I am working with RRBS datasets. I started using Biseq for DMRs and was looking for an alternative package for DMPs/DMCs (I actually tried a few). I really like the smothing approach used in Biseq and in discussions with colleagues we realised that it would be nice to use the Biseq-smoothed methylation for DMPs. I then noticed that I had the statiscal output from the betaregression step on Biseq and therefore I applied an FDR-correction to these p-values and used that as DMPs. I do realise that I lost many sites in the smoothing steps because Biseq only keeps sites that are considered to be clustered together; but I do like the fact that it 'normalises' the methylation for each site considering the methylation in nearby sites (region-wise error control) and most importantly I particularly like the fact that Biseq uses sequencing coverage when performing the smoothing. So although I lose a big junk of my data, in theory the data I keep is of much better quality (i.e. contains less false positives).

All of that said, I would like to know the opinions of the authors of the package and other people that uses / have used Biseq - does using Biseq for DMPs make sense?

Thank you.