search

broadinstitute / gatk

Official code repository for GATK versions 4 and up
https://software.broadinstitute.org/gatk
Other
1.69k stars 588 forks source link

GenotypeGVCFs fails with NullPointerException #3210

Closed eitanbanks closed 7 years ago

eitanbanks commented 7 years ago

In a joint calling run with 11,000 samples, and broken up into over 10,000 scatters, a single one failed with a NPE. I was able to get around it for now by just ignoring that scatter for the output, but that's really not an ideal thing to do for joint calling (and we cannot do that for the CCDG callset). I can't give you the inputs because it was running on so many samples (and via GenomicsDB), but hopefully the stacktrace will help here:

java.lang.NullPointerException at org.broadinstitute.hellbender.tools.walkers.genotyper.AlleleSubsettingUtils.calculateLikelihoodSums(AlleleSubsettingUtils.java:234) at org.broadinstitute.hellbender.tools.walkers.genotyper.AlleleSubsettingUtils.calculateMostLikelyAlleles(AlleleSubsettingUtils.java:199) at org.broadinstitute.hellbender.tools.walkers.genotyper.GenotypingEngine.calculateGenotypes(GenotypingEngine.java:241) at org.broadinstitute.hellbender.tools.walkers.genotyper.GenotypingEngine.calculateGenotypes(GenotypingEngine.java:205) at org.broadinstitute.hellbender.tools.walkers.GenotypeGVCFs.calculateGenotypes(GenotypeGVCFs.java:276) at org.broadinstitute.hellbender.tools.walkers.GenotypeGVCFs.regenotypeVC(GenotypeGVCFs.java:234) at org.broadinstitute.hellbender.tools.walkers.GenotypeGVCFs.apply(GenotypeGVCFs.java:213)

I'm not sure who now owns this code, so will ping @davidbenjamin, @ldgauthier, @droazen.

sooheelee commented 7 years ago

Was the NPE for a sample or shard?

Sent from an iPhone and typed with my thumbs.

On Jul 3, 2017, at 8:31 AM, Eric Banks notifications@github.com wrote:

In a joint calling run with 11,000 samples, and broken up into over 10,000 scatters, a single one failed with a NPE. I was able to get around it for now by just ignoring that scatter for the output, but that's really not an ideal thing to do for joint calling (and we cannot do that for the CCDG callset). I can't give you the inputs because it was running on so many samples (and via GenomicsDB), but hopefully the stacktrace will help here:

java.lang.NullPointerException at org.broadinstitute.hellbender.tools.walkers.genotyper.AlleleSubsettingUtils.calculateLikelihoodSums(AlleleSubsettingUtils.java:234) at org.broadinstitute.hellbender.tools.walkers.genotyper.AlleleSubsettingUtils.calculateMostLikelyAlleles(AlleleSubsettingUtils.java:199) at org.broadinstitute.hellbender.tools.walkers.genotyper.GenotypingEngine.calculateGenotypes(GenotypingEngine.java:241) at org.broadinstitute.hellbender.tools.walkers.genotyper.GenotypingEngine.calculateGenotypes(GenotypingEngine.java:205) at org.broadinstitute.hellbender.tools.walkers.GenotypeGVCFs.calculateGenotypes(GenotypeGVCFs.java:276) at org.broadinstitute.hellbender.tools.walkers.GenotypeGVCFs.regenotypeVC(GenotypeGVCFs.java:234) at org.broadinstitute.hellbender.tools.walkers.GenotypeGVCFs.apply(GenotypeGVCFs.java:213)

I'm not sure who now owns this code, so will ping @davidbenjamin, @ldgauthier, @droazen.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

eitanbanks commented 7 years ago

This tool works on a shard of many samples. You can tell from the stacktrace where in the code it choked.

ronlevine commented 7 years ago

After looking at the code, if the genotype does not have likelihoods, they are set to null and will throw an NPE.