UnifiedGenotyper / Strand Bias (SB)

nguyenqxxnguyenqxx Member Posts: 3
edited January 2013 in Ask the GATK team

Hi,

How to get all SNPs variants with Strand Bias (high SB value) ?

GenomeAnalysisTK.jar -T UnifiedGenotyper  \
    -R  ucsc.hg19.fasta   -D dbsnp_135.hg19.vcf  \
    -stand_emit_conf 1.5 -stand_call_conf 1.5 \
    -o Sample_2.snv.vcf  -I Sample_2.recal.bam -nt 10

Thanks,
Q

Post edited by Geraldine_VdAuwera on

Comments

  • Geraldine_VdAuweraGeraldine_VdAuwera Administrator, Dev Posts: 10,292 admin

    You probably want to take a look at SelectVariants:
    http://gatkforums.broadinstitute.org/discussion/54/using-selectvariants

    Geraldine Van der Auwera, PhD

  • nguyenqxxnguyenqxx Member Posts: 3

    Hi Geraldine,

    Thanks for responding my quesitns. I looked at the 'SelectVariants', which is great option that I will need later on a project. However, what I try to find out is:
    How to get all possible variatants (SNP & INDEL) that "GenomeAnalysisTK.jar -T UnifiedGenotyper" detects WITHOUT filter anything. Because the coverage of our sampels are low, thus I want to have as much as sensitive as possible.

    Thanks for your help.

  • Geraldine_VdAuweraGeraldine_VdAuwera Administrator, Dev Posts: 10,292 admin

    Ah, I see -- I misunderstood your original question, sorry. Take a look at the --output_mode argument of the UG, see if that will give you what you want:
    http://www.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_sting_gatk_walkers_genotyper_UnifiedGenotyper.html#--output_mode

    Geraldine Van der Auwera, PhD

  • nguyenqxxnguyenqxx Member Posts: 3

    Hi Geraldine,

    Thanks for the links, its great one. I've tried all the "--output_mode" options
    EMIT_VARIANTS_ONLY (default) - doesn't call ALL possible Variants (by verifying with IGV)
    EMIT_ALL_CONFIDENT_SITES - Gives similar (little more/less) variants compare with 'EMIT_VARIANTS_ONLY'
    EMIT_ALL_SITES - gives every sites (Doesn't make sense for a current project!) I could use "mpilleup" instead if needed.

    Overall, I am still struggle with calling all possible variant (regardless any filter!) on a single sample. I like to apply a variant-filter later. Do you a better idea?

    Thanks,
    Q

  • ebanksebanks Broad InstituteMember, Administrator, Broadie, Moderator, Dev Posts: 698 admin

    See --standard_min_confidence_threshold_for_calling.

    Eric Banks, PhD -- Director, Data Sciences and Data Engineering, Broad Institute of Harvard and MIT

  • nguyenqxxnguyenqxx Member Posts: 3

    On my original question, I've palyed with '-stand_emit_conf 1.5 -stand_call_conf 1.5' which allow very low phred score! There must be other factors that interfere the variant calling. I have yet to figure-it-out.

    Thanks,
    Q

  • Geraldine_VdAuweraGeraldine_VdAuwera Administrator, Dev Posts: 10,292 admin

    You'll have to go through the tech docs and evaluate which parameters you want to adjust. Please note however that by default the genotyper is already very aggressive in calling variants, so if you lower its thresholds even more you might end up with a lot of junk calls, and we can't help you deal with that since it goes against our best practices recommendations.

    Geraldine Van der Auwera, PhD

Sign In or Register to comment.