This is a test version of Biostars. For the public version, visit https://www.biostars.org.
phasing de novo mutations

Hi folks,

Let's suppose I called the variants in a parent child trio and filtered for de novo mutations in the child. I am now interested in the phase, that is, I would like to know whether the mutation originated in the male or female germ line. Under certain circumstances this is possible: I there is a read covering not only the de novo site but also a heterozygous polymorphism that can only have been transmitted from one parent, this information can be used for phasing.

Let's have a look at the following pseudo vcf file:

#chr pos child father mother
chr1 10   0/1   0/0      0/0
chr1 20   0/1   0/1      0/0

The second line can be phased without any further knowledge:

chr1 20   1|0   0/1      0|0

Now, if the first heterozygous mutation is on the same read as the second, then we know also the phase of this variant:

chr1 10   1|0   0|0      0|0
chr1 20   1|0   0/1      0|0

Vice versa, if the first heterozygous mutation and the second one are not on the same read, the de novo mutation arose in the maternal germ line:

chr1 10   0|1   0|0      0|0
chr1 20   1|0   0/1      0|0

Does anyone know about a software tool that does this kind of phasing?

Thanks a lot!

peter

phasing denovo next-gen
• 5,280 views • link
updated by Ram Ram • written by peter.krawitz • peter.krawitz

thx for your information.

Could you please tell me how to find de novo mutation in trio sequencing data?

thx in advance!

• 0 views • link
written by 897598644 897598644

Hi,

I use GATK Unified genotyper do generate a multiple vcf file. Then I upload the data to GeneTalk, set the affection status and filter for dominant.

If you need further assistance about GeneTalk, don't hesitate to contact me: peter at gene-talk.de

• 0 views • link
updated by Ram Ram • written by peter.krawitz • peter.krawitz

2 answers

They doesn't necessarily have to be on the same read, I think linkage equilibrium can be applied to infer haplotype of origin for de novo mutations within 1-5 kb of a mutation that satisfies mendelian inheritance.

The GATK has a tool for this that works on VCF files, ReadBackedPhasing.

• 0 views • link
updated by Ram Ram • written by Vivek Vivek

Hi Vivek,

Thanks for your answer! I had a look at the documentation of the ReadBackedPhasing tool from GATK. As far as I understood, all possible 2^n haplotypes are constructed, if we consider n variant positions. Although it didn't become clear to me from the parameters it sounds like these potential haplotypes will then be compared to known haplotype data bases to determine the most likeliest haplotype.

However, if this is how it works, it won't help with any de novo variant, as these variants cannot be in any haplotype data base yet. Thus the haplotype probabilities for the 2^n or 2^(n-1) possiblities should be the same no matter whether I include or exclude the variant position of the de novo mutation.

Please let me know if I misinterpreted you.

Cheers,
peter

• 0 views • link
updated by Ram Ram • written by peter.krawitz • peter.krawitz

I don't see the part about comparing to local databases anywhere in the documentation. As far as I know the tool consider all possible haplotypes in a given locus and picks the haplotype with the highest probability using the read information. So if your denovo mutation falls within a haplotype string with other heterozygous mutations that have been phased, you can assign the haplotype of origin.

Here is a bit more information and if you search their help forums there's a bunch of useful threads with the developer clarifying some of the issues.

• 0 views • link
updated by Ram Ram • written by Vivek Vivek

This is a good idea, but I feel the frequency of de novo mutations which occur within 50bp of an informative SNP and in sufficient numbers that you can be sure it is not a sequencing error will leave you with only a tiny handful of de novo mutations.

That being said, I don't know much about phasing de novo SNPs, so maybe there are loads and I just never saw them on my Affymetrix SNP Arrays. If tools exist to phase SNPs based on linkage dissociation, and you can add LD data into your VCF somehow, then I guess you could write a small script which puts the LD to 1 (or whatever the maximum is) for two SNPs which occur on the same read. Your phasing software should just work then.

But before you invest too much time in this, I would seriously consider taking a look at how many un-phasable, homozygous, de novo variations - with more than X reads in total and are 50bp away from an informative SNP - there are in your data. I think they'll be only a few, if any. Like less than 10.

The alternative, perhaps, is to sequence the children of the children - or at least some sperm/eggs. Even with a few new crossing over events, you should be able to correctly phase ALL the de novo SNPs, and not just those covered by informative reads.

• 0 views • link
updated by Ram Ram • written by John John

Log in to answer this question.

More posts like this