Apply VQSR to variants called by other tools

Hi, I have several variant files that were generated by other calling tools with some annotations not defined by GATK. I wonder if I can also apply VQSR on these dataset restricting to those exclusive annotations (using "-an"). Is there any drawback for this application?

Besides, I saw somewhere that fitting only a single Gaussian distribution to each annotation. Is this a proper way to perform variant recalibration?

Tagged:

Best Answers

Answers

  • EmmaDanEmmaDan ChinaMember

    Hi @Geraldine_VdAuwera,
    Thank you very much for the quick response! Would you please name some annotations that are gaussianly distributions? What are the characteristics of them?

  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie

    The annotations that we recommend for VQSR are for the most part gaussianly distributed. That is their most relevant characteristic. Also, we know from past analyses that they are informative as to the quality of variant calls.

    What annotations are you considering to use?

  • tommycarstensentommycarstensen United KingdomMember

    @EmmaDan would it perhaps be an option for you to use VariantAnnotator or recall your variants in GGA mode with UG to avoid local realignment?

  • EmmaDanEmmaDan ChinaMember
    edited May 2015

    Hi @Geraldine_VdAuwera @tommycarstensen ,

    I am very sorry for the late response and thank you very much for your kind comment and suggestion!

    Actually, I currently have no idea which annotations to apply. What I am thinking is that, each calling tool may have different preference and bias towards variants. Therefore, datasets called by different tools may show particular distribution. I would like to train VQSR using dataset (assuming they are all true positive) called by, said SAMtools, and then use this to determine whether a variant (also called by SAMtools) is a false positive or not. Am I on the right track about this? Any suggestion what kind annotations should I use?

    Thank you!

    Emma

Sign In or Register to comment.