Hi! I am performing a differential expression analysis with DESeq2. I built a unique dds dataset and I would like to compare different conditions. In particular I got this combination:

resultsNames(dds)
[1] "Intercept"
[2] "condition_cond1_vs_cond2"
[3] "condition_cond3_vs_cond2"
[4] "condition_cond4_vs_cond2"

Considering that I used the method apeglm to compare cond3 vs cond4 I can not use the option contrast. So, to do this I have done:

dds <- relevel(dds, ref="cond3")
dds <- nbinomWaldTest(dds, betaPrior =TRUE)
res <- lfcshrink(dds, coef=2, type=”apeglm”)

I was wondering if the workflow is correct. In addition considering that with apeglm you can not use contrast function, is there a way to get the symmetric Log2FoldChange, in other words have the same result if I perform cond1 vs cond2 or cond2 vs cond1.

Thank you!

Concetta

betaPrior should be FALSE (TRUE is basically deprecated, but kept for backward compatibility).

Otherwise, this is correct, and what we recommend from the vignette.

We have abandoned the expanded model matrix approach for apeglm and lfcShrink, as it was difficult for users to interpret (it was really a big cause of confusion on the support site while it was the default method for shrinkage). So we don't have any support for the symmetric approach now.