I have simply question related to math:

How to calculate 'fold change' if the data is already normalized by log2. Let's say we have exprdata (log2 already) like this.

      control     control       treat       treat      treat
A -0.08658011 -0.97647981  1.48342630  2.61512874 -0.5412316
B  1.32252443  0.99083635  0.67760030  1.76150083  0.2955919

If I would like to get log2FC of change for gene "A" between treat and control, should I do like this?

mean(expr["A", 3:4]) - mean(expr["A", 1:2])

or should I convert to back 1st?

log2(mean(2**expr["A", 3:4]) / mean(2**expr["A", 1:2]))

Thank you in advance.