Bug Bulletin: The GenomeLocPArser error in SplitNCigarReads has been fixed; if you encounter it, use the latest nightly build.

Scatter/Gather in Queue with UG in indel-mode introduces cycles in graph

Johan_DahlbergJohan_Dahlberg Posts: 85Member ✭✭✭

I've noticed some strange behavior from Queue where in some cases, when I scatter/gather the Unified Genotyper in indel-mode it will introduce Cycles in the graph. This causes to Queue to die with a StackOverflowError which seems to be caused by the graphDepth function in QGraph due to the recursion becoming unbounded. This cause me some headaches yesterday as I tried to figure out how to make the function tail-recursive, before noticing the message: ERROR 17:18:21,292 QGraph - Cycles were detected in the graph this morning.

This leads me to one request and one question. First the request: It would be nice if Queue would exit if the graph validation fails, as it would make identifying the source of the problem simpler. It this possible?

Secondly the question: do you have any ideas as to what might cause the cycles?

I have tried looking at the graphviz files and I cannot identify any cycles from those (though when looking at the s/g-plots it's really difficult to make any sense of it).

My code looks like this:

val candidateSnps = new File(outputDir + "/" + projectName + ".candidate.snp.vcf")
val candidateIndels = new File(outputDir + "/" + projectName + ".candidate.indel.vcf")

// SNP and INDEL Calls
add(snpCall(cohortList, candidateSnps))
add(indelCall(cohortList, candidateIndels))

val targets = new File(outputDir + "/" + projectName + ".targets")
add(target(candidateIndels, targets))

// Take regions based on indels called in previous step
val postCleaningBamList =
  for (bam <- cohortList) yield {
    val indelRealignedBam = swapExt(bam, ".bam", ".clean.bam")
    add(clean(Seq(bam), targets, indelRealignedBam))
    indelRealignedBam
  }

val afterCleanupSnps = swapExt(candidateSnps, ".candidate.snp.vcf", ".cleaned.snp.vcf")
val afterCleanupIndels = swapExt(candidateIndels, ".candidate.indel.vcf", ".cleaned.indel.vcf")

// Call snps/indels again
add(snpCall(postCleaningBamList, afterCleanupSnps))
add(indelCall(postCleaningBamList, afterCleanupIndels))

Where the cohortList is a Seq[File].

Right now I've solved this by setting this.scatterCount = 1 in the indelCall case class, however this doesn't feel quite satisfactory to me, so any pointers for a more robust solution would be greatly appreciated.

Answers

Sign In or Register to comment.