I have adapted my pipeline to the new filtering strategies implemented in gatk and while looking at its output I noticed that the stat file generated by mutect2 has a negative number:

callable -1.538687311E9

I did not find the documentation about this file content but in the gatk forum it is mentioned that is should represent the number of bases above a certain coverage (10 by default). This happened for a WGS sample, while for WES I obtain a more sensible value (well, at least positive and integer, even if maybe a tad high):

callable        8.6884303E7

Has this happened to anyone else? I compared the resulting calls obtained for the WGS with a previous version of mutect2 and they are comparable-ish (98% of the new "ones" are in common with the "old" ones), but I am still worried that something fishy is happening, expecially considering that the filtering step of the pipeline uses this stats file.

  • 1
    $\begingroup$ quick followup: seems to be a bug due to an integer overflow (yay Java!), GATK developers have fixed it and I'm trying it out. $\endgroup$
    – vodka
    Dec 6 '19 at 9:05

This was answered by the OP -- it was a bug that has been fixed in a subsequent release, See the release notes here: https://github.com/broadinstitute/gatk/releases/tag/

Made callable sites a Long to avoid integer overflow (#6303)

Issue ticket: https://github.com/broadinstitute/gatk/issues/6302

  • $\begingroup$ Hey ya, this could be a useful answer if you would provide some details, like the version where the bug is fixed and a link to the changelog. Right now it provides nothing than what OP wrote. $\endgroup$
    – Kamil S Jaron
    Apr 8 '20 at 10:13
  • $\begingroup$ Sure; done. I was just looking to not have it show up as unanswered anymore since the problem was effectively solved. But I get your point. $\endgroup$ Apr 8 '20 at 14:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.