Extended Data Fig. 8: Phylogenetic trees of select species show divergent associations between subclades and type 2 diabetes within each species.
The annotation bars represent metformin use (metf), study, body mass index (BMI), sex, age, and type 2 diabetes (T2D) status, respectively. The boxplots in the bottom represent the posterior mean of the phylogenetic effect of each phylogenetic tree leaf (metagenome) estimated by the phylogenetic generalized linear mixed models (PGLMMs) in anpan (ANalysis of microbial Phylogenies And geNes, see Methods) with whiskers representing the 95% credible intervals of the posterior means. By applying PGLMMs, we compared two generalized linear mixed models with and without incorporating within-species phylogeny as a random effect (Methods). Both models were adjusted for age, sex, body mass index, metformin use, and study membership as fixed effects. We generated within-species phylogenetic trees by randomly splitting the edges based on the Euclidean similarity matrix derived from clustered sets of protein sequences (UniRef90 gene families) after dimension reduction by principal components analysis.
