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!

Clarification of --normal-artifact-lod in FilterMutectCalls

The FilterMutectCalls tool doc says:

If the normal artifact log odds is larger than the threshold, then FilterMutectCalls applies the artifact-in-normal filter. For matched normal analyses with tumor contamination in the normal, consider increasing the normal-artifact-lod threshold.

This is what I understand:

The normal atrifact log odds is the threshold above which an artifactual site detected in the normal will be used to filter any variants at that site (assuming this is just a shared artifact). If there is tumor contamination of the normal, an apparent artifact in the normal may just represent tumor contamination. Hence, this threshold should be raised if we suspect tumor contamination of the normal.

Am I right?

Best Answer


  • jorgezjorgez Member

    Hello Sheila,

    I was wondering if you could please help me to understand a bit more the -normal-artifact-lod parameter in FilterMutectCalls. I am trying to explain below my understanding of this parameter, please let me know if I am off base with the interpretation.

    From the documentation I see it is a log of odds with a zero default, which I understand means that the probability of making a somatic call due to an artifact produced by normal contamination in the tumor is the same as if the tumor was not contaminated (of there was no such artifact). I guess the zero comes from log(0.5/0.5)=0.

    Then I if wanted to express that the probability of making a ‘false‘ somatic call due to normal contamination artifacts is three times larger compared to an non-contaminated tumor I would set this parameter to log(0.75/0.25)=0.47

    Any advice will be much appreciated

  • davidbendavidben BostonMember, Broadie, Dev ✭✭✭

    @jorgez You are expecting that the normal artifact lod is a measure of normal contamination in the tumor, which is totally reasonable. However, we mean something different. The normal artifact lod in M2 actually has nothing to do with the tumor. Rather, it is the result of applying the somatic genotyping model to the normal reads. In FilterMutectCalls we use this apparent evidence for somatic variation in the tumor as a proxy for the possibility of a false positive in the tumor.

    Note that as of recent releases there is no more -normal-artifact-lod threshold. Instead, FilterMutectCalls automatically learns the extent to which normal artifacts suggest tumor artifacts.

Sign In or Register to comment.