search

dnanexus-rnd / GLnexus

Scalable gVCF merging and joint variant calling for population sequencing projects
Apache License 2.0
150 stars 38 forks source link

--bed ranges and gvcf merging #150

Closed HagenC closed 5 years ago

HagenC commented 5 years ago

Hi,

Thank you for creating GLnexus. I have a few question regarding merging of gvcf files from deepvariant.

  1. Does GLnexus change/recalibrate GTs in anyway? I want to use GLnexus for merging gvcf files - not joint-genotyping.
  2. Is it possible to add multiple ranges to the --bed file - or all contigs in the gvcfs?

Thank you, /Hagen

mlin commented 5 years ago

Hi!

  1. The experimental deepvariant configuration does not filter or revise the gVCF variants and genotypes. It may in the future (see this twitter thread) but for now it should do more or less what you want, (with the caveat that the configuration is experimental & still to be exhaustively validated in a real cohort)
  2. The BED file usually does contain multiple regions, for exome sequencing it typically contains the hundreds of thousands of target regions, and for whole genomes one typically provides the 24 full-length reference chromosomes. Did I understand the question correctly?
HagenC commented 5 years ago

Thanks,

  1. Perfect.! I will do a sanity-check on the output to make sure.
  2. Sorry - the question was badly formulated . I can see that a bed-file is required for glnexus_cli and was wondering if there was a option for specifying contigs without specific positions i.e like --chr1 or chr1-22 or maybe --all-records to merge all available data.
mlin commented 5 years ago

Ah yea, WES studies usually have the bed file handy, but it's tedious for WGS studies to have to come up with it. I found this in the code :P

https://github.com/dnanexus-rnd/GLnexus/blob/8ea2fabf5a891bf4a9f8a584f0f820c210ae8846/cli/glnexus_cli.cc#L89

HagenC commented 5 years ago

:) It also works by adding position 0 to 9999999999 for all chromosomes in the bed file.

BTW do you have any recommendations for filtering on RNCs for trio studies? Doing some tests at the moment.