Orchestrating microbiome multi-omics

Bioconductor conference, 2023

Leo Lahti @antagomir

Orchestrating microbiome multi-omics with R and Bioconductor

Leo Lahti @antagomir | Department of Computing, University of Turku, Finland

Organizing microbiome data

Side information
sample & feature metadata

Multiple assays
counts, relabundance, phILR..

Alternative experiments
16S, metagenome, phylogenetic microarrays; pipelines

Cross-kingdom analyses bacteriome, archaeome, virome, mykobiome, eukaryome

Hierarchical data
taxonomic levels, nested designs, sample & feature trees

Fig. by Domenick Braccia (EuroBioC 2020)

Integrating multiple experiments

Combine taxonomic, metagenomic & functional profiling, and host measurements

by extending the Bioconductor MultiAssayExperiment class.

Benefits for microbiome data science

Extended
enhances multi-assay analyses in microbiome studies

Optimized
speed, memory & scale (sample sizes, datasets, cohorts)

Integrated
with other applications & frameworks (e.g. Single Cell, Spatial transcriptomics)

Package and data ecosystem

Bioconductor methods & packages supporting the framework.

Basic wrangling: transformations, agglomeration, split/merge etc.
Alpha & beta diversity
Dimension reduction (scater)
Differential abundance (ANCOMBC, ALDEx2, benchdamic)
Visualization (miaViz)
Time series manipulation (miaTime)
Ecological simulations (miaSim)

Importers for other common formats

Linking with other common formats and microbiome data science frameworks.

Open data resources

Open microbiome data sets readily available in TreeSummarizedExperiment format via curatedMetagenomicData, EBI MGnify, Bioconductor ExperimentHub, and packages.

Minimal example

Cross-correlating taxonomic and metabolomic profiles from a dietary intervention study in mice (Hintikka et al. 2021).

library(mia)

# Import data and rename
data("HintikkaXOData")
mae <- HintikkaXOData
  
# Agglomerate rare taxa by prevalence at the Family level & log10-transform read  counts
mae[["microbiota"]] <- agglomerateByPrevalence(mae[["microbiota"]], rank = "Family")
mae[["microbiota"]] <- transformAssay(mae[["microbiota"]], method = "clr", pseudocount = 1)

# Get cross-correlation between taxa & metabolites
x <- testExperimentCrossAssociation(mae, experiment1 = "microbiota", experiment2 = "metabolites",
                                               assay.type1 = "clr", assay.type2 = "nmr",
                                               mode = "matrix", sort = TRUE)$cor

# Visualize taxa-metabolite associations on heatmap
ComplexHeatmap::Heatmap(x)

Documentation: OMA Gitbook (beta)

Complements other Bioconductor Gitbooks; overlapping methods and analysis strategies: microbiome.github.io/OMA

Community

Coordination: Tuomas Borman, Leo Lahti

Giulio Benedetti, Yağmur Şimşek, Basil Courbayre, Jeba Akewak, Daena Rys, Henrik Eckermann, Chouaib Benchraka, Rajesh Shigdel, Artur Sannikov, Lu Yang, Renuka Potbhare, S. A. Shetty, R. Huang, F. G.M. Ernst, D. J. Braccia, H. C. Bravo; 20+ contributors, 10+ countries. Welcome to join!

Acknowledgments: TreeSummarizedExperiment, SingleCellExperiment, MultiAssayExperiment, curatedMetagenomicData, MGnifyR, phILR, ANCOMBC, ALDEx2, benchdamic, scater, scuttle authors..

Courses in 2023: Pune (IND) | Turku & Oulu (FIN) | Utrecht, Wageningen & Radboud (NLD) | Norwich (UK) |Bergen (NO)

Bioconductor books

Resources

Microbiome R package listing:

github.com/microsud/Tools-Microbiome-Analysis

Microbiome data science workflow

Data containers:

TreeSummarizedExperiment
MultiAssayExperiment

R/Bioc packages:
mia, miaViz

Figure by Domenick Braccia (EuroBioC 2020)

Misc

Figure source: Moreno-Indias et al. (2021) Statistical and Machine Learning Techniques in Human Microbiome Studies: Contemporary Challenges and Solutions. Frontiers in Microbiology 12:11.

TreeSummarizedExperiment

Optimal container for microbiome data integration?

Improve interoperability, ensure sustainability.

Multiple assays
seamless interlinking

Side information
extended capabilities & data types

Hierarchical data
both samples & features

Optimized
speed & memory

Integrated
other applications & frameworks

Data integration in microbiomics

Scaling up sample size (cohorts & meta-analyses)
Complementary pipelines
Alternative experiments (e.g. 16S, metagenome, phylogenetic microarrays)
Multiple assays (e.g. counts, relative abundances, clr, phILR)
Hierarchies in feature & sample space (phylogenies, nested designs)
Cross-kingdom analyses (bacteriome, virome, mykobiome, eukaryome)
Multi-omics (taxonomy, function, host)