wiSDOM: a visual and statistical analytics for interrogating microbiome

Abstract

Motivation

We proposed a wiSDOM (web-based inclusionary analysis Suite for Disease-Oriented Metagenomics) R Shiny application which comprises six functional modules: (i) initial visualization of sampling effort and distribution of dominant bacterial taxa among groups or individual samples at different taxonomic levels; (ii) statistical and visual analysis of α diversity; (iii) analysis of similarity (ANOSIM) of β diversity on UniFrac, Bray-Curtis, Horn-Morisita or Jaccard distance and visualizations; (iv) microbial biomarker discovery between two or more groups with various statistical and machine learning approaches; (v) assessment of the clinical validity of selected biomarkers by creating the interactive receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC) for binary classifiers; and lastly (vi) functional prediction of metagenomes with PICRUSt or Tax4Fun.

Results

The performance of wiSDOM has been evaluated in several of our previous studies for exploring microbial biomarkers and their clinical validity as well as assessing the alterations in bacterial diversity and functionality. The wiSDOM can be customized and visualized as per users’ needs and specifications, allowing researchers without programming background to conduct comprehensive data mining and illustration using an intuitive browser-based interface.

Availability and implementation

The browser-based R Shiny interface can be accessible via (https://lun-ching.shinyapps.io/wisdom/) and freely available at (https://github.com/lunching/wiSDOM).

Supplementary information

Supplementary data are available at Bioinformatics online.

1 Introduction

Cumulative evidence from studies using the 16S rRNA amplicon sequencing technology in the past years has revealed that the compositional change in human microbiota is central to the development and progression of many human diseases (Cho and Blaser, 2012). Leading to a paradigm shift in clinical microbiology. Unlike shotgun metagenomics sequencing that is computationally laborious, 16S rRNA-based strategies, with fast turnaround, are cost-effective and applicable to high host DNA contamination (non-fecal samples). In these microbiome studies, raw sequence reads were initially processed and clustered into operational taxonomical units (OTUs) through a variety of well-developed tool suites, such as QIIME (Caporaso et al., 2010) and mothur (Schloss et al., 2009). Such OTU profiles could be highly dispersed and comprising many zeros, and in many analyses, the phylogenetic distance among these OTUs needs to be taken into account, posing enormous challenges in downstream data mining and presentation.

Current disease-oriented microbiome studies mostly aim to explore (i) the alterations in community compositions as measured by α diversity (within-sample diversity) or β diversity (between-sample diversity); (ii) discriminatory or differentially abundant microbes by classifying or comparative analyses; (iii) the clinical validity of potential microbial biomarkers and (iv) the functional changes in dysbiosis. Many statistical and machine learning approaches have been developed and publicly available for these investigations, however, these approaches aforementioned remain relatively sophisticated and require intensive research efforts due to being implemented by different programming languages such as R, Matlab, Python, etc. To address these burdens, we propose and implement a web-based inclusionary analysis Suite for Disease-Oriented Metagenomics, named wiSDOM, with browser-based R Shiny interface. Browser-based wiSDOM is freely available at the R Shiny server ‘shinyapps.io’ and can be downloaded and run locally at github repository (https://github.com/lunching/wiSDOM) for future update and maintenance. A user’s manual is provided in Supplementary Materials and wiSDOM github repository.

2 Program overview

Six functional modules are included to manage computational details in human microbiome research: (i) data summary; (ii) α diversity; (iii) β diversity; (iv) statistical analysis; (v) receiver operating characteristic (ROC) curve and area under the ROC curve (AUC) and (vi) functional prediction of metagenomes (Fig. 1). More information and function of each module are described in Supplementary Material. In addition, all R packages, their corresponding citations, and required input type of 16S rRNA data in each module are summarized in Supplementary Table S1.

Fig. 1.

Workflow and example output of wiSDOM

3 Discussions

Recent culture-free microbiome studies are largely exploratory, requiring intensive research efforts on data mining. Here, we present the web-based inclusionary analysis suite for disease-oriented metagenomics (wiSDOM), a free web platform with six function modules to manage advanced computational details in human microbiome research. wiSDOM is highly flexible for customization and can be broadly used without programming background, and the analyses in each module can be visualized as per users’ needs and specifications. Its performance has been evaluated in several of our previous studies for exploring microbial biomarkers and their clinical validity as well as assessing the alterations in bacterial diversity and functionality (Chiu et al., 2020; Su et al., 2020; Wu et al., 2020a,b,c). Due to limited space, a brief discussion regarding the comparisons of the wiSDOM with numerous freely available computational pipelines is provided in Supplementary Material.

To link dynamic bacterial consortia to human diseases, additional functions are needed to address several limitations of the present version. Although 16S rRNA amplicon sequencing appears to be the most commonly used strategy and has made important strides in clinical microbiology, deep shotgun metagenome sequencing usually offers more in-depth taxonomic characterization and functional insights into the human microbiomes than 16S rRNA-based methods. Currently, wiSDOM does not support taxonomy- and gene-centered analyses of shotgun data. Also unavailable is the integration of DADA2 (Callahan et al., 2016) to process sequencing reads into amplicon sequence variant (ASV) (Callahan et al., 2017) for subsequent data analysis and mining, although there remains a debate that ASVs should replace OTUs in marker gene analyses. Another issue is that wiSDOM does not provide correlation or association analysis as the connection between microbial communities and host factors has drawn more and more attention in microbiome studies (e.g. microbe-metabolite interaction). To fill these gaps, we intend to implement function modules to the wiSDOM in future releases.