DESeq2 experimental design/analysis

[crh]

Nieve statistical question

We have the following data:
control: 3 biological replicates
treatment 1: 2 biological replicates
treatment 2: 2 biological replicates

Using DESeq2 and running 3 separate comparisons we get:
1. ctl vs treatment 1 ~ 160 genes up-regulated
2. ctl vs treatment 2 ~ 660 genes up-regulated
3. treatment1 vs treatment 3 ~ NO genes up-regulated

So, if we compare data from 1 & 2 we get ~ 500 genes differentiially up-regulated between treatment 1 : treatment2, but none if we compare them directly (#3).

I assume this means T1 & T2 distributions are nearly identical, and differ from the control.

Correct?

Charles

[dpryan]

I don't know about "nearly identical", but they're less likely to be different than the control condition (note my weird wording, which allows for cases where the variance in the treatment groups is high whereas that in the control group is low...which can also lead to findings like this). BTW, you probably want to check down-regulated genes as