How to identify additional SNPs on EPICv2

Question

[Basti]

Hi everyone,

I realized that many CpGs had a 3-distribution DNA methylation profile on EPIC v2 chips, suggesting that they are SNPs.

I've already used dropLociWithSnps with the most recent annotation, so there must be additional unidentified SNPs. To your knowledge, are there any tools that would enable me to eliminate these probes?

I thought of developing an algorithm that would identify this distribution in 3 groups at 0%, 50% and 100%, but there might be a more precise tool based on the identification of these SNPs according to their genomic positions.

Thank you for your suggestions.

Answer 1

[Papyrus]

There is actually an algorithm, you can try the gaphunter function from the minfi package. I typically play with the threshold and outCutoff parameters because they depend on your dataset size, etc. but have found the function to really detect these types of SNP-like probes.

[Basti]

It seems to work pretty well thank you. Now I have some additional interrogations, because some CpGs that I identify with the trimodal distribution 0-50-100 identified by gaphunter or MethylToSNP are not located near known SNPs according to the 1000Genome. Since you seem to have experience in array analyses, I was wondering if you had any further biological explanation that could explain why we observe such a trimodal pattern of DNA methylation. No worries if you do not have any hypothesis !

[Basti]

I just read that there are meQTL, this is possible explanation

[Papyrus]

I'm glad it worked for you! It's possible that some of them are meQTCL, although I think nearby or associated SNPs sometimes do not necessarily impact methylation in such a strong manner to generate such a "clean" 0-50-100 distribution. I guess others must be directly cohort-specific SNPs. Because arrays often have "more than enough" probes to detect biological changes in any study (if they exist), I generally filter them without giving them much thought, really.

[Basti]

Thank you very much, I'm definitely going to test this !

Answer 2

[Joe]

Our recent work provides annotation files for unambiguous probes on the Illumina 450K, EPIC/850K, and EPICv2.0/935K methylation arrays. Additionally, we offer T2T-CHM13-based annotation information, including genomic features, SNPs with a minor allele frequency > 0.05, repetitive elements, and epigenetic marks. These files are available for download at https://github.com/functionalepigenomics/Illumina_Infinium_HumanMethylation_BeadChips_Annotation.

[Basti]

Thank you very much for this valuable new annotation. However, I still observe unambiguous CpGs that are not annotated to any SNPs in your work but that still show similar pattern in my dataset. I will try to check with other EPICv2 datasets, but I'm convinced there is a SNP there.