DEseq2: comparison the same when reversing factor order?

Question

[gt]

Say I want to compare a treatment group to a control group. Will treatment vs. control give me the same results as control vs. treatment? Just a slightly different interpretation? I know that log(A/B) = -1*log(B/A), so I would think so. Is this true for any method I use that computes a log fold change? What if I was using FCROS or edgeR instead?

Answer 1

[swbarnes2]

If you specify with contrast what to compare to what, the factor order doesn't matter, just the order in the contrast.

[gt]

I think I wasn't clear. If I do test vs. control in the contrast and get log fold change values for the genes. Then I do control vs. test in the contrast and get log fold change values again. Are the log fold change values for the two separate contrasts going to be the same? As in... log2(treatment/control) = -1*log2(control/treatment)?

[swbarnes2]

Well, why don't you try it and look for yourself? With DESeq, I think the LFCShrink operation will slightly change the results depending on the chosen control.

[gt]

Yep, will do. Was just wondering if this would be true for any package that calculates a log fold change... such as edgeR and FCROS as well.