Hi all,

I'm struggling with the question of how to properly analyze the RNA-seq data I'm currently working on and your insights are very welcome. As a second question, we're interested in looking at the effect of treatment over time, so a design with time and treatment is the obvious option - with a total of 6 groups.

However, as the experiment was not designed with this in mind, the data consist of 5 groups (n=5/group). For each treatment there is a group at 1.5h and 3h, but the t=0 time point is represented by a single naive group of animals.

My question is whether or not I can analyze the data together (including the t=0 for both groups in some way), or if I should analyze the 1.5h en 3h time points for each treatment apart from the t=0 time point and check separately what the expression does per treatment over time.

Biologically, I expect both groups at t=0 to be similar, but of course there would still be an amount of variation between the groups. Another suggestion I received is to split the t=0 group in two so you do have six groups for the analysis, but this will results in very small groups.

Curious to hear your thoughts on this.

Thank you very much,

Rob

This support site is intended to help people with technical questions about Bioconductor software, not help people with statistical advice. I would in general recommend getting local help from a statistician with whom you can have a collaborative relationship rather than depending on free advice from random internet people, but if that's your thing, you might try at biostars or stackoverflow.

I'd echo James' point that you may do well to discuss the statistical approach with a local collaborator.

For DESeq2 you're probably just best off having a factor group with levels for all the 5 groups. You can then do pairwise comparisons. Unlike the time course example in the vignette, you can't for example, remove the baseline differences (t=0) for the treated samples because those samples don't exist.