We've moved!
This site is now read-only. You can find our new documentation site and support forum for posting questions here.
Be sure to read our welcome blog!

artifact_in_normal (NALOD) negative values


I'm observing a lot of variants with a NALOD negative values in a matched normal exome samples with Mutect2/GATK4.1 e.g. :

1 78438378 . T A . artifact_in_normal CONTQ=93;DP=144;ECNT=1;GERMQ=150;MBQ=30,32;MFRL=176,145;MMQ=25,45;MPOS=64;NALOD=-9.977e+00;NLOD=5.18;POPAF=1.31;SAAF=0.030,0.040,0.045;SAPP=8.620e-03,8.786e-03,0.983;TLOD=6.99 GT:AD:AF:DP:F1R2:F2R1 0/0:73,4:0.063:77:35,1:38,3 0/1:64,3:0.058:67:33,3:31,0

However, I thought the artifact_in_normal was called when normal-artifact-lod was > 0.0 (default threshold).

Thank you for your help on this.



Best Answer


  • AdelaideRAdelaideR Member admin

    Hi Thomas - The log of the odds ratio that a artifact is normal can be negative if the ratio is a fraction.

    Mutect2 works primarily by contrasting the presence or absence of evidence for variation between two samples, the tumor and matched normal, from the same individual. The tool can run on unmatched tumors but this produces high rates of false positives. Technically speaking, somatic variants are both (i) different from the control sample and (ii) different from the reference. What this means is that if a site is variant in the control but in the somatic sample reverts to the reference allele, then it is not a somatic variant.

    You can read more about how Mutect2 is making these contrasts here.

    It is also helpful to remember that:

    This snippet comes from the mathematical deep dive into the Mutect2 method found here

    Another helpful blog post can be found that explains how Mutect2 marginalizes over allele fractions using a Bayesian likelihoods model can be found here.

  • davidbendavidben BostonMember, Broadie, Dev ✭✭✭

    @ThomasD Just want to add that we recently changed the definition of NALOD so that, similar to NLOD negative values mean more evidence in the normal. The reason for this is that the external libraries that we depend on to format VCF fields give nice-looking output for positive values and cumbersome engineering notation such as -9.977e+00 for negative values. We flipped the sign so that the rarer case gets the clumsier output as part of an attempt to make Mutect2 vcfs more readable.

  • ThomasDThomasD Member
    Hi @davidben and @AdelaideR,

    Thank you to both of you for your answers!

    Actually, I think I misunderstood the way the "artifact_in_normal" FILTER was applied because in the documentation for the FilterMutectCalls tool (https://software.broadinstitute.org/gatk/documentation/tooldocs/ , it is written:
    "If the normal artifact log odds is larger than the threshold (e.g. 0.0 ?), then FilterMutectCalls applies the artifact-in-normal filter".

    But now, if I understand correctly, positive values are flipped because of formatting purpose and negative values thus means "more evidence in the normal".
    Although this now explains most of the "artifact_in_normal" FITLER in my .vcf, I do still see few variants with this FILTER flag having positive NALOD values e.g. (NALOD=0.510):
    9 57849155 . GAGAAA G . artifact_in_normal CONTQ=93;DP=86;ECNT=1;GERMQ=72;MBQ=32,32;MFRL=184,216;MMQ=60,60;MPOS=38;NALOD=0.510;NLOD=4.59;POPAF=1.52;SAAF=0.030,0.071,0.075;SAPP=0.014,0.017,0.969;TLOD=7.62 GT:AD:AF:DP:F1R2:F2R1 0/0:33,4:0.090:37:19,3:14,1 0/1:37,3:0.096:40:17,1:20,2

    Is there any other reasons explaining the "artifcat_in_normal" FILTER call?

    Thanks again your help


Sign In or Register to comment.