First of all hello, this is my first post, although I've been reading you for years.
I want to use DESeq2 in order to analyze an experiment with, in my opinion, a tricky design. It consists in a control-treatment experiment split into 4 time points, with two different groups that receive treatment or placebo depending of the time, as shown here:
T1: Group A and group B basal levels, without receiving treatment.
T2: Group A have been receiving treatment for 3 months since T1 while group B received placebo.
T3: Group A and group B haven't received anything during 3 months (washing time).
T4: Group A have been receiving placebo for 3 months since T3 while group B received treatment.
I've been thinking about reducing the time points: T1 and T3 will be an initial time (T0), because they are independent of the treatment; and T2 and T4 the final time (T1).
Would this assumption be correct? The problem I see is that I lose the possible random effects from patient.
After reducing the time variable, I only have 2 times and 2 conditions (control/treatment), that I would analyse with an LRT in DESeq2.
design = ~ Time + Condition + Time:Condition
reduced = ~ Time
I don't think this is a good approximation, but I don't have experience with this kind of designs.
Do Anyone know about similar designs, or have experience with them?
Thank you
Is is not that you have 3 different variables at play here? - group (A/B), time-point (T1 - T4), and treatment (placebo/treatment)? I would merge the group and treatment factors into a single factor, and then run the analysis as an interaction between time-point and this new merged factor.