Genotyping strategy for using UnifiedGenotyper

skblazer

Dear GATK authors,

I have two little questions about the genotyping strategy for using UnifiedGenotyper.

1) Imaging I have a list of candidate sites in hand, then I just want to use UG to genotype my individuals at these sites. Considering the sequencing coverage for each individual is low or median, will there be any difference between genotyping them on individual level and on a cohort of population level?

2) I have a list of candidate sites, including SNPs and complex indels, which were discovered by HaplotypeCaller, Then can I use UG to genotype these sites across individuals to get the correct genotypes? I know HC used a local denovo assembly to discover the variants, but I'm not sure after the variants have been discovered, whether there is any difference in genotyping result between HC and UG.

Best, SK

Geraldine_VdAuwera

Hi SK,

1) You're almost certainly better off genotyping them together. This does assume that it makes sense to group your samples into one cohort of course. Note that that's a question of experimental design which we can't comment on -- you're the best one to assess that.

2) You can genotype at given sites with either the UG or HC, regardless of which caller originally identified the sites of interest. In principle the HC gives superior results thanks to the local assembly/realignment step, especially for complex indels. But it is an experimental tool and still has a few issues, so if you experience any trouble you can fall back on the UG.