DESeq interactions: What genes that are different between two experiments affects the conditions

Question

[compuTE]

Hello!

I'm super noob on DESeq interactions so sorry in advance if this question is duplicated. I just need to ask directly to see if I understand correctly.

I have two experiments performed in young and old mice. I want to test in both for knock down vs control of a certain gene. I want to make a differential expression analysis between conditions (ko vs ctrl) and between experiments (young vs old).

My coldata looks like

    sample     condition    experiment
    mouse1     ctrl            young
    mouse2     ctrl            young
    mouse3     ctrl            young
    mouse4     ko              young
    mouse5     ko              young
    mouse6     ko              young
    mouse7     ctrl            adult
    mouse8     ctrl            adult
    mouse9     ctrl            adult
    mouse10    ko              adult
    mouse11    ko              adult
    mouse12    ko              adult

But I am not really interested on what are all the differences between young and old mice, but only those that have an impact on the ko vs ctrl test.

So my design for the DESeq object is ~ condition + experiment + condition:experiment

> resultsNames(dds)
[1] "Intercept"                "condition_ko_vs_ctrl"     "experiment_young_vs_adult"   "conditionko.experimentyoung"

I am not sure, but the intuitive thing to do would be to take experiment_young_vs_adult. Is this correct?

Also, could someone explain why do I not have "conditionko.experimentadult", "conditionctrl.experimentadult", "conditionko.experimentyoung" in resultsNames(dds)?

Thank you in advance!

Answer 1

[swbarnes2]

I found this website helpful in working out the right way to deal with interactions

https://rpubs.com/ge600/deseq2

I think you likely want "conditionko.experimentyoung", or what's likely more useful is to relevel age so that young is the control, and then the contrast will be "conditionko.experimentadult"

[compuTE]

Thank you! This was very helpful.