Summary and Setup

This lesson teaches the basics of microbiome data analysis with R/Bioconductor. Using real data from a study on colorectal cancer, you will learn the typical steps of a microbiome analysis: importing the data, exploring and preparing it, comparing diversity between groups, finding the bacteria that differ between groups, and predicting an outcome from the microbiome with machine learning.

The lesson uses the OMA framework, a family of R/Bioconductor packages for microbiome data science. After the lesson, you can continue learning with the online book Orchestrating Microbiome Analysis (OMA).

Learning objectives

After this lesson, learners will be able to:

Learning objective Episode
Describe how sequencing reads become the tables we analyse, and how to import them. From reads to tables
Import microbiome data into a TreeSummarizedExperiment object and access its parts. Importing data
Prepare microbiome data by agglomerating, transforming and filtering taxa. Exploring and preparing data
Calculate alpha diversity and test whether it differs between study groups. Alpha diversity
Interpret beta diversity ordination plots, test differences in community composition between groups, and find groups of similar samples with clustering. Beta diversity
Identify taxa whose abundance or prevalence differs between groups. Differential abundance
Predict an outcome from the microbiome with machine learning and evaluate the prediction. Machine learning
Carry out the whole workflow on a new dataset. Capstone

This lesson is a prerequisite for the follow-up multi-omics lesson, which combines the microbiome with other data types, such as metabolites.

Where does this lesson fit?

A microbiome study goes from collecting samples and sequencing them, through processing the reads into tables, to analysing the tables. This lesson covers the whole analysis of a single abundance table: storing the data in R, processing it and the statistical analysis. Studies that combine several tables, for example the microbiome and metabolites of the same samples, need data integration methods. They are covered in the multi-omics lesson, which builds on this one.

Workflow of a microbiome study: data generation, preprocessing, and then five numbered steps: 1. data object, 2. data processing, 3. statistical analysis, 4. data integration and 5. results. Steps 1 to 3 are highlighted as the part covered by this lesson.
Microbiome data science workflow. The highlighted steps are covered in this lesson. Source: OMA.

The data

We analyse stool samples from 60 people in India: 30 people with colorectal cancer and 30 healthy people (Gupta et al. 2019). The samples were sequenced with shotgun metagenomics. The capstone uses a second, small example dataset that comes with the mia package.

Prerequisite

Prerequisites

  • Basic R programming: data frames, factors, functions and plotting with ggplot2, for example from the Introduction to data analysis with R and Bioconductor lesson.
  • Basic statistics: what a p-value is.
  • A general idea of what a microbiome is. No experience in microbiome analysis is needed.

Software

This lesson uses R and RStudio, together with a few R/Bioconductor packages.

  1. Install the newest version of R from CRAN.
  2. Install RStudio Desktop from Posit.

For more detailed instructions, see the setup page of the Introduction to data analysis with R and Bioconductor lesson.

Packages

The lesson uses R packages from Bioconductor and CRAN. We install them in two steps. Run the commands below in the RStudio console. The installation can take several minutes. If R asks whether to install a package from source, you can answer no.

1. Bioconductor and CRAN packages. Bioconductor packages are installed with the BiocManager package, which also installs packages from CRAN. First install BiocManager itself, then the packages:

R 

install.packages("BiocManager")

BiocManager::install(c(
    "mia",       # microbiome data analysis
    "miaViz",    # microbiome data visualization
    "scater",    # principal component analysis
    "maaslin3",  # differential abundance analysis
    "bluster",   # clustering
    "mikropml",  # machine learning
    "ape",       # phylogenetic trees
    "remotes"    # installing packages from GitHub (step 2)
))

The lesson needs the current Bioconductor release. You can check your version with:

R 

BiocManager::version()

If BiocManager says that your version is out of date, update R to the newest release and run the installation again.

2. A package from GitHub. The daa package, which we use to show the idea of differential abundance analysis, is new and is not yet part of Bioconductor. Packages that are still in development are often shared on GitHub. The remotes package installs them:

R 

remotes::install_github("microbiome/daa")

Finally, check that the main packages load and that daa is installed:

R 

library(mia)
library(miaViz)
library(maaslin3)
packageVersion("daa")

If these commands run without errors, you are ready. Messages about masked objects are normal.

Callout

Problems with the installation?

If a package fails to install, check that your R version is recent enough with R.version.string, and try the installation again. If the problem persists, ask for help before the workshop, so that we can solve it together.

Data

You do not need to download anything in advance. We download the data together in the episode RStudio project and experimental data.