Hi, I am a bit puzzled on how to correctly analyze my expression data.

I have 4 different groups:

• cat1
• cat2
• cat3
• cat4

And I want to compare different combinations:

1. cat1+cat2 vs. cat3+cat4
2. cat1+cat3 vs. cat2+cat4
3. cat1 vs. cat3
4. cat2 vs. cat4
5. cat1 vs. cat2 vs. cat3 vs. cat4

I tried to set design = ~group within the DESeqDataSetFromMatrix() call. After the DESeq() call I get:

resultsNames(de)
[1] "Intercept"  "group_cat2_vs_cat1" "group_cat3_vs_cat1" "group_cat4_vs_cat1"

When I call results() with name = 'Intercept', I get the fifth comparison, right?

For the 3rd comparison, I used results() with contrast = c('group', 'cat1', 'cat3'). For the 4th comparison, I used results() with contrast = c('group', 'cat2', 'cat4').

I hope, I am correct so far. But how do I call results() for the first and second comparisons?

I tried contrast = list(c('Intercept', 'group_cat2_vs_cat1'), c('group_cat3_vs_cat1', 'group_cat4_vs_cat1')) but that went horribly wrong, as 99.9 % of the genes were significantly down-regulated, which I cannot believe.

I'd love some advice.

Stop using intercept. That is not at all how you get all to all comparisons. You do an LRT type of comparison to compare all to all. For the 2 vs 2, there are a few ways to do that. A simple way is just to make a new grouping, change the design to that, and use that.