Extended Data Fig. 7: Comparison of culture data vs microbiome profiles and assessment of the interactome as a predictor of ‘time to next exacerbation’ in longitudinal samples.

a, For each patient (P1-17, n = 17 biologically independent patient samples) three stacked barplots indicate the derived 16S rRNA profiles observed at each timepoint (n = 3) from baseline to exacerbation and post-exacerbation (from left to right). Above each set of barplots are the indicated corresponding results from microbiological culture and the antibiotics used to treat the respective exacerbation (AMX – amoxicillin, AMC – amoxicillin-clavulanic acid, CAZ – ceftazidime, CIP – ciprofloxacin, DOX – doxycycline, MEM – meropenem, TZP – piperacillin-tazobactam). *Asterisks indicate the co-isolation of P. aeruginosa and H. influenzae observed in two patients. A sub-analysis of microbiome diversity in patients suffering a subsequent exacerbation in the 12-week period following the initial occurrence versus those who remained exacerbation free for >12 weeks post-exacerbation is illustrated. b, Shannon Diversity Index at baseline (pre-exacerbation), during initial pulmonary exacerbation and post-exacerbation in those experiencing a second exacerbation in <12 weeks (red, n = 9) and >12 weeks (blue, n = 8). c, Berger-Parker index at baseline (pre-exacerbation), during initial pulmonary exacerbation and post-exacerbation in those experiencing a second exacerbation in <12 weeks (red) and >12 weeks (blue). Box plots reflect median and IQRs with whiskers bounding non-outlier values. ns; not significant (Kruskal-wallis). d, Correlation analysis of microbial abundance and (e) microbial interactions associated with time to next exacerbation. Heatmaps illustrate significant (p < 0.05) associated correlations with colour indicating the strength of correlation (Spearman’s ρ). Vertical text font colouration indicates kingdom membership; blue = bacteria, green = fungi, red = virus. Psuedom. = Pseudomonas, Strep. = Streptococcus, Neiss. = Neisseria, Sacch. = Saccharomyces. A multivariate adaptive regression spline (MARS) was implemented with both microbes and interactions (strength of interaction; edge weight) as predictor variables for ‘time to next exacerbation’ defined as; <12 weeks or >12weeks. Feature importance plots based on generalized cross-validation scores (gcv) illustrating the most important (f) microbes and (g) microbial interactions predicting the time to next exacerbation across baseline, exacerbation and post-exacerbation timepoints. Computed R-squared (RSq) and Generalized R-squared metric (GRsq) are indicated in table (h) reflecting goodness of the fit for each model.