case of multiple baseline samples, the sample closest to the diet intervention was used. In the case of multiple diet samples, the last sample during the diet intervention was kept. A single sample was randomly chosen if there were multiple samples from the same person on the same day. No association between diet and partitioning was found for partitions I–VI (P > 0.05). However, a significant association was observed for partition VII (P = 0.003). c, To determine whether diet affects inter-individual differences in gut microbial community structure, we hierarchically clustered 16S rRNA data from the last day of each diet arm. Samples grouped weakly by diet: sub-trees partitioned at the arrowed node showed a minor enrichment for plant-based diet samples in one sub-tree and animal-based diet samples in the other (P = 0.07; Fisher’s exact test). Still, samples from five subjects grouped by individual, not diet (indicated by black nodes), indicating that diet does not reproducibly overcome inter-individual differences in gut microbial community structure.