Calling multiple BAM files for SNPs and vcf

[newbietonextgen]

Hi all,

I am in the process of calling SNPs using GATK. I do know that the unifiedgenotyper will call from multiple BAM files. The question is about the resulting VCF file.
1) Does the VCF file separate out the SNps? In other words, will i know which BAM sample file contributed to what SNP in the VCF file? In such as way i could just parse back individual sample SNPs back, if needed.
2) Also it it better to call SNPs from individual files or pool all the BAM files from the same treatment to get better depth?

The BAMs have unique Read Groups. Any help will a appreciated.

Thanks

[jstjohn]

1. I think GATK uses the Read Group tag to partition the data into separate organisms for variant calling. If you want them treated as the same sample then you give them all the same read group tag, while if you want them treated differently then give them different read group tags. I am actually not 100% sure that "read group" is the right tag for this, so someone please correct me if thats not the case.

There is a program in the Picard package to manipulate the read group along with other tags in bam files.

2. Read through this page which describes the different suggested protocals in order of how good they are thought to perform vs computational time involved. I think the best (and slowest) option is to pool everything. The page gives a good overview of the suggested set of steps to follow to call variants given different computational commitments:
http://www.broadinstitute.org/gsa/wi...th_the_GATK_v2

[newbietonextgen]

Yes, read groups may help in the calling process but when you look a vcf file, can you actually distinguish which BAM file gave rise to particular SNPs?

[d17]

Originally Posted by newbietonextgen

can you actually distinguish which BAM file gave rise to particular SNPs?

If you are doing multi-sample calling with each sample in a separate BAM, it doesn't always make sense to say that one BAM file gave rise to a particular SNP. For example you could have three samples each of which has weak evidence for a particular SNP: not enough to call the SNP if you are analyzing the samples separately but when you pool them together the combined evidence is sufficient to call a SNP.