Hello everyone,

I have an experiment with a 2 x 2 factor design, with condition having two levels (N vs NN) , and treatment having control (C) and stress (S) levels. I want to analyze the interaction effect so I run

design(dds) <- ~ condition + treatment + condition:treatment
dds <- DESeq(dds)
resultsNames(dds)

which returned the following

"Intercept"   "condition_NN_vs_N" "treatment_S_vs_C"   "conditionNN.treatmentS"

Therefore, to analyze the interaction I run

res1 <- results(dds_pol_nod, name = "conditionNN.treatmentS")

where by reading DESeq2 vignette and ?results example 3, I interpret that "conditionNN.treatmentS" means analyzing if the effect of treatment S vs treatment C in condition NN is different of the effect of treatment S vs C in condition N.

Then, I want to know how many genes of that interaction have a padj under than 0.05, so I do

padj.cutoff <- 0.05
res1 %>%
  as.data.frame() %>% 
  pull(padj) -> p1 

which(p1< padj.cutoff) %>%  length()

obtaining

897

After that, I want to analyze the opposite, if the effect of condition NN vs N in treatment S is different from that in treatment C, so I changed the design as below

design(dds) <- ~ treatment + condition + treatment:condition
dds <- DESeq(dds)
resultsNames(dds)

"Intercept"   "treatment_S_vs_C"   "condition_NN_vs_N" "treatmentS.conditionNN"

analyze the interaction by

res1 <- results(dds_pol_nod, name = "treatmentS.conditionNN")

obtain the amount of genes that have a padj under than 0.05 but still obtaining

897

How could it be possible? I was expecting to see a different number. Am I misunderstanding something?

Regards, Mauro

Since changing the order of the terms in the design doesn't actually do anything, it makes sense that the result is the same.

But also, you are asking the same question twice.

S in N = A C in N = B S in NN = C C in NN = D

So your first question, you are looking (A/B) / (C/D). (Or (C/D)/(A/B)...it doesn't matter in terms of how many genes are DE)

Your second question then is (A/C)/(B/D). Both of those are equal to AD/BC.