Fig. 1: Stratification of PEx types. | Nature Communications

Fig. 1: Stratification of PEx types.

From: Microbial community organization designates distinct pulmonary exacerbation types and predicts treatment outcome in cystic fibrosis

Fig. 1

A Covariate bias explaining variance of microbiome data (\(n=880\) samples). Two PERMANOVA models contrasted the covariate effect sizes for ASV count data and non-standard sample descriptors. Partial values served as estimators of effect sizes \({\omega }^{2}\). Subject covariates included subject, age group (<31, 31–37, 38–52 years), clinical state (baseline, exacerbation), sex (female, male), F508del CFTR mutation zygosity (homozygous, heterozygous, n.a.), CFTR mutation (F508del+/+; 3 groups F508del-/+ and one other). B Principal component analysis using non-standard sample descriptors (\({{{{{\rm{explained\; variance}}}}}}=84.3\,\%\), two principle components (PCs) are shown). Model variables included Shannon diversity (Shannon), Chao1 richness (Chao1), relative abundance of the dominant pathogen (pat), ratio of counts of CF pathogens and anaerobes (pat/an), and sample classification by Dirichlet multinomial mixture model (DMM) (n = 789 samples). Sample coloring by subject according to legend. C Sample-wise, hierarchical k-mer clustering and distance tree of ordinated data (PC1-3 indicate principle components). Subject ID and age group, as well as Pearson correlation coefficients of the samples are depicted for additional information. Color code of cohort and age group according to legend. D Identification of optimal k-mer number. The dependency between information gain and increasing cluster number \(k\) is shown. First slope saturation served as a cutoff for the minimal number of clusters. Source data for Fig. 1 are provided in the Source Data file.

Back to article page