Rhea: a transparent and modular R pipeline for microbial profiling based on 16S rRNA gene amplicons.

Lagkouvardos I, Fischer S, Kumar N, Clavel T. · 2017

Read Rhea: a transparent and modular R pipeline for microbial profiling based on 16S rRNA gene amplicons. by Lagkouvardos I, Fischer S, Kumar N, Clavel T. (2017) 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-pmc5234437.txt.

Introduction

Ilias Lagkouvardos, Sandra Fischer, Neeraj Kumar, and Thomas Clavel presented Rhea, a free pipeline for microbial profiling, in PeerJ in 2017. Sequencing the 16S rRNA gene is now the standard way to survey which microbes live in a sample, and cheap sequencing has drawn in many new users. The analysis steps that follow, though, involve choices that can daunt non-specialists and skew results if made badly.

Rhea is a set of scripts in the R language that walks through the standard downstream steps for unit tables from such sequencing. These include normalization, diversity measures within and between samples, taxonomic breakdowns, statistical comparisons, and correlations. Each choice is documented, and a filter drops units too rare to test soundly. The authors pitch Rhea as an easy entry point for beginners and a frame that experts can extend, designed to track best practice as the field's standards evolve.

Catalog record

CREATORLagkouvardos I, Fischer S, Kumar N, Clavel T.
YEAR2017
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC5234437

