add GT info to the VCFs

[shinlin77]

I've successfully run the example. I don't understand the output

out/merged_snp/jurkat_chr1/final.tsv.gz

What is the final call? And is what should I use as the measure of confidence?

Thanks.

[aryarm]

Hi @shinlin77 ,

Thanks for your interest in VarCA! I've attached responses to each portion of your question below.

---

out/merged_snp/jurkat_chr1/final.tsv.gz

I think you're referring to the output of the prepare subworkflow, which is not the final output of the entire pipeline. After the prepare subworkflow, VarCA automatically runs the classify subworkflow.

What is the final call? And is what should I use as the measure of confidence?

VCFs containing the final calls as well as QUAL scores with a measure of confidence for each call are generated at the end of the classify subworkflow. That output is described here. Here's the path for the jurkat_chr1 sample:

out/classify/jurkat_chr1_snp/final.vcf.gz

[shinlin77]

Thanks for getting back to me. That makes a lot more sense. One additional question, though. How are heterozygous calls designated? In the example, it seems like all the ALT are homozygous.

[aryarm]

The VCFs produced by VarCA lack the genotype (GT) field (see #11)

In theory, it should be possible to have 2vcf.py add that information to the VCFs, since I think most of the individual variant callers have it. But it might take a bit of a refactor, and I just haven't had the time to implement that.

Is that a feature that you'd like to have?

[shinlin77]

That would be wonderful. The concept of your program should be useful for the community, but this feature would make it truly practical.

[aryarm]

ok, we'll see if I have some time for this

One caveat to this is that a variant may be incorrectly called homozygous (when, in fact, it's actually heterozygous) in situations where only one of the alleles has open chromatin
In other words, the GT info may be incorrect in loci with allele-specific open chromatin

Does that matter to you?

[shinlin77]

It's a good point that should be disclosed as a caveat, but nevertheless, I think it is important to understand whether REF or ALT or both are detected by the ATAC-seq. Actually, for the effect to which you are referring, you can quantify how often that happens, right? Shin

…

On Mon, Apr 4, 2022 at 10:42 AM Arya Massarat ***@***.***> wrote: ok, we'll see if I have some time for this One caveat to this is that a variant may be incorrectly called homozygous (when, in fact, it's actually heterozygous) in situations where only one of the alleles has open chromatin In other words, the GT info may be incorrect in loci with allele-specific open chromatin Does that matter to you? — Reply to this email directly, view it on GitHub <#43 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALVU6EPXCFGZGSJV7NGOTO3VDMSZTANCNFSM5SJTZP6A> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[aryarm]

you can quantify how often that happens, right?

In theory, it should be possible. In practice, it might be difficult. We could compare the scATAC-seq variants to variants from WGS, but there could be other reasons that the two would differ from each other besides allele-specific open chromatin. So we would have to come up with some measure (and a threshold on that measure) for detecting allele-specific open chromatin, and then filter for sites that pass the threshold.

[shinlin77]

In thinking this over, I don't know if we can distinguish between allele-specific open chromatin versus other reasons (i.e. insufficient coverage or noise). However, it would be good to know how often it happens. E.g. if it happens 1% of the time, maybe it doesn't matter regardless of the reason. Thoughts?

…

On Mon, Apr 4, 2022 at 2:26 PM Arya Massarat ***@***.***> wrote: you can quantify how often that happens, right? In theory, it should be possible. In practice, it might be difficult. We could compare the scATAC-seq variants to variants from WGS, but there could be other reasons that the two would differ from each other besides allele-specific open chromatin. So we would have to come up with some measure (and a threshold on that measure) for detecting allele-specific open chromatin, and then filter for sites that pass the threshold. — Reply to this email directly, view it on GitHub <#43 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALVU6ENBLSNBMSJZHCQHRDTVDNM7DANCNFSM5SJTZP6A> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[aryarm]

Hi @shinlin77 ,

I agree that it would be good to know. Somebody should do it. But developing a method to perform a test for allele specific chromatin is probably a whole project on its own. After all, testing for allele specific expression is kinda its own sub-field. I'm sorry, but I think that specific request is one that I do not personally have the bandwidth to implement within the foreseeable future.

But nonetheless, I'm going to keep this issue open to track progress on your request to add GT info to the VCFs. Afterwards, I will probably add a warning in the documentation explaining this situation and cautioning that the GT info may be inaccurate and should be ignored by the conventual user unless they have a way to account for allele specific open chromatin.

[aryarm]

(I'm just renaming the title of the issue to reflect its new purpose.)

If you have any other feature requests, kindly submit them in a new issue. Or, if you have a question or discussion item, you can add it to the Discussions tab.

[shinlin77]

WGS would be a reasonable gold standard. You also had a gold standard for your paper too. Can you not make some extensions for the work in that paper and measure the discrepancy between genotype calls from ATAC-seq and the gold standard?