GATK HaplotypeCallerQUAL field changes while running per chromosome

fazulur

Dear GATK Team,

I ran GATK haplotypecaller (3.4) on one sample.

1. In normal procedure . Whole bam file submitted to haplotype caller.
   2 .In another way, BAM was splitted by chromosome ( using -L option) and ran haplotype caller on each chromosome separately and concatenate the per chromosomal vcf's at the end.

I got 2 vcf files and I have observed the changes in INFO and QUAL fields while comparing both.

Does haplotype caller calculates QUAL and INFO fields across all chromosomes?

Please help me to resolve the issue.

Thanks & Regards
Fazulur Rehaman

Geraldine_VdAuwera

Yes, you can do all that per chromosome, that's fine. For filtering you should do it on the full callset however.

Geraldine_VdAuwera

Hi Fazulur, it is ok to proceed with default options.

Geraldine_VdAuwera

No, those metrics are calculated per site. What you're seeing is due to differences in downsampling in high-coverage regions. The random seed used for downsampling is based on the interval list, so that is different between your two runs, and can cause marginal differences in annotation values.

fazulur

Dear Geraldine,

Thanks a lot for your quick response.

From the above answer, i understood that i have to use downsampling option based on interval which is chrn in my case).
I will try with that and let you know if i have any queries on the same.

And could you please suggest me, can i process my data per chromosome which will follow
Splitted bam - Gatk preprocessing - Gatk variant calling - Concatenate chromosomal vcfs to final vcf.

Thanks in advance
Fazulur Rehaman

Geraldine_VdAuwera

Yes, you can do all that per chromosome, that's fine. For filtering you should do it on the full callset however.

fazulur

Dear Geraldine,

Thanks for confirming.

Thanks & Regards
Fazulur Rehaman

fazulur

Dear Geraldine,

As discussed above, I performed gatk preprocessing and variant calling on whole bam vs chromosome by chromosome..

I am working on whole genome samples and using -L argument to pass chromosome by chromosome.
From gatk preprocessing to gatk variant calling i am using default options (following best practices).

WIth reference to posts below, I came to know that downsampling is hardcoded and results will be changed by 1% when we perform haplotype caller on whole bam vs chromosome by chromosome and it is due to random seed used for downsampling.

http://gatkforums.broadinstitute.org/gatk/discussion/5008/haplotypecaller-on-whole-genome-or-chromosome-by-chromosome-different-results
http://gatkforums.broadinstitute.org/gatk/discussion/7082/haplotypecaller-gvcf-mode-on-whole-genome-vs-chromosome-by-chromosome

Do i need to use any additional options while performing gatk preprocessing or variant calling using haplotype caller while running per chromosome or is it ok to proceed with default options?

Thanks in advance
Fazulur Rehaman

Geraldine_VdAuwera

Hi Fazulur, it is ok to proceed with default options.

fazulur

@Geraldine_VdAuwera said:
Hi Fazulur, it is ok to proceed with default options.

Dear Geraldine,

Thanks a lot for confirming.

Thanks & Regards
Fazulur Rehaman