Unsupervised method for detecting differentially methylated regions (DMR)

Question

[ResearchR]

Dear all,

I am working on whole-genome-bisulfite data and would like to detect differentillay methylated regions. When having predefine conditions for each sample (i.e. clinical infromation), this is not a big deal. Since I am working on a more or less mining-project, I would like to know if there is information hidden in my data, which has not been annotated so far. Does anyone of you know an unsupervised-like method for detecting DMRs?

Best Chris

[Sean Davis]

Have you tried binning and summarizing (mean, median, for example) your data by genomic coordinate and then making heatmaps?

[arta]

Do you know these papers :

BSmooth, Bisulfighter

[ResearchR]

As far as I know, BSmooth and Bisulfighter detect DMRs based in predefined subgroups.