[BioC] error in dmrFinder of charm package when there is one sample per group

Tarca, Adi atarca at med.wayne.edu
Wed Jun 20 15:13:42 CEST 2012


Dear all,

I was using dmrFinder in a case with one sample per group and noticed that it fails due to a call to the function rowMedians which requires a matrix but it gets a vector when grp1 or grp2 is made of one sample:
"mat  = cbind(rowMedians(l[,grp1]), rowMedians(l[,grp2]))"
I realize that more samples per group would be a better case to be in but, given the function dmrFinder is prepared it seems for situations with one sample per group with a message like " grp1 has only 1 array!", it should still run without an error and give at least the difference in methylation scores per region.

Moreover, I was wondering if anyone used charm with data generated from the MEDIP protocol (instead of McrBC), when the treated channel is methyl-enriched (instead of methyl-depleted).
My take on this is that just switching (ut = "_532.xys", md = "_635.xys") which is default in readCharm with  (ut = "_635.xys", md = "_532.xys") would keep the meaning of the methylation scores intact and hence nothing else needs to change in the analysis, but I would be interested in the experience of other bioconductors with this issue.
Thanks,
Adi Tarca 



More information about the Bioconductor mailing list