Computational protocol: Regime Shift in Sandy Beach Microbial Communities following Deepwater Horizon Oil Spill Remediation Efforts

Similar protocols

Protocol publication

[…] Following sequencing, all failed reads, low quality score sequences (<20), and non-bacterial rRNA sequences were removed using the Ribosomal Database Project (RDP) pipeline, version 10 (http://rdp.cme.msu.edu/) . Remaining pyrosequences were trimmed of barcodes and adapters to a minimum length of 170 bp prior to alignment with RDP’s Infernal Secondary Structure Aware Aligner . Following alignment in RDP, chimeric reads were identified using uchime in mothur (www.mothur.org/wiki/Main_Page) (version 1.21.0), and potential chimeric sequences were removed from the dataset after visual examination. Additional pyrosequence screening using mothur included removing sequences with homopolymeric regions and ambiguous reads, and to generate uniform start positions.Two complementary approaches were used to obtain pyrosequence taxonomic classifications and to identify changes in the dominant members of the communities from the beaches to the genus-level. Pyrosequences were clustered into operational taxonomic units (OTUs) as the basic unit of diversity for general comparisons , but major changes in community composition were evaluated after assigning taxonomy . This was done because OTU assignments can be challenging to interpret due to differences in clustering algorithms, potential pyrosequence noise and error, undetected chimera, and our ability to circumvent these problems or remove data without compromising the analyses –. A Phylip distance matrix was created from the RDP Pipeline, which was then used by mothur to cluster pyosequences using the furthest neighbor (i.e., complete-linkage clustering) algorithm that considers all of the sequences in an OTU have a set cut-off distance from all the other sequences in that OTU. This clustering algorithm is also used within the RDP pipeline , but it has the potential to overestimate the number of OTUs produced , , particularly when assigned to specific taxonomic groups . Therefore, OTUs were generated using a range of 3%, 4%, and 5% cut-off criteria for adding pyrosequences to a cluster , corresponding to 97%, 96%, and 95% sequencing similarity values, respectively. Non-parametric estimators of OTU richness and diversity were calculated using mothur, including Chao1, Shannon Diversity (H’), and Simpson’s Dominance indices –, for the different OTU clustering cut-offs. These values are included in . A mean value of each index for each sample site was used for comparison. An optimized OTU cut-off of 4% was applied after comparing rarefaction curves () to the number of OTUs, as well as to the Chao1 estimates (). At this cut-off, rarefaction curves for most (>70%) of the samples approached saturated coverage and a ratio of the calculated OTU number to Chao1 value was >0.6 for approx.70% of the 112 samples. Although it is possible that OTU richness was overestimated, Chao1 values also can underestimate richness . Consequently, the combined approach provided confidence in the relative diversity for the dataset. Taxonomic classifications were done both by taking the filtered reads pre-clustering and uploading them directly to RDP Classifier and also by using mothur and the OTU cluster data and a reference database of known 16S rRNA genes, according to an 80% confidence level . After comparing the results, pyrosequences with specific taxonomic assignments were binned from phylum to genus levels, where possible. The normalized relative abundances (i.e., presence/absence) of each taxonomic unit were used for statistical evaluation of potential diversity changes over time. […]

Pipeline specifications

Software tools UCHIME, mothur, PHYLIP, RDP Classifier
Application 16S rRNA-seq analysis