Correcting for 16S rRNA gene copy numbers in microbiome surveys remains an unsolved problem.
Louca S, Doebeli M, Parfrey LW. · 2018
Read Correcting for 16S rRNA gene copy numbers in microbiome surveys remains an unsolved problem. by Louca S, Doebeli M, Parfrey LW. (2018) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5828423.txt.
Introduction
Louca and co-authors test whether 16S gene copy counts can be fixed by prediction. Surveys count a marker gene to size up microbe groups. Species carry one to many copies of that gene. Raw counts thus flatter the copy-rich groups.
Tools claim to predict copy numbers from family trees. Checked against 6,800 genomes, the tools often failed. Good guesses held only for close kin of sequenced species. Some tools explained under 10 percent of the variance. Tools also disagreed with each other on real communities. The authors advise against default correction. Readers get a warning label for a common microbiome shortcut.
Catalog record
| CREATOR | Louca S, Doebeli M, Parfrey LW. |
|---|---|
| YEAR | 2018 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Correcting for 16S rRNA gene copy numbers in microbiome surveys remains an unsolved problem By Stilianos Louca, Michael Doebeli, Laura Wegener Parfrey Published in: Microbiome DOI: 10.1186/s40168-018-0420-9 ABSTRACT The 16S ribosomal RNA gene is the most widely used marker gene in microbial ecology. Counts of 16S sequence variants, often in PCR amplicons, are used to estimate proportions of bacterial and archaeal taxa in microbial communities. Because different organisms contain different 16S gene copy numbers (GCNs), sequence variant counts are biased towards clades with greater GCNs. Several tools have recently been developed for predicting GCNs using phylogenetic methods and based on sequenced genomes, in order to correct for these biases. However, the accuracy of those predictions has not been independently assessed. Here, we systematically evaluate the predictability of 16S GCNs across bacterial and archaeal clades, based on ∼ 6,800 public sequenced genomes and using several phylogenetic methods. Further, we assess the accuracy of GCNs predicted by three recently published tools (PICRUSt, CopyRighter, and PAPRICA) over a wide range of taxa and for 635 microbial communities from varied environments. We find that regardless of the phylogenetic method tested, 16S GCNs could only be accurately predicted for a limited fraction of taxa, namely taxa with closely to moderately related representatives (≲15% divergence in the 16S rRNA gene). Consistent with this observation, we find that all considered tools exhibit low predictive accuracy when evaluated against completely sequenced genomes, in some cases explaining less than 10% of the variance. Substantial disagreement was also observed between tools (R2<0.5) for the majority of tested microbial communities. The nearest sequenced taxon index (NSTI) of microbial communities, i.e., the average distance to a sequenced genome, was a strong predictor for the agreement between GCN prediction tools on non-animal-associated samples, but only a moderate predictor for animal-associated samples. We recommend against correcting for 16S GCNs in microbiome surveys by default, unless OTUs are sufficiently closely related to sequenced genomes or unless a need for true OTU proportions warrants the additional noise introduced, so that community profiles remain interpretable and comparable between studies. Electronic supplementary material The online version of this article (10.1186/s40168-018-0420-9) contains supplementary material, which is available to authorized users. INTRODUCTION Amplicon sequencing of the 16S ribosomal RNA (rRNA) gene is widely used for estimating the composition of bacterial and archaeal communities. Global microbial diversity initiatives, including the Human Microbiome Project [1], the Earth Microbiome Project [2], and the Tara Oceans global ocean survey [3], use the 16S rRNA gene to determine which microbes are present by matching 16S rRNA sequence variants to reference databases like SILVA [4] and estimate the proportions of taxa based on relative read counts. Many bacteria and archaea, however, have more than one copy of the 16S gene, which leads to biased cell count estimates when the latter are estimated solely based on 16S rRNA read counts [5]. This has led to efforts to predict the distribution of 16S gene copy numbers (GCNs) across clades based on available sequenced genomes, in order to then correct 16S rRNA read counts to