Computational protocol: Bacterial Community Succession in Pine Wood Decomposition

Similar protocols

Protocol publication

[…] The 16S rRNA data analysis was conducted in Mothur following the standard procedure (). Briefly, sequences were filtered removing reads that did not have a perfect match to the degenerated primers or barcodes, as well as read that had ambiguous bases and more than six homopolymers. Denoising was done using PyroNoise. Sequences were aligned against reference alignment Silva (). Potential chimeras were identified using the UCHIME program () and removed. Filtered sequences were binned into operational taxonomic units (OTUs) based on a 97% dissimilarity cutoff from the distance matrix. One representative sequence of each OTU was assigned through a hierarchical taxonomic annotation using RDP classifier (). For OTU-based analysis, samples were rarefied to 1,080 sequences per sample.The ITS region sequences were filtered removing reads that did not have perfect match to the primers or barcodes, or that had ambiguous bases and more than eight homopolymers. Denoising was accomplished using PyroNoise. Minimal accepted sequence length was set to 200 bp. Sequences were assigned to the taxonomy in the UNITE v.6.0 database (). The assignment was based on the k-Nearest Neighbor algorithm using a cutoff of 80%. ITS samples were rarefied to 2,000 sequences per sample. The sequencing data are available under accession number PRJEB10643 at European Nucleotide Archive (ENA). [...] Correlations between bacterial and fungal community richness and diversity in relation to wood density (proxy of wood decay stage) were calculated using linear regression models. The rarefied OTU table (bacterial community composition) and wood characteristics (pH, moisture, density, nitrogen, carbon, C/N, and ergosterol) data were used for multivariate regression tree (MRT) analysis by using the ‘mvpart’ (Multivariate partitioning) package () in R (statistical programming environment), and the distance matrix was based on Bray–Curtis built by the function “gdist.” Based on MRT results, the samples were classified into three categories: early, middle and late stages.Significant differences in relative abundance (average value of three replicates) of specific bacterial genera (Supplementary Table ) and orders (Supplementary Table ) between categories (early versus middle, middle versus late and early versus late) were determined using Metastats (). Permutations (1000 bootstrap) were applied for estimating the null distribution of the t statistic. In case of the comparison on the genus taxonomical level, false discovery rate (FDR) control for correcting for multiple comparisons was performed in Metastats. Differences between decay stages were analyzed by one-way ANOVA. Post hoc Spjøtfoll–Stoline tests (HSD Tukey for unequal replication numbers) was performed to determine significant differences.For network analysis OTUs were classified at the genus level (both bacterial and fungal). Only genera with five or more representatives across samples were included in the analysis. The correlation matrices of all pairwise Pearson correlations on bacteria and fungi genus level in different categories (early, middle, and late) and between all samples (average value of three replicates) across three wood samples categories were calculated using R. FDR correcting for multiple comparisons was performed according to Hochberg and Benjamini’s method. Only Genera with the correlation estimate <-0.80 or >0.80 and P-value < 0.05 were included in the network analysis. Microbial networks were visualized using Cytoscape ().To test whether neutral or niche-based mechanisms best explained the assembly of the microbial community, we examined the species rank abundance distribution for each wood sample category defined by MRT analysis. Niche-based theory assumes that the rank abundance distribution would fit the pre-emption, broken stick, log-normal and Zipf-Mandelbrot models (; ). On the other hand, the neutral theory predicts that rank abundance distribution would be consistent with ZSM model (). The species rank abundance of each sample were fit to broken stick, pre-emption, log-normal, Zipf, and Zipf-Mandelbrot rank abundance models using the command “radfit” found in the R package vegan (), and the ZSM model using TeTame (). Firstly, we generated the Akaike Information Criterion (AIC), which is a measure of relative quality of a statistical model, providing a means for model selection. These values were calculated based on the equation AIC = -2log-likehood+2∗npar, where npar represents the number of parameters in the fitted model (). In order to compare the models, we calculated the Akaike weight (Wi), which are the weight of evidence in favor of one model being the actual best model in comparison to other tested models (). The Akaike weight was calculated from the equation Wi = exp(-1/2∗ΔAICi)/[sum for all model of exp(-1/2∗ΔAIC)], where ΔAICi = AICi – min(AIC). […]

Pipeline specifications

Software tools mothur, PyroNoise, UCHIME, RDP Classifier, Metastats, vegan
Applications Metagenomic sequencing analysis, 16S rRNA-seq analysis
Organisms Bacteria, Fungi
Diseases Mycoses