It may be a trivial question, but I want to make sure I am not missing anything.
When I am using DESeq2 for DEA with the default DESeq2 options of results or lfcShrink functions (namely lfcThreshold=0), the P-value histogram usually looks perfect (large excess of P-values in the lowest bin, and much fewer P-values across higher bins that are uniformly distributed).
However, always when I am using the lfcThreshold > 0 option (e.g. lfcThreshold = 1), the P-value histogram shows a very low powered test (almost all P values are 1). I do get hundreds of significant genes though.
I understand that the test is much stricter in this case and it can explain that. But would it even be expected to get this type of histogram in this case? Or it is reasonable to expect the well-behaved P-value histogram I get when the null hypothesis is the default (never happened to me).