[maker-devel] maker dying on a contig

Fri Aug 31 16:48:37 MDT 2012

Carson, maker completed on the scaffold without error when I used your
patch. Thanks!

- J

On Fri, Aug 31, 2012 at 3:47 PM, Carson Holt <carsonhh at gmail.com> wrote:
> Looks like there is a slight format variation in one line of the
> RepeatMasker outputs.  Not sure what could have caused it -->
> 249 30.8 1.9 0.9  scaffold_10105  22320 22372 (34998) + (TCCA)n
> Simple_repeat          0    109    (0)
>
> The position/compliment information in columns 12-14 give a start position
> of 0 for part of the alignment.  A 0 is invalid because repeat masker
> doesn't use 0 based coordinates.  This causes a bug downstream when MAKER
> asks for the starting position of the alignment in another part of the
> code.
>
> This may be a RepeatMasker or rmblast bug.  When I replace rmblast with
> wu-blast and then reconfigure RepeatMasker I get this line -->
> 249   30.8  1.9  0.9  scaffold_10105  22320 22372 (34998) + (TCCA)n
> Simple_repeat          1    109    (0)
>
> Notice the 0 becomes a 1 and all logic is restored to the world.
>
> From my end the fix is simple.  If I see a 0 when reading RepeatMaskers
> output, then I just turn it into a 1.
> I've attached a file that you should use to replace the
> ./maker/lib/Widget/RepeatMasker.pm file.  It will fix the failure issue.
> I'll also make a note to the RepeatMasker developers.
>
> Thanks,
> Carson
>
>
>
> On 12-08-25 12:21 PM, "Jeremy Semeiks" <jeremy.semeiks at utsw.edu> wrote:
>
>>Attached.
>>
>>Thanks,
>>J
>>
>>On Fri, Aug 24, 2012 at 5:23 PM, Carson Holt <carsonhh at gmail.com> wrote:
>>> Could you send me the entire directory for just the failed contig?
>>>
>>> Thanks,
>>> Carson
>>>
>>>
>>> On 12-08-24 5:09 PM, "Jeremy Semeiks" <jeremy.semeiks at utsw.edu> wrote:
>>>
>>>>After doing this, I still get the same errors as I attached before.
>>>>There was also nothing in the log supporting that maker tried this
>>>>more than once. (Initially, I had tries=2, but I changed it to
>>>>tries=10 for this run. I'm guessing that's what you mean by "retry
>>>>count".) It also doesn't matter if I keep the same old contig
>>>>directory and master log before rerunning, or if I delete them both.
>>>>
>>>>FWIW, maker's own version of blast is 2.2.26+, which is the same as my
>>>>native version.
>>>>
>>>>Thanks,
>>>>Jeremy
>>>>
>>>>On Fri, Aug 24, 2012 at 2:17 PM, Carson Holt <carsonhh at gmail.com> wrote:
>>>>> Let maker install it's own blast (a version I know should work).
>>>>>
>>>>> cd ./maker/src
>>>>> ./Build blast
>>>>>
>>>>> Then to finish up your job (this command will just update executable
>>>>> locations in the maker_exe.ctl file) -->
>>>>> maker -EXE
>>>>>
>>>>> Then run maker again with a higher retry count.  Let me know if the
>>>>>contig
>>>>> still fails.
>>>>>
>>>>> Thanks,
>>>>> Carson
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On 12-08-24 3:13 PM, "Jeremy Semeiks" <jeremy.semeiks at utsw.edu> wrote:
>>>>>
>>>>>>Carson,
>>>>>>
>>>>>>$ blastx -version
>>>>>>blastx: 2.2.26+
>>>>>>Package: blast 2.2.26, build Feb 22 2012 16:04:28
>>>>>>
>>>>>>- J
>>>>>>
>>>>>>On Fri, Aug 24, 2012 at 7:24 AM, Carson Holt <carsonhh at gmail.com>
>>>>>>wrote:
>>>>>>> Could you send the blast version informations for reference
>>>>>>>purposes?
>>>>>>>I
>>>>>>> can see this being caused by BLAST.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Carson
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On 12-08-24 3:44 AM, "Michael Thon" <mike.thon at gmail.com> wrote:
>>>>>>>
>>>>>>>>I'm not sure if this is relevant to your problem but I had problems
>>>>>>>>with
>>>>>>>>maker failing on some contigs but not others.  I reinstalled maker
>>>>>>>>and
>>>>>>>>used the Build script to have it install all of its own
>>>>>>>>dependencies.
>>>>>>>>Now maker is running fine, although the culprit seems to have been
>>>>>>>>the
>>>>>>>>blast version I was using before.
>>>>>>>>
>>>>>>>>
>>>>>>>>On Aug 21, 2012, at 6:26 PM, Jeremy Semeiks
>>>>>>>><jeremy.semeiks at utsw.edu>
>>>>>>>>wrote:
>>>>>>>>
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> I'm trying to annotate a mammalian genome with maker-2.26-beta. I
>>>>>>>>>have
>>>>>>>>> ~36,000 contigs, and per the master log maker FINISHED all of them
>>>>>>>>> except one (scaffold_10105, ~60 kbp, not obviously weird) on the
>>>>>>>>>first
>>>>>>>>> run. It didn't report ERROR for that contig in the master log, but
>>>>>>>>>it
>>>>>>>>> did output two DIED lines in the contig's run.log and then just
>>>>>>>>>hung
>>>>>>>>> for a week until I killed it.
>>>>>>>>>
>>>>>>>>> I reran maker on just that contig by deleting the START line from
>>>>>>>>>the
>>>>>>>>> master log. maker again died on that contig.
>>>>>>>>>
>>>>>>>>> I attach the relevant part of the stderr stream; the rest was just
>>>>>>>>> "--Next Contig--"-type lines. Any ideas?
>>>>>>>>>
>>>>>>>>> A related question. Is deleting lines in the master log, as I did
>>>>>>>>> here, the best way to get maker to rerun on just a few contigs? I
>>>>>>>>> often find myself needing to do this, sometimes because of ERROR
>>>>>>>>>lines
>>>>>>>>> or other assumed quirks in maker and sometimes just because I
>>>>>>>>>needed
>>>>>>>>> to pause a big run for a while. One disadvantage is that when I
>>>>>>>>> restart, maker needs to manually scan over many completed contigs,
>>>>>>>>> which takes a while.
>>>>>>>>>
>>>>>>>>> (I'd think the best solution would be for maker to automatically
>>>>>>>>> recognize whether it's really currently running on contigs marked
>>>>>>>>> STARTED but not FINISHED. It doesn't seem to do this now, but I
>>>>>>>>> realize that might be hard to implement.)
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Jeremy
>>>>>>>>> Grad student, Grishin lab
>>>>>>>>> UT Southwestern, Dallas TX
>>>>>>>>> 510.384.8959
>>>>>>>>> <maker.stderr>_______________________________________________
>>>>>>>>> maker-devel mailing list
>>>>>>>>> maker-devel at box290.bluehost.com
>>>>>>>>>
>>>>>>>>>http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab
>>>>>>>>>.o
>>>>>>>>>rg
>>>>>>>>
>>>>>>>>
>>>>>>>>_______________________________________________
>>>>>>>>maker-devel mailing list
>>>>>>>>maker-devel at box290.bluehost.com
>>>>>>>>http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.
>>>>>>>>or
>>>>>>>>g
>>>>>>>
>>>>>>>
>>>>>
>>>>>
>>>
>>>
>