When Using Deseq Does One Need To Estimate The Size Factor And Dispersion For Each Condition?

Question

[camelbbs]

Hi all,

I want to ask a question in using DESeq, do I also need to import two conditions each time to do the estimatesizefactor and estimatedispersions? Can I import all conditions (>=3) together, and then do sizefactor and dispersion together, and then use res<-nbinomTest(cds,"cond1","cond2") to select which two conditions I want to compare?

In DESeq document, I see in the chapter "working partially without replicates", it indicates to do estimatesizefactor to new cds, but in the chapter " working without any replicates", I see there don't need to do that. So What is the affection if I do estimatesizefactor with all or partial conditions, since I see the sizeFactor is different.

thanks a lot!

Ch

Answer 1

[kevin.dalgaard]

I have been wondering about the same, since it would ease up the analysis. However, I get different sizefactor for Ctrl depending on if I import all three conditions at the same time or whether I do it pairwise e.g. Ctrl VS KO and Ctrl vs HET.

[camelbbs]

Hi kevin, I think this is a question I asked several month ago, now i got the answer is you need to import your conditions pairwise every time.

[gaelgarcia05]

Hi camelbbs,

Could you expand on your answer? I am going through the same dilemma currently.

Thanks! C

Answer 2

[camelbbs]

Anybody can look at this? Thanks!!