DMRcate produces DMRs containing CGs that have different methylated status (hyper or hypo)

[Ευθύμιος Λαδουκάκης]

Hello everyone,

I am using DMRcate to detect differentially methylated regions between two groups defined by either the subject had a dose of a drug (case) or not (control). I have methylation data for my subjects from an EPIC array and I have 3 subjects as control and 2 subjects as cases.

The thing is that the detected DMRs contain CGs that don't follow the same methylation pattern (hyper or hypo). In some DMRs two CGs are hypermethylated, the next two hypomethylated, the following hyper again etc.

I checked this from the $input data in dmrcoutput variable (see my code below) but I also confirmed it by plotting it using dmr.plot.

My R code is the following:

#model with dose
X <- model.matrix(~dose)

#dmr detection

myannotation <- cpg.annotate(datatype="array", object=bVals, arraytype = "EPIC", what="Beta", analysis.type="differential", design=X, coef=2, fdr=0.1)

dmrcoutput <- dmrcate(myannotation, lambda=1000, C=2, mc.cores=8)

res<-extractRanges(dmrcoutput,genome="hg19")

#plotting -- not the actual plotting code but simplified for this example

groups <-c(dose="magenta", control="forestgreen")

type <- factor(table$dose)

cols <- groups[as.character(type)]

DMR.plot(ranges=res, dmr=5, CpGs=bVals, what="Beta",

         arraytype = "EPIC", phen.col=cols2, genome="hg19")

What am I missing? Is DMRcate supposed to give DMRs like this? Can this even be considered a DMR if the CGs inside it have different methylation status?

The following image is from DMR.plot displaying group means of consecutive CGs (green is control group and purple is case group).

Thank you in advance for any answers.