get_binomial_gte_n_success_exact_pval error

[igordot]

I've been using manta for a while (mostly for exomes) without any issues. I tried running it on a small (~100kb) targeted panel. For some reason, one of my samples is giving this error:

[2017-10-15T23:04:35.352315] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] Failed to complete command task: 'generateCandidateSV_0233' launched from master workflow, error code: 1, command: '/software/manta/manta-1.2.1/libexec/GenerateSVCandidates --align-stats /proj_path/logs-snvs-strelka/4-33N-manta/workspace/alignmentStats.xml --graph-file /proj_path/logs-snvs-strelka/4-33N-manta/workspace/svLocusGraph.bin --bin-index 233 --bin-count 256 --min-candidate-sv-size 8 --min-candidate-spanning-count 3 --min-scored-sv-size 51 --ref /ref/hg19/genome.fa --candidate-output-file /proj_path/logs-snvs-strelka/4-33N-manta/workspace/svHyGen/candidateSV.0233.vcf --diploid-output-file /proj_path/logs-snvs-strelka/4-33N-manta/workspace/svHyGen/diploidSV.0233.vcf --min-qual-score 10 --min-pass-qual-score 20 --min-pass-gt-score 15 --somatic-output-file /proj_path/logs-snvs-strelka/4-33N-manta/workspace/svHyGen/somaticSV.0233.vcf --min-somatic-score 10 --min-pass-somatic-score 30 --skip-remote-reads --edge-runtime-log /proj_path/logs-snvs-strelka/4-33N-manta/workspace/svHyGen/edgeRuntimeLog.0233.txt --edge-stats-log /proj_path/logs-snvs-strelka/4-33N-manta/workspace/svHyGen/edgeStats.0233.xml --align-file /proj_path/BAM-GATK-RA-RC/33N.dd.ra.rc.bam --tumor-align-file /proj_path/BAM-GATK-RA-RC/4.dd.ra.rc.bam'
[2017-10-15T23:04:35.353860] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] Error Message:
[2017-10-15T23:04:35.355448] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] Anomalous task wrapper stderr output. Wrapper signal file: '/proj_path/logs-snvs-strelka/4-33N-manta/workspace/pyflow.data/logs/tmp/taskWrapperLogs/000/517/pyflowTaskWrapper.signal.txt'
[2017-10-15T23:04:35.356675] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] Logging 6 line(s) of task wrapper log output below:
[2017-10-15T23:04:35.357967] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] [taskWrapper-stderr] [2017-10-15T23:04:30.787878] [node011.cm.cluster] [11287_1] [pyflowTaskWrapper:generateCandidateSV_0233] [wrapperSignal] wrapperStart
[2017-10-15T23:04:35.359465] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] [taskWrapper-stderr] [2017-10-15T23:04:30.805218] [node011.cm.cluster] [11287_1] [pyflowTaskWrapper:generateCandidateSV_0233] [wrapperSignal] taskStart
[2017-10-15T23:04:35.360758] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] [taskWrapper-stderr] [2017-10-15T23:04:31.790544] [node011.cm.cluster] [11287_1] [pyflowTaskWrapper:generateCandidateSV_0233] [wrapperSignal] taskExitCode -6
[2017-10-15T23:04:35.362840] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] [taskWrapper-stderr] [2017-10-15T23:04:31.798555] [node011.cm.cluster] [11287_1] [pyflowTaskWrapper:generateCandidateSV_0233] [wrapperSignal] taskStderrTail 2
[2017-10-15T23:04:35.364987] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] [taskWrapper-stderr] Last 1 stderr lines from task (of 1 total lines):
[2017-10-15T23:04:35.366435] [node011.cm.cluster] [11287_1] [WorkflowRunner] [ERROR] [generateCandidateSV_0233] [taskWrapper-stderr] [2017-10-15T23:04:31.170409] [node011.cm.cluster] [11287_1] [generateCandidateSV_0233] GenerateSVCandidates: /builder/src/c++/lib/blt_util/binomial_test.cpp:161: double get_binomial_gte_n_success_exact_pval(double, unsigned int, unsigned int): Assertion `(p >= 0.) && (p <= 1.)' failed.

The only difference (as far as I can tell) between this sample and the others that work is that the coverage is lower. This is a somatic workflow and the average coverage for the N is ~80X. Although lower than others, it seems acceptable.

Do you know what may be causing this error?

Also, I understand it's probably pointless running manta on such a small panel, but I am using manta to generate indels for strelka.

[ctsa]

Thank you for reporting this, it is almost certainly a bug. Is this data you can share?

Also can describe (1) the manta version number and (2) whether this is our binary release or a version you've compiled?

[igordot]

This is the latest binary (manta-1.2.1.centos6_x86_64.tar.bz2).

I am posting the BAMs and the full error log (temporarily): https://www.dropbox.com/s/h8rpxjf64rj62fy/manta.zip?dl=0

[ctsa]

Copied bams and reproduced issue here, thank you. Will update once we get to the bottom of this.

[igordot]

Good to hear it's not just an issue for me. Hopefully you can resolve it.

[ctsa]

Update:

We've found the issue and are testing a solution. The alignments used in your example have some unusual properties which exposed a latent bug in manta. I want to highlight a few points just to check that you're expecting this signature in your alignments -- taking as an example 33-N-IA33.dd.ra.rc.bam we observe:

1. Excluding secondary/supplemental reads, ~22% of reads are split (24988/113396)
2. The count of secondary/supplemental split segments (567691) greatly outnumbers the number of primary split reads (24988)

[igordot]

Thank you for the update. Good to hear you found the source of the problem.

These were degraded FFPE samples with low DNA, so quality issues are not surprising. For example, there were ~90% duplicates (removed at a previous stage) for the sample you mention. Of course, that is easier to explain than the secondary reads. Do you know if there is a good explanation for those? Should I be worried?

I had a few other sample pairs in the same batch and they all ran without causing errors, so the pattern must be at least partially due to the actual library rather than some analysis step.

[ctsa]

The fix to stability and estimation of chimeric read fraction required for the data discussed in this ticket appear in MANTA-1261 in today's v1.3.0 release here:

https://github.com/Illumina/manta/releases/tag/v1.3.0