I had been doing analysis using DESeq2 such that I would get the results from DESeq2, do a FDR (adj p-value) cutoff of 0.05 and then from that resulting list do a log(2) cutoff of greater than 1 and less than -1. I will call this Method A, 'sequential cutoffs'
I have come to realize that the final list of genes obtained from this 'sequential cutoffs' way is not something that can be labelled as FDR 0.05, log(2) 1. Pardon the gross statistical ignorance, but this is not readily intuitive to a biochemist. If I want a list of genes that can be labelled as FDR 0.05, log(2)1, I now use lfcThreshold of 1 in my DESeq2 command and from those results do a FDR cutoff of 0.05. I will call this Method B.
However, in a comparison of gene lists obtained from a number of samples analyzed using each of the above mentioned ways (Method A and Method B), I found that using the lfcThreshold in the command always results in a smaller gene list than when the lfcThreshold is not used but two sequential cutoffs (FDR 0.05 followed by a log(2) 1 cutoff) is used.
Intuitively I would expect both methods to provide equal gene lists. Being a bit more statistically savvy would obviously be helpful to me, so I was looking for an explanation as to why Method B (using lfcThreshold) results in a smaller number of genes than Method A.