DEG with several conditions

Question

[Kylie]

Dear all,

I have RNA-Seq count data I want to analyze with DESeq2. My data has the following characteristics:

• there are 2 different cell lines
• there are 2 different treatments
• there are 3 different timepoints of measurement after treatment
• For both the cell lines I also have control samples

My main question is: should I create multiple DESeqDataSetFromMatrix objects, or can I just create one object and then perform all the analysis on that object? If yes, how would such a design formula look like?
Furthermore, how can I make generalized conclusions out of all these pairwise comparisons?

Thank you in advance for your inpunt!

[ATpoint]

First of all, which are the comparisons you want to test? The question on whether to split or keep together usually comes down to whether you want to compare between cell lines, and whether one line has more spread than the other. Consider checking data using PCA, see vignette, to get an idea how the spread is. The question on "general conclusions" again depends on what you want to compare.

[Kylie]

Ah good point. I want to do multiple things:

• Comparison of treated vs control samples (to see the effect of treatment on the expression)
• Comparison of the different treatments
• Investigate whether different timepoints have an effect on the two points above
• Comparison of the 3 points mentioned above in the two different cell lines

I already found with PCA that most of the variance (91%) is explained by difference in cell line.

[BioinfGuru]

Take a look at this blog, it was the biggest help to me: A guide to designs and contrasts in DESeq2

Section 4 will have the core of what you need but I'd strongly recommend also using the author's contraster() function so you can check what you are comparing (the comparison wording in the code is more intuitive).