What are restrictions on interval creation?

inkink Posts: 4Member

GATK Queue implements a Scatter/Gather algorithm to create a set of intervals in order to parallelise data alalysis. If -scatter_gather option is issued, a respective number of interval files will be created and the input BAM files will be processed using these intervals. However a question arises what happens if the point of subsequent analysis is at or near the start/end of an interval? Are all the codes which support -l/-L options robust in the respect to interval positions? Since the input files are not actually spliced, all information is available to the processing program which could make the right decisions so that no artefacts are produced.
Are there any restrictions on interval creation? Perhaps it should be at least a few read lengths. Anything else?
Thanks in advance!



  • Geraldine_VdAuweraGeraldine_VdAuwera Posts: 8,428Administrator, GATK Dev admin

    Hi there,

    I'm not sure exactly what you're asking here, but basically the program is designed to handle "edge of intervals" locations appropriately. There are no restrictions on what you can use as intervals.

    Geraldine Van der Auwera, PhD

  • pdexheimerpdexheimer Posts: 501Member, GATK Dev, DSDE Dev mod

    Note that the restrictions on what the intervals can look like vary depending on the walker used. For instance, UnifiedGenotyper looks at each genomic position independently, while IndelRealigner must consider each read as a unit. This article describes some of the partitioning options and how they're controlled

  • inkink Posts: 4Member
    edited February 2013

    Thank you both. Yes, I am asking about handling of edge of intervals. I've seen the article but not sure how it relates to the default scatter/gather performed by GATK/Queue.

    Post edited by ink on
  • Geraldine_VdAuweraGeraldine_VdAuwera Posts: 8,428Administrator, GATK Dev admin

    Actually these differences are very likely due to downsampling. When you use different intervals, the seeds used for downsampling are different, and that can cause minor differences here and there that affect borderline calls.

    Geraldine Van der Auwera, PhD

Sign In or Register to comment.