IlluminaHumanMethylationEPICanno.ilm10b5.hg38 coordinates

[mgallardogomez]

Hi,

I have noticed that the genomic coordinates in this annotation are the same as those from the IlluminaHumanMethylationEPICanno.ilm10b4.hg19 version 0.6.0 package, therefore they are not adapted to the hg38 assembly. I have checked it with several probes, comparing the coordinates with those of the EPICv2 hg38 annotation and checking the conversion between hg19-hg38 coordinates at the UCSC converter . For example for the cg06303772 probe:

EPIC ilm10b4 hg19 annotation:

> IlluminaHumanMethylationEPICanno.ilm10b4.hg19::Islands.UCSC["cg06303772",]
DataFrame with 1 row and 2 columns
                     Islands_Name Relation_to_Island
                      <character>        <character>
cg06303772 chr19:17716644-17717..             Island

> IlluminaHumanMethylationEPICanno.ilm10b4.hg19::Locations["cg06303772",]
DataFrame with 1 row and 3 columns
                   chr       pos      strand
           <character> <integer> <character>
cg06303772       chr19  17717530           +

EPIC ilmb10b5 hg38 annotation:

> IlluminaHumanMethylationEPICanno.ilm10b5.hg38::Islands.UCSC["cg06303772",]
DataFrame with 1 row and 2 columns
                     Islands_Name Relation_to_Island
                      <character>        <character>
cg06303772 chr19:17716644-17717..             Island

> IlluminaHumanMethylationEPICanno.ilm10b5.hg38::Locations["cg06303772",]
DataFrame with 1 row and 3 columns
                   chr       pos      strand
           <character> <integer> <character>
cg06303772       chr19  17717530           +

EPICv2 (930k) 20a1 hg38 annotation:

> IlluminaHumanMethylationEPICv2anno.20a1.hg38::Islands.UCSC[grep("cg06303772", rownames(IlluminaHumanMethylationEPICv2anno.20a1.hg38::Islands.UCSC)),]
DataFrame with 5 rows and 2 columns
                          Islands_Name Relation_to_Island
                           <character>        <character>
cg06303772_BC21 chr19:17606051-17606..             Island
cg06303772_BC22 chr19:17606051-17606..             Island
cg06303772_BC23 chr19:17606051-17606..             Island
cg06303772_BC24 chr19:17606051-17606..             Island
cg06303772_BC25 chr19:17606051-17606..             Island

> IlluminaHumanMethylationEPICv2anno.20a1.hg38::Locations[grep("cg06303772", rownames(IlluminaHumanMethylationEPICv2anno.20a1.hg38::Locations)),]
DataFrame with 5 rows and 3 columns
                        chr       pos      strand
                <character> <integer> <character>
cg06303772_BC21       chr19  17606721           -
cg06303772_BC22       chr19  17606721           -
cg06303772_BC23       chr19  17606721           -
cg06303772_BC24       chr19  17606721           -
cg06303772_BC25       chr19  17606721           -

I would appreciate any comments or hints if I am using the annotation in a wrong way to explain that coordinates.

Thank you in advance!

[rafalcode]

I think the Illumina b5 product files are a bit awkward as they retain the hg19 values in place and merely append the hg38 values with special names as if they were in a rush. I make my own data frame (anflt) out of the annotation files anyway with a smaller set of columns and overwrite the chr and pos columns like so anflt$pos <- 1 + as.integer(anflt[,'Start_hg38']) anflt$chr <- anflt[,'CHR_hg38']

[rafalcode]

oh don't forget the strand as well anflt$Strand <- anflt[,'Strand_hg38'] An interesting insight is to chase up a gene that has changed chromosome from hg19 to hg38. Let's take cg13873802 on gene MUCIN2, it has been moved from chr7 to chr5 in hg38. So the Islands_Name entry still gets chr7, which appears to say that CpG Island analysis has not been recalculated on hg38. Or at least Illumina haven't bothered with that.

[ehugos]

I noticed this as well, do any of you have resources to recommend for correctly running programs such as minfi and DMRcate with correct hg38 annotation for EPICv1. There seems to exist EPICv2 annotation for hg38, but to my knowledge no hg38 annotation packages exist for EPICv1 (seeing how this annotation is still providing hg19 coordinates rather then hg38)?

[rafalcode]

oh yes EPICv2 annotations for hg38 definitely do exist and are accessible .. I downloaded them and I intended to generate an EPICv1 hg38 from them, but I predict an issue with some v1's not being in v2. I am not sure, but I also ran out of time ...

[Checunmily]

I also got the same coordinates when I trying to annotate epic chip using 10b4.hg19 and 10b5.hg38, and these coordinates can be match to those annotated by 20a1.hg38 for epicV2 after using Liftover to transform them. Is there any solution or update for this package? I don't want to do that every time...