Merge and dnmtools hmr-rep

[vagan21]

Hello! Long time user of MethPipe/DNMTools here (Emily Hodges Lab). This isn't a freature request so much as a general question.

If I were to call HMRs from a methcounts file generated with dnmtools merge, would those HMRs be the same if I didn't initially create a merged methcounts file for technical replicates and just input them straight into the dnmtools hmr-rep program? The caveat being that dnmtools hmr-rep just gives you the HMRs and you don't get the methcounts file that is a product of merging individual methcounts file to produce a single file with higher coverage.

Hope that makes sense!

[andrewdavidsmith]

No, in general they wouldn't be the same. While the segmentation will produce a set of HMRs, and the estimated transition points (into and out-of) the hmrs will be shared, separate parameters are estimated in 'hmr-rep' for each methcounts input file. If you just combine it all at the start, then only one set of parameters will be used. The reason for having this 'rep' mode is in case things like the baseline methylation level inside HMRs might be 0.03 in one rep and 0.05 in another, for technical reasons, and we want to overcome that. But we've never really had great gold-standard data to prove that it's actually helping.