Hello, I have both 450K, Epic and Epicv2 data that I want to compare. I am very new to methylation analysis and therefore to minfi. To handle the Epicv2 I first installed the add-on of jokergoo.

BiocManager::install("remotes")
install("jokergoo/IlluminaHumanMethylationEPICv2manifest")
install("jokergoo/IlluminaHumanMethylationEPICv2anno.20a1.hg38")

I uploaded the data:

RGset450 <- read.metharray.exp(targets = targets450)
RGsetEpic <- read.metharray.exp(targets = targetsEpic)
RGsetEpic_v2 <- read.metharray.exp(targets = targetsEpic_v2)
RGsetEpic_v2@annotation <- c(array = "IlluminaHumanMethylationEPICv2", annotation = "ilm10b4.hg19")

I also preprocessed (Raw) and calculated the RatioSet for the 3 Sets.

EPIC
If I focus on the Epic data:

betaEpic <- getBeta(ratioEpic)

convert_ratio_Epicto450 <- convertArray(ratioEpic,
                  outType = "IlluminaHumanMethylation450k",
                  verbose = TRUE)

beta_convert_ratio_Epicto450 <- getBeta(convert_ratio_Epicto450)

dim(betaEpic)
dim(beta_convert_ratio_Epicto450)

The output is, as expected:

[1] 866238      8
[1] 452832      8

450K
Regarding the 450K data I double converted the ratio. First to the epic and then back to 450K. As the intersection of the epic/450K targets is not perfect, I expect to have a loss of targets during this double conversion. It should then provide the same number of targets than the Epic to 450K conversion (452832).

dim(beta450)
dim(beta_convert_ratio_450toEpic)
dim(beta_convert_ratio_450toEpicto450)

[1] 485512     10
[1] 453093     10
[1] 453093     10

Surprisingly the number of targets is different: 453093 VS 452832

EPICv2
Regarding the v2, I converted the ratio in parallel to Epic or to 450K to compare the behaviour of the package.

dim(betaEpicv2)
dim(beta_convert_ratio_Epicv2toEpic)
dim(beta_convert_ratio_Epicv2to450)

The output:

[1] 936990      8
[1] 0 8
[1] 0 8

It seems that something is wrong but I can't figure out what...

Do you have any idea concerning these 2 problems?

Thanks a lot, Alex

In the double conversion, the output you supplied indicates that the number of probes is the same (453093)?

The EPICv2 is only partially supported. You can now get the manifest files directly from Bioconductor, as Zuguang Gu submitted them for the last release. But this array is different from the EPICv1, having probes that are on both strands. It appears that convertArray is mistakenly being dispatched on the EPICv2 array as if it were an EPICv1 array, and that's why you get no probes returned.