No announcement yet.

"allele balance ratio" and "quality by depth" in VCF files

  • Filter
  • Time
  • Show
Clear All
new posts

  • "allele balance ratio" and "quality by depth" in VCF files

    In a recent paper entitled "Exome sequencing as a tool for Mendelian disease gene discovery", the authors write:

    "The analysis pipeline consisted of: alignment using the Burrows–Wheeler alignment tool; recalibration; realignment around insertion–deletions and merging with the Genome Analysis Toolkit (GATK)91; and removal of duplicates with PICARD. Variants were called using the following parameters: quality score > 50, allele balance ratio < 0.75; homopolymer run > 3; and quality by depth < 8."

    I have three questions:

    1. Once the removal of duplicates step/PICARD step is done, you will be left with a bam (or sam) file. So what subsequent steps are taken in order to find the four numbers that they're filtering on?

    2. If I run things through the GATK pipeline up through the UnifiedGenotyper step, I end up with a vcf file. This site describes elements in that file:;s_VCF_files

    Is "allele balance ratio" the same as "##INFO=<ID=AF,Number=.,Type=Float,Description="Allele Frequency, for each ALT allele, in the same order as listed">"?

    3. The only mention I see of "quality by depth" is here:

    ##INFO=<ID=QD,Number=1,Type=Float,Description="Variant Confidence/Quality by Depth">

    I assume that "Variant Confidence" is the same as "QUAL", in which case I could calculate "Quality by Depth", but this seems wrong because they want LOW "quality by depth" scores. Do they instead mean this QD ratio when they say "quality by depth"?

    Thank you.


  • #2
    Looks like old hard filtering parameters for the VariantFiltration walker, but are some of them backwards? See e.g:

    (note that the FilterExpression filters calls OUT if true).

    GATK and its filtering recommendations have evolved since then, and AB is no longer in the output by default, see e.g.:

    In the most recent pipeline, an entirely different approach to filtering is recommended (VQSR):


    • #3
      Hi RDW,

      Thanks so much for the help. This makes sense and I very much appreciate the links.

      Best wishes,