I performed variance stabilized transformation on raw gene counts. Now for the future purpose, I want to combine replicates and calculate the log fold change values between control and experimental condition (I don't want to perform differential expression analysis). How can I combine the VST counts? I tried combining them by mean() which is basically geometric mean as the values are in log. But I noticed that log fold change distribution is drastically shifted (from zero) for some experimental conditions with lot of replicates (more than 2). Any suggestions?

We recommend using original counts for computing LFC (discussed in workflow).