Hello everyone! I have doubts regarding the conceptual design of experiments in DESeq2. I tried to solve my problems by reading the vignette, but I think I should write down my doubts an ask for validation.
Let's consider I have a very simple experiment and I have 3 replicates for samples in the treatment group vs 3 replicates for my controls. I also assume that I don't have relevant problems with batch effects and the treatment seems the only variable of interest. Therefore I specify design= ~condition_group in my design when defining my object.
The question is: if I specify in the results function a contrast like this one contrast = c("condition_group","treatment","control") the GLM model computed will include an intercept, right? (And in this case the intercept will be computed using "control" as baseline)
Is it also correct to state that I want an intercept because my goal is to make contrasts in expression between groups by using a specific group as baseline. Not using an intercept would mean computing this difference for each group to the absence of signal, which would be not really meaningful in this circumstance. Plus, not setting intercepts introduces the assumption that the model has to be fitted in order to pass trough the origin, which is usually a bias. (The problem that this bias introduces may also have a link with the concept of the basal noise that you can have in the data regardless the group?)
Thank you in advance for the help! Eisuan
