Deseq2 statistics treatment vs control

Question

[JE]

Hi!

In Deseq2 it has to be set which is control and which is treatment. Let's say I have 2 treatments - A and B and I run analysis so that A is control. If I now run the analysis again so that B is control and A treatment, would that change my result statistics or would just the up-regulated become down-regulated and vice versa?

Thanks!

Answer 1

[russhh]

It should only change the sign of your log-FC differnetial expression calls

[Devon Ryan]

I moved this to an answer, since it's the answer.

Answer 2

[JE]

Ok, thanks! So if I accidentally run the analysis with wrong control I could just multiply the lfc value by (-1) to get the values correct? The original analysis was run already some time ago and when I re-run it now with treatment as a control I get more DEGs with changed stats. Also the MA plot looks different (not just inverted positive and negative). So the reason for the difference is not changing the control but something else has been changed?

[russhh]

Is it possible that some other aspect of the alignment/quant/diffex workflow has changed in that time? Are you using the same versions of the tools etc

[JE]

The analysis was run in Oct 2016 so it is very likely that something has changed. The number of comparisons is bigger than 2 in my experiment and the control needs to be changed only for part of them. For all comparisons I should use exactly the same workflow and packages, right? So if I do not want to re-run the analysis for all data, I could just invert the lfc for the one I need? Just want to be absolutely sure I can do that :) Otherwise, if I re-run part of comparisons with changed packages I should repeat the analysis for all the rest of comparisons as well (that were done with correct control).