Pooled sample parameters to be used for haplotype callers

Hi Everyone,
I want to use haplotype caller for detecting snps, but this time I am dealing with pooled samples.
Before pooling and doing sequencing on this WGS exp, I want to ask how GATK HAPLOTYPE CALLER deal with pooled samples. My samples are haploid in their ploidy level, so if I pool 9 samples into one, then ploidy would be 9. Can GATK haplotype caller take this as ploidy level or is it too high?

To tackle this thing I already have 9 samples from my previous sequencing, each sequenced individually. I have already called snps on them. But now I am pooling them together(concatenating the fastq files) to test how will the result look like. I would like to tell you that coverage is not an issue as I am using yeast genome. All the above 9 samples which I want to test have 100X coverage(even more).

I would like you to point me as to how should I change parameters that I can detect snps when i pooled my samples.

First of all can you please explain these 2 parametrs
--standard_min_confidence_threshold_for_calling and --standard_min_confidence_threshold_for_emitting

How can I change them when I have only 1 sample which is a 9 sample pool.

Also can you explain me this parameter

--max_alternate_alleles. By default it is 6. But I think at a position you can have 4 alleles(A,C,G,T)??

In the end should I choose 1/9th the default value of --standard_min_confidence_threshold_for_calling and --standard_min_confidence_threshold_for_emitting so as to call variants now.

Let me know if I am not clear.

Hope to hear from you soon.



  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie admin
    Yes, HaplotypeCaller can take a ploidy of 9. You just need to specify it with the -ploidy argument. The QUAL score (and the thresholds for calling, which are applied to QUAL) does not scale linearly with the number of samples in a pool. The calculation takes ploidy into account and should therefore not require adjusting the threshold. That being said, if you are concerned about sensitivity you can always drop the thresholds to around 10, with the understanding that this will increase the number of false positive calls, which you'll need to filter accordingly.
Sign In or Register to comment.