I have a methylation dataset from EPICv2 assays, and am hoping to do a DMR analysis using DMRcate. I would like to remove non-variable probes which are likely to not be differentially methylated. Is there a current best practice I could follow? I am struggling to find any recent papers which filter non-variable probes from EPICv2 beta matrices.

Thank you for your help!

Sasha

I wouldn't remove the CpGs. The idea behind DMR identification is to find regions with consistent changes in methylation, and if you selectively remove CpGs prior to identifying those regions, you are likely to bias your results. As an example, say there is a stretch of 1000 bp with 100 CpGs, of which 95 are not variable. That region might not look very promising because at most 5 CpGs are changing. But if you remove the 95 CpGs, now you have a region with five CpGs, all of which are changing, and it may look quite promising even though the truth of the matter is that most of the CpGs in that region are not actually differentially methylated.