Tools / Packages to assess allele specific methylation?

Question

[Ankit]

Hi everyone,

Are there any tools/packages/pipelines to assess allele specific methylation for human data?

I have used the following methods:

1. TrimGalore > Bismark Alignment > SNPsplit (Read split)> Bismark (methylation call)

and

2. TrimGalore > Bismark Alignment > Damefinder (Coverage/Meth on SNP)

But are other approaches available?

Thanks

Answer 1

[jrose]

Hi,

What are you looking for specifically? Genome wide search of ASM?

https://github.com/nloyfer/wgbs_tools has some tools for ASM analysis but not a genome wide search.

Answer 2

[colindaven]

Nanopore data is a lot better for allele specific methylation, since you can align and differentiate haplotypes better with long reads. This is impossible with short reads.

Then I'd suggest methylartist

https://github.com/adamewing/methylartist

Nanopore themselves have produced tutorials for ASM too

https://labs.epi2me.io/gm24385-5mc-remora

Others: https://github.com/scottgigante/haplotyped-methylome

[Ankit]

Well I agree with you that Nanopore sequences will be much better for allele-specific analysis. But I have tested with short reads too and it works fine. More important is the availability of accurate genotype information of parents and offsprings.

[colindaven]

Maybe if you have trio data - but bisulfite is expensive so I didn't have this in some projects. However ONT doesn't need trios to resolve haplotypes. I'm pretty sure no one here needs convincing that reads which are 100+ times as long are far, far easier to resolve into haplotype blocks.

[Ankit]

True. Splitting the reads into haplotype blocks is possibe with longer reads.

Can you please elaborate how will you accurately predict allelic origin of reads (paternal and maternal allele) in absence of trio/duo ?

It would be interesting to know such softwares and will be helpful in my study.

[colindaven]

You can separate the reads into haplotype blocks, but could not confidently assign maternal vs paternal genome wide (after the blocks end). You'd need further data, like 1-5 x short read data from the parents, if that was important.