SNV calling using GATK with data from multiple lanes

[Jolin]

Hi,

I am using exome sequencing data to call SNVs with unifiedgenotyper of GATK. I have two lanes for each sample. So I merged two bam files into one with two read groups. But in the VCF file, I got two columns like GT:ADP:GQ:PL 0/1:20,3:23:56:56,0,576 0/1:23,9:32:99:153,0,676.
My questions are
(1) whether GATK treated these two as two samples because there are two read groups?
(2) does GATK called SNVs in these two lanes separately or merge the reads of them?
(3) when I calculate the minor allele frequency, shall I use both columns of GT:ADP:GQ:PL?

Eager to know the answer.
Thank you in advance.

[N311V]

In regards to your first question, I do think GATK UnifiedGenotyper would have treated each different read group as a different sample (http://gatkforums.broadinstitute.org...bout-bam-files).

Is there a particular reason you're using the UnifiedGenotyper? HaplotypeCaller is it's successor (http://www.broadinstitute.org/gatk/g...-discovery-ovw).

[Jolin]

Hi N311V,

Thank you very much. If they treat different read groups as different samples, then the read groups of each lane are supposed to be the same, right? But this is not mentioned at all in GATK website.

I just called SNPs not indels. So unified genotyper seems to be faster. Did HaplotyperCaller run better than Unified Genotyper in your project?

[westerman]

From the GATK web page:

The HaplotypeCaller is a more recent and sophisticated tool than the UnifiedGenotyper. Its ability to call SNPs is equivalent to that of the UnifiedGenotyper, and its ability to call indels is far superior. We recommend using HaplotypeCaller in all cases, with only a few exceptions:

If you want to analyze more than 100 samples at a time (for performance reasons)
If you are working with non-diploid organisms (UG can handle different levels of ploidy while HC cannot)
If you are working with pooled samples (also due to the HC’s limitation regarding ploidy)
In those cases, we recommend using UnifiedGenotyper instead of HaplotypeCaller.

Personally I am not sure which is better. Getting different results bioinformatically is not a proof of correctness.

[athomson]

In regards to your first question, I do think GATK UnifiedGenotyper would have treated each different read group as a different sample (http://gatkforums.broadinstitute.org...bout-bam-files).

If you look at the desc of the SM tag in that page, its seems GATK would treat all read groups with the same SM as coming from the same sample

GATK tools treat all read groups with the same SM value as containing sequencing data for the same sample. Therefore it's critical that the SM field be correctly specified, especially when using multi-sample tools like the Unified Genotyper.

[N311V]

If they treat different read groups as different samples, then the read groups of each lane are supposed to be the same, right? But this is not mentioned at all in GATK website.

I did read somewhere on the GATK website that each sample needs a unique read group, sorry don't have a link right now. To keep track of lane perhaps you could use picard tools AddOrReplaceReadGroups.jar and specify the library name as the lane.

I just called SNPs not indels. So unified genotyper seems to be faster. Did HaplotyperCaller run better than Unified Genotyper in your project?

I was interested in SNPs and indels which made HaplotypeCaller an great all-in-one solution. Also, I was only interested in a couple of genes so speed was not a concern. I haven't compared the SNP results from HaplotypeCaller to UnifiedGenotyper so can't say if they're the same. I assume so but better check.

[Jolin]

Hi Westerman, Thank you. Actually our lab used Unified Genotyper all the time and did some PCR validation on the predicted SNVs. It seems that UG works well in SNV detection.

[Jolin]

Thanks a lot, N311V