#168 BOG assertion failure during splitBadMates

Crash (103)

In my latest assembly with an up-to-date CVS version of wgs, BOG throws an assertion error. Here are the relevant error logs:

checkUnitigMembership()-- nutg=139429302 nfrg=392491466 lost=0 found=392491466
MATE HAPPINESS (dove/dove): unmated 9051040 mated 21198423 sameTig: happy 6798556 grumpy 79778 diffTig: end-end 12223002 end-int 2071101 int-int 25986
MATE HAPPINESS (dove/cont): unmated 22737408 mated 79334552 sameTig: happy 18716011 grumpy 0 diffTig: end-end 48045113 end-int 12182754 int-int 390674
MATE HAPPINESS (cont/cont): unmated 0 mated 73783289 sameTig: happy 25824245 grumpy 0 diffTig: end-end 34893619 end-int 11708829 int-int 1356596

SPLIT ERROR: unitig 140632570 frag 108215069 373,472 missed the split point 274,298.

buildUnitigs: AS_BOG_MateChecker.cc:314: UnitigBreakPoints* UnitigGraph::computeMateCoverage(Unitig*, int): Assertion `loc.bgn <= bad.end+1 || loc.end <= bad.end+1' failed.


  • Jason Miller

    Jason Miller - 2011-06-01
    • labels: --> Crash
    • milestone: --> unitigger
    • assigned_to: nobody --> jasonmiller9704
  • Jason Miller

    Jason Miller - 2011-06-01

    Could you provide some more background? What kind of data was input? If Illumina, was it paired end or mate pair? How much coverage? How were the reads treated initially? That is, what were the fastqToCA or sffToCA options? And how were the reads trimmed? That is, what where the options like doOBT on runCA? Send the runCA spec file if possible. Thanks.

  • Robert Reid

    Robert Reid - 2011-09-12

    I am getting the same error.

    buildUnitigs: AS_BOG_MateChecker.cc:1046: UnitigBreakPoints* MateChecker::computeMateCoverage(Unitig*, BestOverlapGraph*, int): Assertion `loc.bgn <= bad.end+1 || loc.end <= bad.end+1' failed.

    The dataset being added is a collection of paired 454 reads, an Illumina GAII dataset and a large Hiseq dataset.

    The last file successfully created was the breaks.ovl file.

    Could this be caused by short reads in the frg file ?

    The 454 data was 1st filtered and then converted to frg using a matepaire file and Celera's fasta-to-frg script.
    The Illumina data was quality filtered, and trimmed for adapter removal resulting in MANY smaller sized fragments less than 74BP.

    How would one utilize a set of hiseq reads that are ~70 bp long?

  • Jason Miller

    Jason Miller - 2011-09-12

    With high coverage in 70bp Illumina reads, you are at the brink of what CA can support. The minimum read length after trimming is 64bp, and the minimum overlap is 40bp. You can change those parameters in source code at your own risk.

    Your run stopped while BOG checks mate happiness in unitigs. We are aware that BOG's use of mates to confirm unitigs is not ideal for Illumina data. We could try to exchange data and fix the code, but here are some less painful suggestions. With BOG, set the minimum happy mate depth very high (it is 7 by default) or turn off mate-based splitting altogether. Alternately, try unitigger=bogart in the latest code on CVS. That module does not use mates for splitting.

    If either Illumina data set is mate pair (not paired end), then you will want to try the ClassifyMates (DNC) module that we are documenting now. Stay tuned.

  • Robert Reid

    Robert Reid - 2011-11-16

    If I skp the Hiseqs and use just the 454 data, Celera runs flawlessly.

    Adding a highly filtered Hiseq dataset (mated pairs all of which are 100 BP), I still get bogged down at this step (See the pun ?). I tried upping the bogBadMateDepth but the error remains.

    Going to try latest code in CVS........

  • Jason Miller

    Jason Miller - 2011-11-21

    Since you are using 454+Illumina data and the CVS tip, try running with the new unitigger=bogart option.

  • Brian Walenz

    Brian Walenz - 2013-02-22

    A work around is now in place. The problem persists, but at least we can handle it now.

  • Brian Walenz

    Brian Walenz - 2013-02-22
    • status: open --> closed

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:

No, thanks