Data processing, bioinformatics analysis, and statistical analysis
Data were analyzed through QIIME2 (v 2018.6) pipeline (27) using the various plugins. DADA2 (28) was used for denoising, filtering, merging, and chimera removal from the raw Illumina reads. Sequences having a minimum Phred score of 30 were considered for analysis and hence truncated at 285 bp for the merging of forward and reverse sequences. For phylogenetic classification, the sequences were aligned with MAFFT (29) before the construction of a phylogenetic tree using FastTree2 (30). Taxonomic assignment was performed by training a Naïve-Bayes classifier specific to V3-V4 using the SILVA 132 database (31). Taxonomic classification was collapsed and assigned at the genus level. For functional prediction, closed reference OTUs picking was performed using VSEARCH (32) on 97% similarity clustered OTUs using the Greengenes 13_5 database. Weighted UniFrac distance matrices were used to perform principal coordinate analysis (PCoA) in qiime2. Volatility analysis (33) was done based on the Shannon index for different rhizosphere niche samples along with different fallow periods in qiime2 to observe the rate of change in the microbial diversity. Pearson correlation analysis was done in the Base R package to evaluate the relationships among bacterial diversity of rhizosphere niche samples and fallow periods. Microbial functional prediction, closed reference OTUs picking was performed using VSEARCH (34) on 97% similarity clustered OTUs using the Greengenes 13_5 database. The QIIME2 artifacts were converted to BIOM format and subjected to PICRUSt (Phylogenetic investigation of communities by reconstruction of unobserved states) analysis to predict the functional features of the bacterial (34). OTUs were normalized by copy number, and predicted functional categories were created using the Kyoto Encyclopedia of Genes and Genomes (KEGG) (35). The input was put in the STAMP (Statistical Analysis of Metagenomic Profiles) software package to determine the statistical difference for the functional group in between different fallow periods fields samples (36). P-values were calculated using the two-sided Fischer’s exact test (37), while confidence intervals were calculated using the Newcombe– Wilson method (38), and the correction was made using Bonferroni (39). Comparisons of means were done with General Linear Models in Minitab v. 19.