2.8. Bacterial Community Data Analysis

The analyses were performed using QIIME2 and R (version 3.5.1). Alpha-diversity of Shannon index and community balance of Pielou’s evenness index was conducted using the QIIME2. A repeated measures two-way Analysis of variance (ANOVA) model with Tukey post-hoc test was used to assess differences between times (Baseline, Week 6 and Week 12) or between treatment groups (CTRL, LD IMD and HD IMD), while between-group differences in shift (log2 transformed fold-change: Week12/Baseline) were determined by Analysis of covariance (ANCOVA) with permutations test (n = 999), adjusted by food consumption. Post hoc test was performed by Tukey′s procedure. To assess beta-diversity, Euclidean distance between samples was calculated from CLR transformed ASVs data using the R package vegan [38]. Principal coordinates analysis (PCoA) was conducted via R package ape [39]. Significant differences of beta-diversity between treatment groups were assessed via permutational multivariate analysis of variance (PERMANOVA, n = 999), adjusted by the food consumption using the Adonis function in vegan and pairwise comparison function in R package EcolUtils [40]. Comparisons of phyla and genera (CLR-transformed counts) between baseline and Week 12 in each treatment group were performed by paired Wilcoxon tests. Comparison between-group differences in CLR-transformed shift (Δ Week 12—Baseline) were determined by Kruskal-Wallis test, then unpaired Wilcoxon test was performed to identify pairwise comparisons. The p-values of Kruskal-Wallis test were adjusted by Benjamini-Hochberg false discovery rate (FDR) method and referred as q-values. Differences with q-values < 0.15 were considered significant.