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!

Joint calling cases/controls

Dear Team,

I have aprox.100cases and 100controls (WES). Would you rather run joint calling (according to GATK best practises) on cases and controls separately or jointly call cases and controls?
If I have a rare variant that is present in only few cases and not present in controls, will the variant be rescue in jointly called cases only? Will it also be rescue in cases and controls called together ?
What are the caveats of both methods (jointly/separately).




  • SheilaSheila Broad InstituteMember, Broadie admin

    Hi Andy,

    You should joint call the samples all together. Have a look at this thread and this article for more information.


  • andy1andy1 DMember

    Dear Sheila,

    Thank you for that. Can I just double check with you one thing. If I run cases and controls together I may lose some variants that are present in small proportion of cases or controls only, is that right?
    For example if I have 1 good quality variant that is present in 1/100 cases and not present in controls , will that be recorded in file when I joint call case and controls together (1/200)? Will that be recorded when I joint call cases only (1/100)?

    Thanks again

  • SheilaSheila Broad InstituteMember, Broadie admin

    Hi Andy,

    There is a slight chance that low quality singletons will be lost in the workflow. However, if those singletons have a high GQ, they should not be lost. For example, if you have a variant in one sample that has a GQ of 99, that variant should be present in the final VCF, even if no other sample has the variant.

    Have a look at this thread for more information and what you can do to rescue singletons. Please note, if you set the confidence threshold to 0, you will have more false positives along with rescuing singletons. It is up to you to determine what is important to you.


Sign In or Register to comment.