Subset DEG from Deseq2-use padj from results() or lfcShrink()?

Question

[emi]

Hi all,

I am doing RNAseq analysis for DEG using Deseq2 package. I am trying to subset my list of DEG using both padj value and log2FC. I have analysed my results using the Deseq() function and extracted the results using results(). I have also shrunken my Log2FC using lfcShrink().

My question is simply, when I am subseting my list of DEG, should I used the padj values from results() or lfcShrink(), as they are different. (I will be using the log2FC results from lfcShrink). I have read that performing the lfcShrink() function shouldnt change the list of DEG but obviously the list will change if I use the padj from lfcShrink().

Is the padj from lfcShrink a more reliable value or is it just a more conservative method? Thank you.

Answer 1

[ATpoint]

If you set contrasts/coefficients properly the results should/must be the same in terms of p-values since the function adds shrunken fold changes, it does not recompute significances. Please show code.

[emi]

I think I figured out what the problem was: Myalpha was set at 0.05 for results()

My initial code for results() was:

res<-results(dds,alpha= 0.05)

and for lfcShrink():

lfcres<-lfcShrink(dds,coef=2,type="apeglm",res=res)

The above code gave me different padj values. However, when I changed my results() code to:

res<-results(dds,alpha= 0.1)

The padj values were the same for both. Is there an option to set lfcShrink() to alpha=0.05?

Thank you.