how to increase our call confidence

Our call variants analysis, has more far more variants than we expect, including some low-call-confidences and non-coding variants on whole exome. Is there any suggestion on how to increase our call confidence and decrease the total number of variants. Specifically, I had question about the out_mode option.

below is a sample code of our UnifiedGenotyper,

java -Xmx8g \
-jar /medpop/mpg-psrl/Parabase/Tools/GenomeAnalysisTK-2.3-9-ge5ebf34/GenomeAnalysisTK.jar \
-T UnifiedGenotyper \
-R /medpop/mpg-psrl/Parabase/hg19/ucsc.hg19.fasta \
-I /medpop/mpg-psrl/Parabase/Projects/final/final.list \
-o /medpop/mpg-psrl/Parabase/Projects/final/S284.snp.small.vcf \
-stand_call_conf 50.0 \
-stand_emit_conf 10.0 \
--dbsnp /medpop/mpg-psrl/Parabase/hg19/dbsnp_137.hg19.vcf \
-l INFO \
-A DepthOfCoverage \
-A HaplotypeScore \
-A InbreedingCoeff \
-A LowMQ \
-glm SNP \
-dcov 200 \
-nt 4 \
-nct 3 \



  • CarneiroCarneiro Charlestown, MAMember

    are they passing filters? your emit_conf is pretty low, so you'll probably see a lot of variants between 10-50 that will be emitted but not pass your Q50 threshold.

    what is your question about the -out_mode?

  • Thanks; I will try a higher value for emit_conf,

    Regarding -out_mode, I expect to see less variants by setting it to EMIT_ALL_CONFIDENT_SITES, than EMIT_VARIANTS_ONLY, but I got more ! can you explain a little bit the differences ?!

    Thank you

  • CarneiroCarneiro Charlestown, MAMember

    emit all confident sites will emit all sites including reference that pass your Q10 confidence threshold for emitting.

Sign In or Register to comment.