Adjusting priors for heterozygosity for pooled samples in UnifiedGenotyper

jdavila

Hi to all,

I am calling variants in a pooled sample of 3 human samples, using UnifiedGenotyper and a ploidy of 6 and I was wondering if it is necessary to adjust the heterozygosity rate (--heterozygosity) to be 0.003 instead of the default of 0.001, or if the model used in the pooled sample calling already takes this into account.Thanks a lot,

Jaime

delangel

Hi - in theory you don't need to adjust that, the internal model math takes that already into consideration. (See explanation from Eric below)

ebanks

Ah, I see. Yes, heterozygosity is the probability that a variant occurs on any one haplotype - so you should not need to adjust it even when working with pools.

ebanks

I think it's important here to give a more precise explanation. Jaime, you are telling the UG to treat your 3 samples as each having a ploidy of 6; so you expect 18 haplotypes in the data at each position. Clearly this is not what you want. You want the UG to treat your samples as diploids - and it will automagically (assuming your BAM headers were processed correctly) separate the read data for each of your samples when calculating likelihoods. The heterozygosity too is an organism-level value and not dependent on the number of samples included. You should leave this value alone unless you know that your human samples somehow have a different heterozygosity than other humans.

jdavila

Eric

Sorry for not being more clear in my earlier post. I am running UnifiedGenotyper multi-sample calling on several samples (say 20). Each one of the samples is a pool of three individuals without barcoding or multiplexing, hence I am using ploidy of 6. Are you saying that even when I have pooled individuals in a sample I need not change the heterozygosity rate? I was thinking that since each "sample" is a combination of three individuals the chances of a heterozygous call go up by three.

Thanks a lot for your detailed answer,

Jaime