#120 bad.end crash persists with Illumina

unitigger
closed
Jason Miller
Crash (103)
9
2013-02-22
2010-08-22
Alexie
No

While running buildunitigs on datasets containing illumina (tried with two different insect genomes), bog crashes with the dreaded error

buildUnitigs: AS_BOG_MateChecker.cc:1066: UnitigBreakPoints* MateChecker::computeMateCoverage(Unitig*, BestOverlapGraph*, int): Assertion `loc.bgn <= bad.end+1 || loc.end <= bad.end+1' failed

but only when the 'error limit' is not 0 (tried with as low as 1.5). (This is with the CVS version checked out 2010-08-06).

Any ideas?
thanks
a

Discussion

  • Alexie
    Alexie
    2010-08-22

    p.s. all illumina has been quality trimmed with minimum length of 80 bp (to save cabog some time). The UTG error rates tried where 3 - 5 % (and I think it also crashes at the default 1.5 % but I can't remember). It does not crash at 1.5 % and 0 errorlimit.

    Is there a way to make it behave differently when it sees illumina data? E.g. it is better to discard the unitig rather than try to multialign a 1200 bp solexa unitig...

     
  • Alexie
    Alexie
    2010-08-22

    ps.2 kmer size is 18....

     
  • Jason Miller
    Jason Miller
    2010-08-23

    • priority: 5 --> 9
    • assigned_to: nobody --> jasonmiller9704
     
  • Jason Miller
    Jason Miller
    2010-08-23

    This is a serious issue. Several users have encountered it. We will work on it after the CAUG meeting, that is, in September. For faster relief, we suggest editing the code near the assert to get past the assert. You might change "assert (!condition)" to "if (condition) continue". Ultimately, the problem is due to high arrival rates. Using 100bp reads at 100X coverage, arrival rate is 1bp. Thus, another possible work-around is to sample down the Illumina reads to 50X-100X.

     
  • Hi Jason

    That's exactly what I did ;-) Now failed as it tries to split the unitig into sub-unitigs and thus asserts near line 398 of AS_CGW/SplitChunks_CGW.c i'm currently editing it to drop the affected unitig...

    the illumina is much lower than 50x, more like 20x, maybe less. Lets' chat about this after CAUG.
    cheers
    alexie

     
  • Brian Walenz
    Brian Walenz
    2013-02-22

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

    A work around is (finally) in place. The problem remains, however.