Attention:
The frontline support team will be offline as we are occupied with the GATK Workshop on March 21st and 22nd 2019. We will be back and available to answer questions on the forum on March 25th 2019.

Filtering out M2 Mutect Calls made in regions where only 1 type of read orientation is dominant

Hello,

I have applied filter criteria based on OxoG as discussed in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3616734/. However, I have regions that are present consistently across samples (and cancers) that have reads either in the forward direction or reverse. And also these transversions are different from what is discussed in the above paper (for transversions, C>A/G>T ).

How do I go about informatically removing these positions from my call set?

Thank you for your input!

~MW

PS, My dataset comes from WES T/N data.

Issue · Github
by Sheila

Issue Number
1633
State
closed
Last Updated
Assignee
Array
Milestone
Array
Closed By
vdauwera

Answers

  • SheilaSheila Broad InstituteMember, Broadie, Moderator admin

    @mikawin
    Hi MW,

    I am checking with the team. We will get back to you asap.

    -Sheila

  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie admin

    Hi @mikawin,

    Sorry for the somewhat late reply; the wording of your question had us confused, thinking about OxoG and somatic-specific sequence properties, for which as I've mentioned before we aren't currently able to provide support. But after further review it sounds like what you're really asking for is some way to detect strand bias -- is that right? In the germline world you'd want to use the StrandBias (SB), FisherStrand (FS) and StrandOddsRatio (SOR) annotations to weed out these artifacts. I'm not sure what you get annotated by default with MuTect (which one are you using, btw?) but you could always re-annotate your callset using VariantAnnotator to get them, then filter using VariantFiltration as explained in the germline filtering documentation. Like I said I can't guarantee that's the right thing to do for somatic data but it's the best we can offer right now.

Sign In or Register to comment.