In IFB-ElixirFr/ProMetIS: Multi-omics phenotyping of the LAT and MX2 knockout mice

knitr::opts_chunk$set(fig.width = 8,
                      fig.height = 8,
                      fig.path = 'figures/temp/tutorial/',
                      message = FALSE,
                      warning = FALSE)

This vignette describes how to upload the data from the package.

Note: The MutiDataSet/ExpressionSet Bioconductor framework is used throughout this vignette [@Hernandez-Ferrer_2017_MultiDataSetPackageEncapsulating].

Data upload

The path to the post-processed data stored in the package can be accessed as follows:

data_dir.c <- ProMetIS::post_processed_dir.c()

In this tutorial, we will focus on the metabolomics datasets:

met_sets.vc <- grep("metabolomics_",
                    ProMetIS::sets.vc(),
                    value = TRUE)

Each dataset is stored as 3 tabulated files containing the matrix of intensities, and the sample and variable metadata, respectively. The tables can be uploaded as a MutiDataSet object with the phenomis package from Bioconductor:

met.mds <- phenomis::reading(data_dir.c,
                             subsets.vc = met_sets.vc,
                             output.c = "set",
                             report.c = "none")

We further restrict to the liver data from the MX2 mice:

mx2_liv_met.mds <- ProMetIS::subsetting(met.mds,
                                         genes.vc = c("WT", "MX2"),
                                         tissues.vc = "liver")

Data analysis

We provide below two examples of single and multi-omics analysis.

The comprehensive analysis and biologicial interpretation of the datasets are currently beeing finalized by the consortium and will be soon made publicly available. Feel free to contact us for any question (etienne.thevenot\@cea.fr).

Single-omics

Let us first focus on the data obtained with the HILIC chromatographic column in the negative mode:

mx2_liv_met_hilic.eset <- mx2_liv_met.mds[["metabolomics_liver_hilic_neg"]]

The genotype information can be extracted from the sample metadata:

gene.fc <- factor(Biobase::pData(mx2_liv_met_hilic.eset)[, "gene"],
                  levels = c("WT", "MX2"))

Let us compute an Orthogonal Partial Least Square - Discriminant Analysis (OPLS-DA) model of the genotype response:

ropls::opls(t(Biobase::exprs(mx2_liv_met_hilic.eset)), gene.fc,
            predI = 1, orthoI = 1)

Multi-omics analysis

Let us now provide an example of exploration of the relationships between the datasets by using Multiple Co-Inertia [@Chessel_1996_AnalysesCoinertieNuages].

We first extract the list of matrices:

mx2_liv_met_mn.ls <- MultiDataSet::as.list(mx2_liv_met.mds)

We then perform the analysis by using the implementation in the omicade4 package [@Meng_2014_MultivariateApproachIntegration]:

mx2_liv_met.mcia <- omicade4::mcia(mx2_liv_met_mn.ls, cia.nf = 3)

For each sample, the 2 blocks as well as the reference structure maximizing the covariance can be plotted (here in the 2nd and 3rd dimensions) with:

require(ade4)
omicade4::plot.mcia(mx2_liv_met.mcia, axes = 2:3,
                    phenovec = gene.fc)

Session Info

sessionInfo()

References

IFB-ElixirFr/ProMetIS documentation built on May 21, 2024, 8:02 p.m.