Hello!

I was trying to use DMRcate to analyze a large Illumina BeadChip 450k dataset. My phenotype variable has multiple levels: apart from the control samples we have samples from 8 different hematologic cancers.

I construct my design matrix as I used to when I used the 'limma' package, but I was wondering if, instead of re-ordering my phenotype variable so that I take the control samples first, I could skip this step and use the makeContrasts() function and then cpg.annotate() with contrasts=TRUE argument to find the differentially methylated regions between each type of cancer and my control samples.

Any help would be much appreciated!

It's definitely worth utilising Limma's makeContrasts function for this purpose. From the vignette, if you look at the cpg.annotate function, you can do something like the below code. 

myannotation       <- cpg.annotate(getM(norm.filtered), 
                                   analysis.type = "differential", 
                                   design        = design, 
                                   coef          = "A - B",
                                   contrasts     = T, 
                                   cont.matrix   = cont_mat)

Where you feed the coef argument the column name from your contrast matrix. This is assuming you have a design matrix (design) and a contrast matrix (cont_mat)