Fig. 2

Impact of COVID-19 severity on the taxa composition and longitudinal changes of the gut microbiota. A, B Differential abundance analysis and db-RDA analysis in ventilated and non-ventilated COVID-19 patients. A Genera differentially abundant among the two groups were identified by a complex model including ventilation, antibiotics and timepoint with longitudinal sampling correction using negative binomial and zero-inflated mixed model (NBZIMM). The taxa with p-value < 0.01 and effect size (log10) > 2.5 for each group are presented in the bar plot (**p < 0.01, ***p < 0.001). B RDA triplot based on Bray–Curtis distance visualized with type 2 scaling. The correlation between relative abundance of key taxa identified by NBZIMM and clinical variables—ventilation, antibiotics and timepoint—is represented. C, D Beta diversity of ventilated and non-ventilated groups was represented using NMDS plot based on Jaccard distance. Beta diversity at all time points is displayed with ellipses, where the color of each dot indicates the location of the patient at the sampling time point. Samples from patients in antibiotics (ATB) and non-antibiotics (NATB) groups are represented as circles and triangles, respectively. Subsequent samples from the same patients were linked by arrows. Ventilated patients (red arrow) have been ventilated at more than one time point and non-ventilated patients (blue arrow) had no ventilation history during their hospitalization. The four samples of one patient, who had been continuous antibiotic treatment with multiple antibiotics for 2 months before inclusion, clustered distantly from other samples and were hence excluded from the representation. E The mean Jaccard distance, a binary measure of bacterial presence and absence, between time-ordered samples for each patient was compared between ventilated and non-ventilated groups