Statistical Analysis 16S rRNA and WGS
Samples were clustered using supervised canonical correspondence analysis (CCA) (including ‘Variable heights’ H1 to H4 as explanatory variables). The community richness was identified by rarefraction analysis for each sample using Calypso pipeline (Zakrzewski et al 2016). Differences in bacterial alpha diversity (Shannon diversity, Richness, Evenness, Chao & Fischer Alpha) between groups were calculated.
Differentially abundant and significant taxa were evaluated using Wilcoxon rank test (odds ratio) at a p-value cut off of 0.05 followed by false discovery rate (FDR) correction of the p value to determine statistical significance (i.e., p < 0.05/number of tests) for reproducibility. We also generated quantile – quantile plots (QQ plots) of the—log (observed p value) versus the—log (expected p values) within each pair wise comparison for all taxonomic levels and gene categories to ascertain potentially statistically significant associations after correction for multiple comparisons.
Beta diversity was calculated using weighted UniFrac distances, whereas, differences were calculated using Permutational Analysis of Multivariate Dispersions (PERMDISP2) through the betadisper function. Differences in the composition of the fecal microbiota between groups were assessed using the Linear discriminant analysis Effect Size (LEfSe) workflow, by comparing each height (H2, H3 & H4) to baseline (H1).