Thanks a lot for your reply. I was running all different packages and having the same result, now it is clear for me.
Dear all,
I have some RRBS data with very low coverage and high number of NAs which they are aligned with BSMAP (more than 900,000 rows contain NAs). I did try many different packages but in all of them the DMRs and DMcs are completely different for example if gene A is detected as significantly methylated DMR in promoter region, there is no DMC CpG sites assigned to this gene from significantly methylated DMC output. Do you have any idea it? Many thanks in advance for your suggestions, Rahel
2 answers
That sounds like the aggregation of read counts for all CpG over the region lifts the statistical power to make a significant call. That is the problem with low-coverage data. I have only little experience with sparse datasets, but I think there is not much you can do if the data do not have the power to make single CpG calls.
There's no reason to expect DMRs to overlap DMCs. As ATpoint aptly noted, DMR calling works by aggregating information across CpGs. Typically an individual CpG will not be sufficiently differentially methylated itself, but a region will be. DMCs themselves are less likely to have biological effects unless they clump into DMRs (there are exceptions to this when a transcription factor preferentially binds to a motif with a particular methylation state).
Many thanks for your reply. Now it is more clear for me why I have such an output.
Log in to answer this question.