We've moved!
This site is now read-only. You can find our new documentation site and support forum for posting questions here.
Be sure to read our welcome blog!

positive and negative set in model training by variantRecalibrator


Thanks for develop this tool set and share with others with good supports, it really contains a lot of wonderful tools.

I just try to understand VQSR more into detail. If I give the resources (dbsnp, hap map and 1kg omni data) recommended by the best practice with default settings (which one is in training, which one is the true set ...), Does the positive set contain all variants which recorded in the resources having train=TRUE, but how does the tool select negative set? Does it order the variants from high to low by the QUAL value, and pick up the 5% from the bottom (if the percentBad = 0.05)? Will there be some overlap between positive set and negative set? And is there any quality filtration on the data, e.g. one date point is more than a standard deviation away from average...


Best Answer


  • ying_sheng_1ying_sheng_1 Member ✭✭

    Thanks Geraldine. I think you answer all my questions.

  • tanu06tanu06 CanadaMember


    I had a followup question on this, should I keep those SNPs for downstream analysis or should be removed.
    I have already removed the Fail snps from VQSR vcf file .
    Kindly suggest.

    Thank You,
    With Regards,

  • SheilaSheila Broad InstituteMember, Broadie ✭✭✭✭✭

    Hi Tanushree,

    Sorry for the delay. We were at a workshop. I am not sure what you mean? You should not need to remove any variants other than non-PASS variants after VQSR and determining the sensitivity.


Sign In or Register to comment.