Why are dbSNP indels excluded from indel realignment?

igorigor New YorkMember ✭✭

According to the known variants documentation (https://software.broadinstitute.org/gatk/guide/article?id=1247), only Mills indels and 1KG indels are recommended for indel realignment (RealignerTargetCreator and IndelRealigner). Although dbSNP contains indels (including 1KG indels) and is also in the bundle, it is not recommended. Why not? Is there a problem with using too many known indels or that may contain too many false positives?

Best Answer

  • SheilaSheila Broad Institute admin
    Accepted Answer

    @igor
    Hi.

    I think the simple answer is that during testing, we found that using dbSNP does not make much of a difference in Indel Realignment. There is no problem using too many indels or too many false positives, because the tool will only choose intervals that really show evidence for needing realignment.

    -Sheila

    P.S. Have a look at this post which announces that we are deprecating Indel Realignment as a pre-processing step. However, you still should use it if you are using a position based variant caller, and not a realignment based-caller.

Answers

  • SheilaSheila Broad InstituteMember, Broadie, Moderator admin
    Accepted Answer

    @igor
    Hi.

    I think the simple answer is that during testing, we found that using dbSNP does not make much of a difference in Indel Realignment. There is no problem using too many indels or too many false positives, because the tool will only choose intervals that really show evidence for needing realignment.

    -Sheila

    P.S. Have a look at this post which announces that we are deprecating Indel Realignment as a pre-processing step. However, you still should use it if you are using a position based variant caller, and not a realignment based-caller.

Sign In or Register to comment.