Fig. 5
Network analysis of the interactome in frequent exacerbators. Classification of the COPD patients in the EndAECOPD cohort according to exacerbator phenotype. (A-E). Comparison of breathlessness (mMRC) score (A), lung function (FEV1% predicted) (B), sputum total cells (C) and cytokines (D, E) between exacerbator phenotypes. Of the 19 detected cytokines, IL-1β (D) and S100A8 (E) were significantly elevated in frequent exacerbators. Box plots reflect median and IQRs, with whiskers (5–95 percentile) bounding non-outlier values. ns, not significant (Mann–Whitney U-test). IFE = infrequent exacerbator, FE = frequent exacerbator. F. Barplots of microbial abundance displaying similar taxonomic composition between exacerbator phenotypes which was confirmed by LEfSe analysis revealing no differentially abundant taxa. (G). PCoA illustrating similar beta diversity (Weighted UniFrac distance) between exacerbator phenotypes. (H, I). Comparison of Shannon index (H) and bacterial load (I) between exacerbator phenotypes with Mann–Whitney U-test. (J). Opposing network analysis of exacerbator phenotypes. The frequent exacerbators had significantly lower proportion of negative interactions in microbial network compared with infrequent exacerbators (P < 0.0001, χ2 test). # denotes paraphyletic group; • denotes unclassified genus; •• denotes classified but unnamed genus. K. Change of negative interactions (%) for the stability-related taxa (blue, up) and exacerbation-related taxa (purple, down) identified in above network analysis in frequent versus infrequent exacerbators. The dashed line represents ± 5%