account for variable 16S GCNs in cells [5–8]. These corrections can substantially affect community profiles and diversity patterns, since some clades have over 10 copies of the 16S rRNA gene [5, 7]. It is thus important to carefully evaluate the accuracy [9] of predicted 16S GCNs across the wide range of microbial clades encountered in microbiome surveys. Inaccurate prediction of 16S GCNs can introduce substantial noise to community profiles, which can be worse than the original GCN-related biases, particularly when prediction methods differ between studies. An accurate prediction of 16S GCNs relies heavily on the assumption that 16S GCNs are sufficiently phylogenetically conserved. That is, 16S GCNs must be autocorrelated among related taxa at least across phylogenetic distances typically covered by available sequenced genomes [10]. Kembel et al. [5] found that 16S GCN exhibits a strong phylogenetic signal, as measured by Blomberg’s K statistic [11], and concluded that 16S GCNs may be predictable based on phylogenetic placement with respect to genomes with known 16S GCN. A similar conclusion was reached independently by Angly et al. [7], based on a strong phylogenetic signal as measured by Pagel’s λ [12]. However, neither Blomberg’s K nor Pagel’s λ make any statement about time scales (nor phylogenetic scales) over which traits vary. While 16S GCN variation is relatively rare within species, variation increases with taxonomic distance [13] and this may lead to inaccurate predictions for the many clades which are distant from sequenced genomes. To date, no independent evaluation of existing 16S GCN prediction tools has been published. To resolve these uncertainties, we assessed the phylogenetic autocorrelation of 16S GCNs across bacteria and archaea (prokaryota) in a phylogenetic tree comprising ∼ 570,000 OTUs (99% similarity in 16S rRNA), based on ∼ 6800 quality-checked complete sequenced genomes. The tree was constructed from sequences in SILVA and partly constrained using SILVA’s taxonomic annotations. We predicted 16S GCNs using several common phylogenetic reconstruction methods and examined the accuracy achieved by each method for OTUs in the SILVA-derived tree. We assessed the predictive accuracy as a function of an OTU’s nearest-sequenced-taxon-distance (NSTD), that is, the minimum phylogenetic distance (mean nucleotide substitutions per site) of the OTU to the nearest sequenced genome. We note that the average NSTD for a particular microbial community, weighted by OTU frequencies, is known as its nearest sequenced taxon index (NSTI; [6]). Further, we systematically assessed the predictive accuracy of three recent tools for correcting 16S GCNs in microbiome surveys, PICRUSt [6], CopyRighter [7], and PAPRICA [8], which together have been cited over 1000 times. While PICRUSt and PAPRICA were mainly designed to predict community gene content based on 16S amplicon sequences, they automatically include an intermediate step for predicting and correcting for 16S GCNs. We evaluate the accuracy of these tools using the known GCNs of the aforementioned sequenced genomes, as a function of a genome’s NSTD. To further evaluate these tools under more realistic scenarios, we also compare all tools to each other for OTUs in 635 prokaryotic communities sampled from diverse natural environments, including the ocean, lakes, hot springs, soil, bioreactors, and animal guts. We find that 16S GCNs are moderately phylogenetically conserved and that prediction of 16S GCNs for the large number of clades without sequenced genomes from close relatives will generally be inaccurate. This conclusion is verified by our finding of low predictive accuracies by CopyRighter, PICRUSt, and PAPRICA, both for the sequenced genomes as well as when compared to each other on microbiomes. Using phylogenetically predicted 16S GCNs to correct 16S read counts in microbiome surveys, as previously suggested [5–7], worked well only for a small number of microbiomes. RESULTS AND DISCUSSION How predictable are 16S GCNs from phylogeny? We found that the autocorrelation function of 16S GCNs, that is the correlation between the GCNs of two randomly picked OTUs at a certain phylogenetic distance, decays moderately with increasing phylogenetic distance (Fig. 1a), dropping below 0.5 at a phylogenetic distance of ∼ 15% and to zero at a phylogenetic distance of ∼ 30% (nucleotide substitutions per site in the 16S gene). Hence, predictions of 16S GCNs are expected to be inaccurate for clades with an NSTD greater than about 15% and close