Premium
An improved method to set significance thresholds for β diversity testing in microbial community comparisons
Author(s) -
Gülay Arda,
Smets Barth F.
Publication year - 2015
Publication title -
environmental microbiology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.954
H-Index - 188
eISSN - 1462-2920
pISSN - 1462-2912
DOI - 10.1111/1462-2920.12748
Subject(s) - species evenness , biology , rarefaction (ecology) , diversity (politics) , estimator , pyrosequencing , operational taxonomic unit , set (abstract data type) , community structure , ecology , alpha diversity , microbial population biology , species richness , statistics , computer science , mathematics , 16s ribosomal rna , genetics , sociology , anthropology , gene , bacteria , programming language
Summary Exploring the variation in microbial community diversity between locations ( β diversity) is a central topic in microbial ecology. Currently, there is no consensus on how to set the significance threshold for β diversity. Here, we describe and quantify the technical components of β diversity, including those associated with the process of subsampling. These components exist for any proposed β diversity measurement procedure. Further, we introduce a strategy to set significance thresholds for β diversity of any group of microbial samples using rarefaction, invoking the notion of a meta‐community. The proposed technique was applied to several in silico generated operational taxonomic unit (OTU) libraries and experimental 16 S r RNA pyrosequencing libraries. The latter represented microbial communities from different biological rapid sand filters at a full‐scale waterworks. We observe that β diversity, after subsampling, is inflated by intra‐sample differences; this inflation is avoided in the proposed method. In addition, microbial community evenness ( G ini > 0.08) strongly affects all β diversity estimations due to bias associated with rarefaction. Where published methods to test β significance often fail, the proposed meta‐community‐based estimator is more successful at rejecting insignificant β diversity values. Applying our approach, we reveal the heterogeneous microbial structure of biological rapid sand filters both within and across filters.