Tutorial 4: Convert between SummarizedExperiment and CommunityProfile

Both SummarizedExperiment (SE) and CommunityProfile (CP) containers efficiently integrate microbiome profile data into one comprehensive object, from which information is easy to retrieve and analyse. The former originates from SummarizedExperiments.jl, whereas the latter belongs to the EcoJulia framework, and the way they internally organise the data is slightly divergent. Nevertheless, interoperability and conversion between them is possible.

From CP to SE

First, a CommunityProfile is created from its building blocks and some metadata about the sampling sites (origin) is added.

using Microbiome

# generate array with sample data
samps = MicrobiomeSample.(["s$i" for i in 1:10]);

# generate array with feature data
taxa = [[Taxon("s$i", :species) for i in 1:10]; [Taxon("g$i", :genus) for i in 1:10]];

# generate matrix with random entries
mat = rand(20, 10);

# create cp instance
comm = CommunityProfile(mat, taxa, samps)

# add some metadata about origin
for (sample, value) in zip(1:length(samples(comm)), ["origin$i" for i in 1:length(samples(comm))])

    set!(samples(comm)[sample], :origin, value)

end

Next, the CommunityProfile can be reshaped into a SmmarizedExperiment object by redifining its object type, i.e., SummarizedExperiment(comm::CommunityProfile). This method can also be applied to functional profiles, where GeneFunction instances are found in place of Taxon ones.

# convert cp to se
se_converted = SummarizedExperiment(comm)

20x10 SummarizedExperiment
  assays(1): comm
  rownames: s1 s2 ... g9 g10
  rowdata(3): name species genus
  colnames: s1 s2 ... s9 s10
  coldata(2): name origin
  metadata(0):

From SE to CP

A SummarizedExperiment object can also be converted into a CommunityProfile. As an example, we will use the se constructed in the first tutorial, which is redefined through CommunityProfile(se::SummarizedExperiment).

# view se
se
# convert se to cp
comm_converted = CommunityProfile(se)

CommunityProfile{Float64, Microbiome.Taxon, Microbiome.MicrobiomeSample} with 20 features in 501 samples

Feature names:
g1, g2, g3...g19, g20

Sample names:
t1, t2, t3...t500, t501

It is highly recommended for the conversion to succeed that the rowdata of the original SummarizedExperiment object contains columns named by and specifying at least one of the following:

• strain
• subspecies
• species
• genus
• phamily
• order
• class
• phylum
• kingdom

Further applications

BioBakeryUtils.jl implements MetaPhlAn and HUMAnN ports for the Julia language. Both these workflows return an abundance or gene assay in the form of a CommunityProfile, which can be easily converted into a SummarizedExperiment object. Thus, results from several sources, such as MetaPhlAn and HUMAnN, can be integrated into a single MultiAssayExperiment and analysed in parallel, as explained in tutorial 2.