Skip to main content
Fig. 2 | Respiratory Research

Fig. 2

From: Saliva as a non-invasive specimen for COPD assessment

Fig. 2

Unsupervised clustering analysis of the microbiota of people with COPD. A Dendrogram representing Neighbour joining clustering of Weighted Unifrac (samples rarefied by 4000 sequences, with 5000 iteractions). Numbers close to the internal nodes represent bootstrap support. Two major clusters containing 90% of people with COPD emerged: Cluster I and Cluster II. The bar chart represents microbiota composition of each patient at phylum level (Orange—Firmicutes; Blue—Bacteroidetes; Red—Proteobacteria; Green—Fusobacteria; Pink—Actinobacteria). Orange, blue, red and green circles represent the dominant phylum of each sample. The heatmap shows patient status according to “hospital admissions” and “Gold group”. Shading from white to black is proportional to severity level, white less severe and black most severe. B PCoA analysis using Emperor of Weighted UniFrac distance matrix Clusters I and II have a significantly distinct microbiota composition (PERMANOVA adjusted for PY, p = 0.001). Grey arrows represent the 3 most relevant ASVs for cluster segregation. One ASV of Prevotella melaninogenica (d0b698c7298bf04110a6d2f220879bfb) is the major contributor for segregation of Cluster II, while one ASV of Haemophilus parainfluenzae (e27680d4009f98f30248d823bc17fb8e) and another for Streptococcus sp. (a5189f77a2cfeab3bc1602ff5c8ac3e9) contribute for segregation of Cluster I. C The microbiota of Cluster I is less diverse than microbiota of Cluster 2 (Mann–Whitney U-test, U = 271, p = 0.008). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Back to article page