- I am applying sPLS to integrate my gene expression and metabolomics data, can I also apply rCCA ?
Yes, sPLS has two modes to model either a unidirectional (regression mode) or bidirectional relationship (canonical mode) between the data sets. In your case, you may want to explain the expression of the metabolites (Y) by the expression of the genes (X) with a regression mode. Similar to sPLS canonical mode, rCCA models a bidirectional relationship to uncover the correlation structure between the two data sets. As the number of variables is much larger than the number of samples, you will have to choose the regularization parameters, see our website for more details.
There is a full illustrated example of a case study with sPLS canonical mode on our web-interface http://mixomics.qfab.org -> dataset examples->Highlight associations between gene expression and CNV to give an overview of such analysis.
- Can mixOmics perform an analysis for 3 datasets at one time or should I perform a pairwise analysis of two datasets?
Unfortunately, you will have to do pairwise comparisons. We are currently working on developping further methodologies to integrate multi omics data sets.
Since version 5.0-1 released in october 2013, mixOmics, linked to the RGCCA package enables to deal with more than 2 packages simultaneously.
- Can MixOmics handle multiclass data?
Yes in the case of a classification context, you can use PLS-DA if you are interested in discriminating the classes. The sparse version sPLS-DA is useful to identify discriminative features. The other approaches are purely unsupervised and therefore do not take into account the classes of the samples
- Can it combine SNP data with gene and/or metabolite expression data?
As long as the data are measured on the same matching samples, it should work.