Notice:
If you happen to see a question you know the answer to, please do chime in and help your fellow community members. We appreciate your help!

Test-drive the GATK tools and Best Practices pipelines on Terra


Check out this blog post to learn how you can get started with GATK and try out the pipelines in preconfigured workspaces (with a user-friendly interface!) without having to install anything.

i want to understand how to interpreat AD and calulate PL from it

dhwanidhwani indiaMember

I have a following INFO field from a haploid genome VCF file

0:33,100:133:99:0,141 (GT:AD:DP:GQ:PL).

What i understand is that
1) Genotype is 0 (reference).
2) 33,100 is the depth of reference and alternate
3) 133 is the total read depth (100+33)
4) 99 is the genotype quality
5) 0,141 is the PL.

My question is

I) if the supporting reads for alternate is more than the reference than can call be the reference genotype.
II) PL is basically used to confidently denote the genotype and according to this post low values mean a genotype is more likely, and high values means it’s less likely [https://gatkforums.broadinstitute.org/gatk/discussion/5913/math-notes-how-pl-is-calculated-in-haplotypecaller] . Thus probability of call being reference is high

If we get such contradictory results how to proceed.
Also based on the above info filed can we calculate the PL field? if yes can you please elaborate, I have checked few of your blogs but i could not find it [https://gatkforums.broadinstitute.org/gatk/discussion/5913/math-notes-how-pl-is-calculated-in-haplotypecaller].

In the above blog

# Conditional probabilities calculated by HC
P(AA | Data) = 0.000001
P(AT | Data) = 0.000100
P(TT | Data) = 0.010000

You havent clearly explained what is Data here?

Answers

  • bshifawbshifaw Member, Broadie, Moderator admin
    edited July 18

    here is an updated link to the blog you posted: Calculation of PL and GQ by HaplotypeCaller and GenotypeGVCFs

    I) if the supporting reads for alternate is more than the reference than can call be the reference genotype.

    I think this is asking whether the number of reads is the absolute definer on whether a call will be made, and i would say no since haplotypecaller takes into account several other aspects into consideration before making a call. I'm sure the document mentioning how the calls are made does a much HaplotypeCaller in a nutshell better job at explaining these process than i would.

    II) PL is basically used to confidently denote the genotype and according to this post...low values mean a genotype is more likely, and high values means it’s less likely. Thus probability of_ [this] _call being reference is high
    If we get such contradictory results how to proceed. Also based on the above info filed can we calculate the PL field?

    The blog does mention the GQ gets capped at 99 so it may be higher and just not indicated in the vcf.

    In the above blog
    ...
    You have not clearly explained what is Data here?_

    PL = -10 * \log{P(Genotype | Data)}
    P(Genotype | Data) = conditional probability of the genotype G given the observed data D. In other words the probability of the Genotype given the sequence Data that we have observed.
    P(G|D) = P(G) P(D|G) / sum( P(G_i) P(D|G_i) )
    This is calculated for you by the Haplotypecaller and the formula for this is shown in Assigning per-sample genotypes (HaplotypeCaller). No, I don't believe you have values to calculate the PL from the INFO field.

  • dhwanidhwani indiaMember

    @bshifaw with respect to my first question: I forgot to mention that I have used unified genotyper not haplotypecaller. Can send me a link for the same ?

  • bshifawbshifaw Member, Broadie, Moderator admin

    No, there doesn't seem to be any documentation listed PL calculation. Second best link i could find is the source code being used to calculate the PL here

    Why not move to GATK4 with Haplotypecaller?

Sign In or Register to comment.