Hello,

I have a multi-factor design and I have successfully used DESeq2 to extract my comparisons of interest.  I wanted to use DESeq1 in order to compare (long story) but I do not know how to extract specific comparisons.  THere is actually the following comment on the vignette:

"Fixme: Can we add a paragraph on what to do if we only want to make speci c comparisons, e.g. let fac be a factor
with three levels A, B and C, and we want to test pairwise for di erences between A and B (without C), between A
and C (without B) etc"

If I just did for example:

fit <- fitNbinomGLMs( cds , count ~ treatment+gender+treatment:gender )

I see how to get the log2fold changes, how can I get p-values and adjusted p-values for the specific comparisons???

Thanks,

Ramiro

How can I make specific

I'm not sure about this. I added a tag 'deseq' and changed the title, so it's clearly a DESeq question.

If you don't get a better answer, you can subset to A and B samples to make a pairwise comparison, although this is not the same as the contrast in DESeq2, which uses all samples for dispersion estimation.

A while ago, I wrote some convenience functions to use edgeR's glmFit function on a DESeq1 CountDataSet. I've posted them here:

Using this allows you to use edgeR's glmLRT function for testing specific contrasts. I verified by both reading the code and empirical testing that this uses the same formulas and produces identical results to DESeq's workflow. (Both are simply fitting a GLMs and performing a likelihood ratio test, which are well-understood methods).