Data analysis

Data were analyzed using the Qiime 1.8.0 software package (Quantitative Insights Into Microbial Ecology). First, data were filtered to remove forward primers, reverse primers and barcode sequences. Operational Taxonomic Units (OTUs) were selected (97 % similarity) using the pick_closed_reference_otus.py script with the uclust method and Greengenes 13_8 reference. Shannon and Simpson indices were then determined. An OTU table of the samples was generated for downstream analysis with R Studio v0.98.1091. A Venn diagram was created to determine specific and common bacterial species of the different samples. A non-metric multidimensional (MDS) plot using the Bray-Curtis similarity index was used to visualize the differences between samples. Analysis of similarities (ANOSIM) was performed to test whether there were significant differences between different groups. Moreover, similarity percentage analysis (SIMPER) was used to define the taxon at the root of dissimilarity between sample types. Frequency distribution matrices of the sample sequencing results were used for these three types of analysis. Correlations between the evolution of the major bacterial genera found in BFT water and environmental parameters of rearing were tested using the critical value table for Spearman’s rank correlation coefficient rho at the 5 % alpha level.