Add support for ANOVA-like analyses in multiGSEA()

[lianos]

Currently we only support testing a specific contrast (either via a coefficient or a contrast vector), but we often do differential expression analysis where we want to test mean shifts in several groups of samples, which we would like to support here.

For instance, if we had a 3-group experiment, ie. a group factor had "control", "treatment01", and "treatment02" levels, we might do a limma analysis like so:

des <- model.matrix(~ group)
fit <- eBayes(lmFit(y, des))
res <- topTable(fit, coef = 2:3, ...)

We want to support that in a GSEA-like manner.

[dktanwar]

Hi @lianos

Is there any update on this?

[lianos]

Yes, there is some basic support for ANVOA analyses followed up with enrichment testing.

I'm only supporting GSEA testing using the "enrichtest" method, which is a wrapper to limma::kegga (which, itself is like goseq).

You can see an example of how that works in this unit test.

I just added that test along with some bugfixes to support that in the latest @develop branch, so you will have to devtools::install_github("lianos/multiGSEA@develop") to get it to work.

Alternatively, you can run the ANOVA (or whatever) test outside of the whole multiGSEA pipeline, and run your enrichment testing using the enrichtest method directly. The examples in ?enrichtest should be enough to get you going.

Let me know if you have any issues.