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Rhea: a transparent and modular R pipeline for microbial profiling based on 16S rRNA gene amplicons By Ilias Lagkouvardos, Sandra Fischer, Neeraj Kumar, Thomas Clavel, Blanca B. Landa Published in: PeerJ DOI: 10.7717/peerj.2836 ABSTRACT The importance of 16S rRNA gene amplicon profiles for understanding the influence of microbes in a variety of environments coupled with the steep reduction in sequencing costs led to a surge of microbial sequencing projects. The expanding crowd of scientists and clinicians wanting to make use of sequencing datasets can choose among a range of multipurpose software platforms, the use of which can be intimidating for non-expert users. Among available pipeline options for high-throughput 16S rRNA gene analysis, the R programming language and software environment for statistical computing stands out for its power and increased flexibility, and the possibility to adhere to most recent best practices and to adjust to individual project needs. Here we present the Rhea pipeline, a set of R scripts that encode a series of well-documented choices for the downstream analysis of Operational Taxonomic Units (OTUs) tables, including normalization steps, alpha- and beta-diversity analysis, taxonomic composition, statistical comparisons, and calculation of correlations. Rhea is primarily a straightforward starting point for beginners, but can also be a framework for advanced users who can modify and expand the tool. As the community standards evolve, Rhea will adapt to always represent the current state-of-the-art in microbial profiles analysis in the clear and comprehensive way allowed by the R language. Rhea scripts and documentation are freely available at https://lagkouvardos.github.io/Rhea. INTRODUCTION The analysis of complex microbial communities by high-throughput sequencing of 16S rRNA gene amplicons has become very popular. However, this rapid popularization is in stark contrast to the slow transfer of knowledge about data analysis. The hope generated by pioneering next-generation sequencing studies community was high, but recent findings have highlighted the needs for standardization in the field (Clavel, Lagkouvardos & Hiergeist, 2016; Hiergeist, Reischl & Gessner, 2016; Sinha et al., 2015; Walker et al., 2015). With respect to data processing, important parameters such as the choice of clustering methods and thresholds of relative abundances must be considered for proper analysis of high-throughput 16S rRNA gene amplicon datasets (Edgar, 2013; Martínez, Muller & Walter, 2013). The common output of raw sequence processing is a contingency table with read counts assigned to sequence clusters (usually at 97% similarity) termed Operational Taxonomic Units (OTUs). These tables are commonly referred to as OTU-tables and also contain information on the taxonomic classification of sequences. There are several options for the processing of raw sequence data down to OTUs-tables: the most popular software (or package thereof) are QIIME (Caporaso et al., 2010), mothur (Schloss et al., 2009) and UPARSE method (Edgar, 2013) offered by the USEARCH package (Edgar, 2010). The UPARSE method has also been integrated into the pipelines LotuS (Hildebrand et al., 2014) and IMNGS (Lagkouvardos et al., 2016) that can also be used for initial processing of raw sequencing data. In particular, IMNGS is a web-based platform (http://www.imngs.org) with no installation requirements. Downstream analysis of OTU-tables is carried out for the calculation and visualization of diversity and composition of complex microbial communities of interest. Further statistical analysis can allow the identification of OTUs or taxonomic groups that are differentially abundant between sample groups, or can highlight correlations between microbial taxa and metadata characterizing the environment of interest. OTU-tables and associated sequences can currently be analysed by specialized software suites like QIIME (Caporaso et al., 2010) and mothur (Schloss et al., 2009), or by individual pipelines using a general statistical platform like R (R Core Team, 2013). As it is challenging to establish analysis environments that satisfy the needs of a wide diversity of users with respect to ecosystem of interest, experimental design, and expectations, these specialized tools with their respective strengths will continue contributing substantially to standardized analysis of amplicon datasets. Nevertheless, there is also room for further development of modular, transparent, and user-friendly approaches. In contrast to specialized software suites, usage of open statistical environments like R is flexible. The broadness of available tools and the simple syntax can quickly accommodate very complicated needs and allow adjustment to novel requirements, reducing the need for dedicated programs. That is why the R environment has established as a very versatile, powerful, and comprehensive platform for the development of analysis pipelines. The only drawback is the scripting required that can discourage new R users. Pre-packaged functions dedicated to the analysis of microbial data already exist in R, including the phyloseq package (McMurdie & Holmes, 2013), greatly reducing the required personal implementation. Furthermore, the development of Shiny-phyloseq (McMurdie & Holmes, 2015), which adds an abstraction layer over the R console with a browser based graphical control of the analysis, is minimizing the need for knowledge about the R language. Here we took advantage of the easiness and flexibility of R to create a full assembly of analytical steps for diversity and composition analysis of OTU-tables. These scripts were optimized for the analysis of host-associated microbial profiles, most specifically those from the gut environment. We have already utilized the pipeline in different studies (Lagkouvardos et al., 2015; Schaubeck et al., 2015; Würth et al., 2015), which speaks in favour of its validity and usefulness for the assessment of relevant biological information from 16S rRNA gene amplicon data. These scripts do not represent an exhaustive set of all possibilities that exist to analyse sequence datasets, but rather reflect our personal choices among available methods together with novel ideas integrated in the selected analytical steps. We believe that this pipeline can be of use to many scientists with limited or no programing experience or to those with more experience who wants to build on an existing framework, thanks to a simple and clear step-wise processing with very detailed documentation. We packaged this pipeline under the name of the ancient Greek mythology goddess “Rhea,” as a reference to its ever flowing modular architecture, and make it freely available at https://lagkouvardos.github.io/Rhea. METHODS, RESULTS AND DISCUSSION Overview The Rhea pipeline consists of six main R scripts that perform the tasks of normalization of input tables, calculation of alpha-diversity, beta-diversity, taxonomic relative abundances, serial group comparisons, and correlations. These scripts rely on the R packages ade4, GUniFrac, phangorn, Hmisc, corrplot, plotrix, PerformanceAnalytics, reshape, ggplot2, gridExtra, grid, ggrepel, gtable, and Matrix that by themselves have several dependencies. The installation of the packages is performed automatically within the scripts when run for the first time. For the purpose of demonstrating and illustrating the different features of Rhea, the publicly available sequence data from the study by Müller et al. (2016), with ENA accession PRJEB13041, were analysed with the web platform IMNGS (http://www.imngs.org) and the output OTU-table and files (also available for download through the GitHub repository) were used for analysis. In this template study, the impact of housing conditions and diet on mouse faecal microbiota and gut barrier was investigated. For demonstration of the variability of OTU-specific relative abundances among technical replicates, a small study of 10 amplicon libraries constructed from the same human faecal sample were sequenced by Illumina MiSeq. The raw data of this study were deposited to ENA and are available under accession PRJEB14963. Those data were also

Read the complete text