Reproducible Research Workflow in R for the Analysis of Personalized Human Microbiome Data

Abstract

This article presents a reproducible research workow for amplicon-based microbiome studies in personalized medicine created using Bioconductor packages and the knitr markdown interface. We show that sometimes a multiplicity of choices and lack of consistent documentation at each stage of the sequential processing pipeline used for the analysis of microbiome data can lead to spurious results. We propose its replacement with reproducible and documented analysis using R packages dada2, knitr, and phyloseq. This workow implements both key stages of amplicon analysis: the initial filtering and denoising steps needed to construct taxonomic feature tables from errorcontaining sequencing reads (dada2), and the exploratory and inferential analysis of those feature tables and associated sample metadata (phyloseq). This workow facilitates reproducible interrogation of the full set of choices required in microbiome studies. We present several examples in which we leverage existing packages for analysis in a way that allows easy sharing and modification by others, and give pointers to articles that depend on this reproducible workow for the study of longitudinal and spatial series analyses of the vaginal microbiome in pregnancy and the oral microbiome in humans with healthy dentition and intra-oral tissues.

The Landscape Ecology and Microbiota of the Human Nose, Mouth, and Throat

“…It is unclear in general whether most surveys of the microbiota characterize single patch types or whether they pool multiple patch types since the spatial extent of CSTs is rarely defined. Further, there is some debate in the field of microbial ecology as to how community state types should be defined (Callahan et al, 2016a). …”

“…Of the studies that have analyzed independent samples of each tooth surface most treat the unit of spatial variation – the physical location of a tooth in the mouth – as a categorical variable such as tooth number, tooth class or tooth aspect (Haffajee et al, 2009; Mager et al, 2003). We have recently shown in a pilot experiment that microbial communities inhabiting the exposed tooth surfaces of healthy humans vary not only based on tooth aspect and tooth class, but as a function of the physical distance separating sites in a manner that is consistent with a spatial gradient (Callahan et al, 2016a). …”

Statistical proof? The problem of irreproducibility

“…For example, modern neuroscience involves inspecting images containing hundreds of thousands of voxels, any of which could be associated with a phenomenon of interest. These lead to multiple hypotheses and variables on which statisticians have to make exponentially many preprocessing choices [19]. This allows the possibility for scientists to choose their outcomes after looking at the data.…”

“…As shown in [19], if one counts the number of possible analyses on the same data-allowing for the choice of up to nine outliers, different transformations of the data, choice from 40 different possible distances, and five different ordination methods-the result is more than 200 million possibilities. No multiple hypotheses correction can protect the user.…”

“…This is because the incentive structure has not been changed; publishing one's data and code is costly in terms of recognition, time, and money. From experience, I know that a complete data workflow paper with a confirmatory data set takes almost a year more to publish [19,23]. This is not a problem where mathematics can help, but mathematicians have set an example and a publication record where transparency is key and theorems are proved: mathematicians show their work and so should statisticians.…”