In beadyallen/MGnifyR: R interface to EBI MGnify metagenomics resource

library(knitr)
knitr::opts_chunk$set(
    collapse = TRUE,
    comment = "#>",
    cache = TRUE
)

# Get already loaded results
path <- system.file("extdata", "vignette_MGnifyR.rds", package = "MGnifyR")
vignette_MGnifyR <- readRDS(path)

Introduction

MGnifyR is a package designed to ease access to the EBI's MGnify resource, allowing searching and retrieval of multiple datasets for downstream analysis.

The latest version of MGnifyR seamlessly integrates with the miaverse framework providing access to cutting-edge tools in microbiome down-stream analytics.

Installation

MGnifyR is hosted on Bioconductor, and can be installed using via BiocManager.

BiocManager::install("MGnifyR")

Load MGnifyR package

Once installed, MGnifyR is made available in the usual way.

library(MGnifyR)

Create a client

All functions in MGnifyR make use of a MgnifyClient object to keep track of the JSONAPI url, disk cache location and user access tokens. Thus the first thing to do when starting any analysis is to instantiate this object. The following snippet creates this.

mg <- MgnifyClient(useCache = TRUE)
mg

The MgnifyClient object contains slots for each of the previously mentioned settings.

Functions for fetching the data

Search data

doQuery() function can be utilized to search results such as samples and studies from MGnify database. Below, we fetch information drinking water samples.

# Fetch studies
samples <- doQuery(
    mg,
    type = "samples",
    biome_name = "root:Environmental:Aquatic:Freshwater:Drinking water",
    max.hits = 10)

samples <- vignette_MGnifyR[["samples"]]

The result is a table containing accession IDs and description -- in this case -- on samples.

colnames(samples) |> head()

Find relevent analyses accessions

Now we want to find analysis accessions. Each sample might have multiple analyses. Each analysis ID corresponds to a single run of a particular pipeline on a single sample in a single study.

analyses_accessions <- searchAnalysis(mg, "samples", samples$accession)

analyses_accessions <- vignette_MGnifyR[["analyses_accessions"]]

By running the searchAnalysis() function, we get a vector of analysis IDs of samples that we fed as an input.

analyses_accessions |> head()

Fetch metadata

We can now check the metadata to get hint of what kind of data we have. We use getMetadata() function to fetch data based on analysis IDs.

analyses_metadata <- getMetadata(mg, analyses_accessions)

analyses_metadata <- vignette_MGnifyR[["analyses_metadata"]]

The returned value is a data.frame that includes metadata for example on how analysis was conducted and what kind of samples were analyzed.

colnames(analyses_metadata) |> head()

Fetch microbiome data

After we have selected the data to fetch, we can use getResult()

The output is r BiocStyle::Biocpkg("TreeSummarizedExperiment") (TreeSE) or r BiocStyle::Biocpkg("MultiAssayExperiment") (MAE) depending on the dataset. If the dataset includes only taxonomic profiling data, the output is a single TreeSE. If dataset includes also functional data, the output is multiple TreeSE objects that are linked together by utilizing MAE.

mae <- getResult(mg, accession = analyses_accessions)

mae <- vignette_MGnifyR[["mae"]]

mae

You can get access to individual TreeSE object in MAE by specifying index or name.

mae[[1]]

TreeSE object is uniquely positioned to support SummarizedExperiment-based microbiome data manipulation and visualization. Moreover, it enables access to miaverse tools. For example, we can estimate diversity of samples...

library(mia)

mae[[1]] <- estimateDiversity(mae[[1]], index = "shannon")

library(scater)

plotColData(mae[[1]], "shannon", x = "sample_environment..biome.")

... and plot abundances of most abundant phyla.

# Agglomerate data
altExps(mae[[1]]) <- splitByRanks(mae[[1]])

library(miaViz)

# Plot top taxa
top_taxa <- getTopFeatures(altExp(mae[[1]], "Phylum"), 10)
plotAbundance(
    altExp(mae[[1]], "Phylum")[top_taxa, ],
    rank = "Phylum",
    as.relative = TRUE
    )

We can also perform other analyses such as principal component analysis to microbial profiling data by utilizing miaverse tools.

# Apply relative transformation
mae[[1]] <- transformAssay(mae[[1]], method = "relabundance")
# Perform PCoA
mae[[1]] <- runMDS(
    mae[[1]], assay.type = "relabundance",
    FUN = vegan::vegdist, method = "bray")
# Plot
plotReducedDim(
    mae[[1]], "MDS", colour_by = "sample_environment..biome.")

Fetch raw files

While getResult() can be utilized to retrieve microbial profiling data, getData() can be used more flexibly to retrieve any kind of data from the database. It returns data as simple data.frame or list format.

publications <- getData(mg, type = "publications")

publications <- vignette_MGnifyR[["publications"]]

colnames(publications) |> head()

The result is a data.frame by default. In this case, it includes information on publications fetched from the data portal.

Fetch sequence files

Finally, we can use searchFile() and getFile() to retrieve other MGnify pipeline outputs such as merged sequence reads, assembled contigs, and details of the functional analyses.

With searchFile(), we can search files from the database.

dl_urls <- searchFile(mg, analyses_accessions, type = "analyses")

dl_urls <- vignette_MGnifyR[["dl_urls"]]

The returned table contains search results related to analyses that we fed as an input. The table contains information on file and also URL address from where the file can be loaded.

target_urls <- dl_urls[
    dl_urls$attributes.description.label == "Predicted alpha tmRNA", ]

colnames(target_urls) |> head()

Finally, we can download the files with getFile().

# Just select a single file from the target_urls list for demonstration.
file_url <- target_urls$download_url[[1]]
cached_location <- getFile(mg, file_url)

cached_location <- vignette_MGnifyR[["cached_location"]]

The function returns a path where the file is stored.

# Where are the files?
cached_location

sessionInfo()

beadyallen/MGnifyR documentation built on Nov. 3, 2024, 4:23 p.m.