GATK UnifiedGenotyper question

[yl01]

Hello all,

I was setting up my human exome sequencing analysis pipeline mostly according to the How-to/exome analysis manual. My question is in the producing raw SNP call step using UnifiedGenotyper. From the best practises on GATK web site I know that it is better to give multiple BAM files to UnifiedGenotyper at the same time and output one multi sample VCF file. However, my impression from the How-to and other exome pipelines is that giving one file to UnifiedGenotyper each time to obtain one single sample VCF file is more common. And I also find it is easier to manage the different files when I have one VCF file for each individual. I did a comparison of the two methods and number of obtained SNPs are different. So what is the benefit of calling variants using UnifiedGenotyper with multiple BAM files?

[vivek_]

Here is a link on this discussion from the GATK forums.

https://getsatisfaction.com/gsa/topics/multi_sample_vs_single_sample_snp_calling_using_unifiedgenotyper

When you call each sample independently the genotyper will not output homozygous reference calls. Whereas these calls will be output in multisample calling if atleast one of the samples being called has indication of a non reference allele present at that location.

This could explain why you are seeing more variants when calling the samples together.

I normally do multisample calling when I have family data as it makes subsequent downstream analysis easier.

[yl01]

Thank you vivek_ for poiting out the link. It answered my question well. The short answer seems to be multiple sample calling is always better.