Download the latest Picard release at https://github.com/broadinstitute/picard/releases.
GATK version 4.beta.5 is out. See the GATK4 beta page for download and details.

Is UGCalcLikelihodds available?

Hi,

We're interested in implementing Pasaniuc et al.'s (Nature Genetics 2012) method to use off-target next-generation sequencing data to 'genotype' some of our samples. In this paper, GATK’s UGCalcLikelihoods mode was used to compute the genotype likelihoods prior to imputation. Unfortunately, however, it doesn’t appear that UGCalcLikelihoods has been made available for public access. Is this true and can we anticipate that this might be available soon? Is there any way for members outside of Broad to access this?

If this is indeed unavailable this is a bit of a shame, both because this was published in such a high profile journal and, more importantly, because I'm certain that other researchers would like to use this method.

Can your suggest other tools that one might be able to use for this purpose?

Thanks!

Best Answer

  • ebanksebanks Broad InstituteMember, Broadie, Dev
    Accepted Answer

    Hi there,

    UGCalcLikelihoods is a private tool used for QCing the calling and not intended for general use; so no it will not be made available. That being said, had my colleagues who wrote that paper consulted me I would have suggested an alternative approach that would be easier (while still achieving the same results).

    You should use the Unified Genotyper, with '--genotyping_mode GENOTYPE_GIVEN_ALLELES' (since you want likelihoods for just the 1000 Genomes alleles), '--alleles 1000Genomes.vcf -L 1000Genomes.vcf' (to tell it which alleles from 1000 Genomes to genotype and just process those locations), and '--output_mode EMIT_ALL_SITES' (to produce likelihoods even at low confidence sites). This produces the same file that was described in the paper.

Answers

  • ebanksebanks Broad InstituteMember, Broadie, Dev
    Accepted Answer

    Hi there,

    UGCalcLikelihoods is a private tool used for QCing the calling and not intended for general use; so no it will not be made available. That being said, had my colleagues who wrote that paper consulted me I would have suggested an alternative approach that would be easier (while still achieving the same results).

    You should use the Unified Genotyper, with '--genotyping_mode GENOTYPE_GIVEN_ALLELES' (since you want likelihoods for just the 1000 Genomes alleles), '--alleles 1000Genomes.vcf -L 1000Genomes.vcf' (to tell it which alleles from 1000 Genomes to genotype and just process those locations), and '--output_mode EMIT_ALL_SITES' (to produce likelihoods even at low confidence sites). This produces the same file that was described in the paper.

  • Thank you very much for both a prompt and extremely useful reply!

    Best regards!

Sign In or Register to comment.