Allele-Specific analysis for human WGBS data

Question

[Ankit]

Hi everyone,

I need to perform allele-specific methylation analysis for human whole genome bisulfite sequencing data. As Allele-specific analysis is dependent on SNPs/Polymorphic sites, I stuck with two queries:

1. Is there a way to get these polymorphic sites from WGBS data ? (I don't have any other data for these samples, I mean no exome or WGS to call variants). Has anybody tried using BS-SNPer? I also thought to use SNPs from public sources but it might not be able to resolve two parental alleles as both parents may be heterozygous for the same polymorphisms and it is difficult to tag allele to either mother or father.

1. Can allele-specific analysis possible on single sample without the data of parents?

Any suggestions would be helpful.

Thank you

Kind regards

Ankit

[Ventrilocus]

Try Bis-SNP: https://people.csail.mit.edu/dnaase/bissnp2011/

[Ankit]

Thanks I will try that tool also The other query is

Can allele-specific analysis possible on single sample without the data of parents?

[Ventrilocus]

Some background first: in allele-specific DNA methylation (ASM), you use heterozygous SNPs to label each epiallele. For simplicity, imagine you have a read with a biallelic SNP (for example, Adenine (A)/ Guanine(G)) and a CpG (Methylated (M)/Unmethylated (U)). ASM would calculate two methylation ratios:

methylation_ratio_A = M_A/(M_A + U_A)
methylation_ratio_G = M_G/(M_G + U_G)

If methylation_ratio_A ≠ methylation_ratio_G, then you say you have allele-specific methylation (ASM). Back to your question, it is possible to find heterozygous SNPs without the genotype of the parents (having trios can help make this more reliable but it is not required per se; see for example: https://academic.oup.com/bioinformatics/article/36/11/3549/5823297). Thus, it follows that trios are not required to quantify ASM.

[Ankit]

Ok. Good explanation

For example, if you look at this IGV SNAPshot (BISULFITE MODE) do I need to count Methylated / Unmethylated bases (left color) and number of SNP positions (right )?

[Ventrilocus]

Yes, that sounds right. Sometimes, you have several nearby genetic variants. Then you can talk about haplotype-allele dependent DNA methylation. You don't have to do all of this by hand. Googling a bit, I found this tool (https://www.nature.com/articles/s41467-020-19077-1; https://github.com/jordiabante/CpelAsm.jl); I haven't used it but it looks good.

[Ankit]

Thank you for your valuable suggestion.

This tool CPEL, based on Julia is bit new and I can't understand it well.

I am looking forward to use more established tools like SNPSplit.

How about aligning the reads to N-maked genome and then separate two allele using SNPSplit?

I am thinking to use dbSNP variants to create N-masked genome.

Let me know if it is feasible.

[Ankit]

Also can you tell if SNP looks like as in the below picture of IGV than is it possible to separate two allele and predict maternal and paternal methylation. This SNP is polymorphic site.