Minimizing spurious features in 16S rRNA gene amplicon sequencing
Author and article information
Abstract
The 16S rRNA gene amplicon sequencing is a widely used high-throughput method for the taxonomic inference in microbial communities. Many data analysis pipelines have been developed to enhance the accuracy in reflecting the real taxonomy, in order to better guide the downstream identification, isolation and mechanistic studies. Though rigorous quality filtration steps were adopted in these pipelines, with well-designed mock and simulated data sets, we found that there were still a widely divergent number of spurious features due to the “pseudo sequences” artificially generated during the PCR and sequencing process. These pseudo sequences were in low abundances, and were unreliable determined through a weighted re-sampling test. To minimize their influences on the characterization of taxonomy, we proposed an approach that contains two steps, an abundance filtering (AF) step and the subsequent AF-based OTU picking and remapping (AOR) step, which can efficiently decrease the spurious OTUs, sequences or oligotyping features, and improve Matthew's Correlation Coefficient (MCC) values in OTU clustering. The approach can be easily integrated with the popularly-used 16S rRNA sequencing data analysis pipelines, to make the number of OTUs, alpha and beta diversities from divergent pipelines more consistent with the real structure of microbial communities.
Cite this as
2018. Minimizing spurious features in 16S rRNA gene amplicon sequencing. PeerJ Preprints 6:e26872v1 https://doi.org/10.7287/peerj.preprints.26872v1Author comment
This is a preprint submission to PeerJ Preprints.
Sections
Supplemental Information
Figure S1 The OTUs obtained by AOR approach in Mock data
(a-c) The number of OTUs decreased to 22 at thresholds; (d-f) the total ratio of sequences remapped back to OTUs also maintained at >99%; (g-i) the MCC values increased to >0.95, indicating ideal OTU delineation quality. The alternative x axis at the bottom indicates how many sequences did not attending initial OTU delineation at each threshold levels. After OTU delineation, qualified unique sequences were remapped to OTUs with 97% similarity threshold. Dots indicate the original results of corresponding OTU delineation methods.
Figure S2 Coefficient of variation(a-d) and the 99% confidential intervals of bootstrapped abundance (e-h) in (a, e) PWS, (b, f) Ultra, (c, g) River and (d, h) Water data
The Coefficient of variation decreased quickly along with the sequences’ abundances. The distribution of bootstrapped abundance included zero when the abundances were really low. Dashed vertical lines showed the abundance thresholds for OTU delineation.
Figure S3 The OTUs obtained by AOR approach in (a) PWS, (b) Ultra, (c) River and (d) Water data sets
The vertical dashed lines indicates the threshold set by bootstrap resampling. Different pipelines obtained close number of OTUs at these thresholds. Dots indicate the original results of corresponding OTU delineation methods.
Figure S4 The MCC value in (a) PWS, (b) Ultra, (c) River and (d) Water data sets increased along with the threshold
After OTU delineation, all “qualified sequences” were remapped to OTUs with 97% similarity. Dots indicate the original results of corresponding OTU delineation methods.
Figure S5 AOR resulted in less OTUs but comparable alpha diversity in PWS (a-d), Ultra (e-h), River (i-l) and Water (m-p) data
(a, e, i, m) Number of OTUs, (b, f, j, n) Chao1 indices, (c, g, k, o) Simpson indices and (d, h, l, p) Shannon indices per sample were calculated. Multiple comparison was performed using Wilcox test, p values were adjusted by FDR method.
Figure S6 AOR resulted in more consistent beta diversity among methods in (a) PWS, (b) Ultra, (c) River and (d) Water data
Mantel r Statistics were obtained by comparing beta diversity distance matrices between each pair of analysis methods with (Red) original results, (Blue) AOR approach incorporated.
Table S2 The 87 references used in simulated data
Table S3 The average error rates of the raw sequences reported by sequencing machine, QC sequences passing different quality control methods, the final qualified sequences for OTU delineation, and the qualified sequences pre-clustered with up to 1 differe
Table S4 The number of sequences passed quality filtration using different methods
Table S5 The abundance threshold of unreliable sequences
Additional Information
Competing Interests
The authors declare that they have no competing interests.
Author Contributions
Jing Wang conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, prepared figures and/or tables, authored or reviewed drafts of the paper, approved the final draft.
Qianpeng Zhang performed the experiments.
Guojun Wu authored or reviewed drafts of the paper.
Chenhong Zhang performed the experiments.
Menghui Zhang conceived and designed the experiments, analyzed the data, contributed reagents/materials/analysis tools, authored or reviewed drafts of the paper, approved the final draft.
Liping Zhao conceived and designed the experiments, authored or reviewed drafts of the paper, approved the final draft.
DNA Deposition
The following information was supplied regarding the deposition of DNA sequences:
The mock communities and PWS datasets supporting the conclusions of this article are available in the NCBI Short Read Archive repository under BioProject PRJNA306596 (http://www.ncbi.nlm.nih.gov/bioproject/PRJNA306596).
Data Deposition
The following information was supplied regarding data availability:
The 16S rRNA gene clones used in mock data is provided in the supplementary table 1. The 16S rRNA gene reference sequences used in simulated data is provided in the supplementary table 2. The custom R script (resample_uniques_ci.r) used to perform the bootstrapping approach is available in the supplementary.
Funding
This work was supported by grants from the National Science and Technology Major Project of China (2012ZX10005001-009), the National Natural Science Foundation of China (31330005, 30730005, 81401141 and 20875061), and the Science and Technology Commission of Shanghai Municipality (14YF1402200). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.