How to combine different "groups" in DESeq2 constrast

Question

[biotrekker]

Hello I have three groups A, B, and C I want design to be (1) A vs (B and C) (2) B vs (A and C) and (3) C vs (A and B). However with contrast parameter inside of results from DESeq2, it seems that I can only choose two at a time? What should I do? Thanks

Answer 1

[LChart]

The official way is to specify a contrast matrix, which can be created something like this:

> limma::makeContrasts(groupAvsBC=2*A-B-C,groupBvsAC=2*B-A-C, groupCvsAB=2*C-A-B,levels=c('A', 'B', 'C'))
      Contrasts
Levels groupAvsBC groupBvsAC groupCvsAB
     A          2         -1         -1
     B         -1          2         -1
     C         -1         -1          2

That said, many times I simply cheat and set dat$metaGroupA = ifelse(dat$group == 'A', 'A', 'B_or_C') and keep a simple contrast.

[biotrekker]

Seeing as how the rest of my pipeline is in DESeq2, and solutions that are provided in this package?

[LChart]

You can pass the contrast matrix into DESeq2 results directly:

result(dds, contrast=contrast.matrix[,'groupBvsAC'])

"Alternatively, a numeric vector of the length of resultsNames(dds) can be provided, for manually specifying the linear combination of terms."

The call to limma:makeContrasts simply helps to get the correct numeric vector; but you can create it manually.

Answer 2

[swbarnes2]

Just make new columns in colData that have the groups the way you want them. That's simple and readable.