(Data) Reproducibility of soil microbiome data (2011)
Published in early 2011, this work compared 24 separate soil microbial communities under four treatment conditions via multiplexed/barcoded 454-pyrosequencing of PCR-amplified 16S rRNA gene fragments. The authors found differences in the composition and structure of microbial communities between soil treatments. As expected, the soil microbial communities were highly diverse, with a staggering 16,825 different OTUs (species) observed in the included dataset. Interestingly, this study used a larger number of replicates than previous studies of this type, for a total of 56 samples, and the putatively low resampling rate of species between replicated sequencing trials (“OTU overlap”) was a major concern by the authors.
This dataset contains an experiment-level (phyloseq-class) object,
which in turn contains the taxa-contingency table and soil-treatment table
as otu_table-class and sample_data-class components, respectively.
This data was
imported from raw files supplied directly by the authors via personal communication
for the purposes of including as an example in the phyloseq-package.
As this data is sensitive to choices in OTU-clustering parameters, attempts to recreate
the otu_table from the raw sequencing data may give slightly different results
than the table provided here.
abstract from research article (quoted):
To determine the reproducibility and quantitation of the amplicon sequencing-based
detection approach for analyzing microbial community structure, a total of 24 microbial
communities from a long-term global change experimental site were examined. Genomic DNA
obtained from each community was used to amplify 16S rRNA genes with two or three
barcode tags as technical replicates in the presence of a small quantity (0.1% wt/wt)
of genomic DNA from Shewanella oneidensis MR-1 as the control. The technical
reproducibility of the amplicon sequencing-based detection approach is quite low,
with an average operational taxonomic unit (OTU) overlap of 17.2%+/-2.3%
between two technical replicates, and 8.2%+/-2.3% among three technical
replicates, which is most likely due to problems associated with random sampling processes.
Such variations in technical replicates could have substantial effects on estimating
beta-diversity but less on alpha-diversity. A high variation was also observed in the
control across different samples (for example, 66.7-fold for the forward primer),
suggesting that the amplicon sequencing-based detection approach could not be quantitative.
In addition, various strategies were examined to improve the comparability of amplicon
sequencing data, such as increasing biological replicates, and removing singleton sequences
and less-representative OTUs across biological replicates. Finally, as expected, various
statistical analyses with preprocessed experimental data revealed clear differences in
the composition and structure of microbial communities between warming and non-warming,
or between clipping and non-clipping. Taken together, these results suggest that amplicon
sequencing-based detection is useful in analyzing microbial community structure even
though it is not reproducible and quantitative. However, great caution should be taken
in experimental design and data interpretation when the amplicon sequencing-based detection
approach is used for quantitative analysis of the beta-diversity of microbial communities.
(end quote)
Jizhong Zhou, et al.
Zhou, J., Wu, L., Deng, Y., Zhi, X., Jiang, Y.-H., Tu, Q., Xie, J., et al.
Reproducibility and quantitation of amplicon sequencing-based detection.
The ISME Journal. (2011) 5(8):1303-1313. doi:10.1038/ismej.2011.11
The article can be accessed online at http://www.nature.com/ismej/journal/v5/n8/full/ismej201111a.html
# Load the data
data(soilrep)
################################################################################
# Alpha diversity (richness) example. Accept null hypothesis:
# No convincing difference in species richness between warmed/unwarmed soils.
################################################################################
# Graphically compare richness between the different treatments.
man.col <- c(WC="red", WU="brown", UC="blue", UU="darkgreen")
plot_richness(soilrep, x="Treatment", color="Treatment", measures=c("Observed", "Chao1", "Shannon"))Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.