ChAMP probe.features displays only CHR and MAPINFO, and it entails epic data, whereas probe.features.epic does not exist

Question

[Line Heylen]

Hi,

I updated R and Bioconductor to use the new ChAMP package for EPIC arrays. In contrast to what is written in the manual, it seems that probe.features.epic does not exist, and that probe.features entails now the EPIC data and not the 450K data. How is this possible?

> dim(probe.features)
[1] 867531      2

> dim(probe.features.epic)
Error: object 'probe.features.epic' not found

In addition, the probe.features data entails only the columns CHR and MAPINFO and not all the other columns (gene, feature,..) it should have.

> head(probe.features)
           CHR   MAPINFO
cg07881041  19   5236016
cg18478105  20  61847650
cg23229610   1   6841125
cg03513874   2 198303466
cg09835024   X  24072640
cg05451842  14  93581139

Does anyone know how to solve this?

Thank you!

Answer 1

[Charles Warden]

Have you tried updating to the latest (non-devel) version of R?

At least for minfi, I had to update to R version 3.3.0 to get the read.metharray functions to work (although they also don't work if you are using devel version of R, at least when I tried v3.4.0)

Answer 2

[Line Heylen]

I did, this does not seem to be the problem. When I restart R; load Champ and enter data(probe.features); it successfully loads the 450K annotation with all columns. If I then load the epic annotation by data(probe.features.epic) probe.features refers to the epic array and not the 450K array and includes only the first 2 columns with chr and mapinfo.
It seems to be a problem of the ChAMP package.

[Charles Warden]

Ok - maybe they will eventually update the features for the EPIC array.

You should still be able to use champ.DMR() to define regions without the use of features (using their "ProbeLasso" or "Bumphunter" algorithms), but I've had issues with the regions called with this function in the past.

Also, I see metrics for annotated regions on the 450k array in the ChAMP user guide, but I wasn't aware of a function in ChAMP to define regions based upon the pre-existing annotations.

I know RnBeads provides region-level metrics for the EPIC array. I've also added some EPIC-array functionality to COHCAP in the standalone version, and I will eventually update the Bioconductor version (after I do some more testing with the EPIC array).

https://sourceforge.net/projects/cohcap/

You can also see some EPIC array annotations in this Bioconductor package:

http://bioconductor.org/packages/release/data/annotation/html/IlluminaHumanMethylationEPICanno.ilm10b2.hg19.html

Answer 3

[Line Heylen]

The solution offered by Tian Yuang: load data(illuminaEPICGr), which contains the information. When you load the data illuminaEPICGr, the variable in R will be named as "illumina450GR", but it's actually the EPIC array's annotation.