Estimating prokaryotic diversity: When are 16S rDNA libraries large enough?

Abstract

As a necessary step in the study of prokaryotic diversity using 16S rDNA libraries, authors should evaluate how well their libraries represent diversity in the source environment. Phylotype-richness estimates can be used to judge whether a library represents diversity sufficiently for its intended purpose. We have argued that richness estimates are most useful if libraries are first shown to be large enough to yield stable estimates. In this article, we (1) evaluate two potentially suitable, non-parametric richness estimators (S ACE and S Chao1 ), tested against model libraries and libraries drawn from natural prokaryotic communities; (2) evaluate whether stable richness estimates are also unbiased; and (3) examine characteristics of prokaryotic libraries that influence the usefulness of richness estimators. Richness estimates consistently reached a stable asymptotic value for libraries that sampled diversity exhaustively. Stable estimates appear to be unbiased or minimally biased estimates of phylotype richness. The S ACE estimator was often undefined, sometimes overestimated phylotype richness at intermediate sampling efforts, and sometimes stabilized at a larger library size than the S Chao1 estimator. The S Chao1 estimator appears well suited for estimating phylotype richness from prokaryotic 16S rDNA libraries. Libraries judged too small to yield a stable richness estimate typically had a highly uneven frequency distribution of phylotypes, with a preponderance of phylotypes that occurred only once in the library. Libraries considered large enough typically had a more even frequency distribution of phylotypes. A software tool is provided to aid others in assessing whether their libraries are large enough to yield stable phylotype-richness estimates. LIMNOLOGY and OCEANOGRAPHY: METHODScedure gave us greater confidence that the richness estimates and the comparisons based on them were valid.The assessment procedure we applied was rudimentary, and critically important questions remain unanswered. Stable phylotype-richness estimates are not necessarily unbiased estimates, and as Hughes et al. (2001) commented "To test for bias, one needs to know the true richness to compare against the sample estimates. As yet, this comparison is impossible for microbes, because no communities have been exhaustively sampled." Furthermore, the relationship of richness estimators to library size has been examined in several recent studies and by subsampling 16S rDNA libraries, and results have been mixed. Hughes et al. (2001) observed that in several large (128 to 284 clones) prokaryotic libraries, richness estimates based on S Chao1 first increased and then usually stabilized with increasing subsample size and were independent of sample size thereafter. However, S Chao1 estimates did not stabilize for bacteria in a high-productivity aquatic mesocosm, and S ACE did not yield stable estimates for any library. Hill et al. (2003) observed that S Chao1 estimates stabilized with one but not with a second large bacterial libra...

Diversity and abundance of diazotrophic microorganisms in the South China Sea during intermonsoon

“…Comparative diversity studies are often hampered by the need to create large clone libraries to characterize true sequence diversity (Kemp and Aller, 2004;Hughes and Hellman, 2005). Once the clone libraries are developed, another large effort is required if separate targets are quantified by realtime PCR.…”

Soil eukaryotic functional diversity, a metatranscriptomic approach

“…The abundance-based richness estimators S Chao1 and S ACE were computed for different subsamples of different sizes drawn from the entire 18S rDNA data set as described by Kemp and Aller (2004) and as implemented at http://www.aslo.org/ lomethods/free/2004/0114a.html.…”

“…Estimations of the abundance-based richness estimators S Chao1 and S ACE were computed by resampling our data set according to Kemp and Aller (2004) (Figure 4, inset). For sample sizes above 60 sequences, S CHAO1 values levelled off to reach a stable asymptotic value of ca.…”

Vertical segregation and phylogenetic characterization of ammonia-oxidizing Archaea in a deep oligotrophic lake

“…Chao diversity indices were calculated using the FastGroupII server http://fastgroup.sdsu.edu/ fg_tools.htm). Coverage of libraries was calculated using the diversity tool (http://www.aslo.org/ lomethods/free/2004/0114a.html) developed by Kemp and Aller (2004) Distance matrices for 16S rRNA and amoA genes were constructed with UniFrac (http://bmf.colorado.edu/unifrac) (Lozupone et al, 2006). To assess the sources of variation in the 16S rRNA and amoA UniFrac matrices, we carried out permutational multivariate analysis of variance based on 1000 permutations (McArdle, 2001) using the function adonis in vegan package (Oksanen et al, 2008).…”