DESeq2 lfcshrink, normal vs. apeglm

Question

[OST]

Hello!

I want to kindly ask a question regarding the lfcshrink function of DESeq2.

We realized that the log2fold changes in our data are abnormally large (above 20), so we decided to use the lfcShrink() function. However, I have been doing some reading and I realize there are two main types of lfcShrink to deflate large log2fold values, type="apeglm" and type="normal". Does it make a systematic difference on which type we use? I have attempted both types on our data, and the results seem similar, although apeglm had a larger range of log2fold absolute changes (0 to 4) while normal was between 0-1 absolute log2fold changes. Which one do you recommend using?

Thank you so much in advance!

Answer 1

[Kevin Blighe]

You should probably ask this on the Bioconductor Support Forum: https://support.bioconductor.org/

However, there, you will probably be directed to the extensive vignette (it's really really good): Alternative shrinkage estimators

Kevin