Re: [wgs-assembler-users] CA 8.0 on large genome using PacBio

[Walenz, B. <bw...@jc...>]

Hi-

The obtStore.err is claiming that the inputs (*.ovb.gz) are not in gzip
format.  At this stage, all it is doing is 'gzip -l
0-overlaptrim-overlap/001/000462.ovb.gz' to see the true length of the
compressed file.  Gzip itself is complaining that the file is not gzip
format.  Are these files valid (non-empty)?  Were there any error reported
in the overlap job output (0-overlaptrim-overlap/*.err)?  You can also clean
up this directory (remove *.err and the 001/ directory) and run a few jobs
by hand (sh overlap.sh 1, sh overlap.sh 2, etc) to see that they run OK.


I don't see anything wrong in your spec, but would suggest some changes.

unitigger=bogart
utgGraphErrorRate=0.05
utgMergeErrorRate=0.05
batMemory=X (X in gigabytes)
batThreads=Y (default is to use all CPUs on the machine)

The bogart unitigger seems to work much better than 'bog', but it more
expensive to run.  It needs to load non-best overlaps into memory.  If you
leave this unset, it will default to using all physical memory on the
machine.

I increased the allowed error rate from 3% to 5%.  This should result in
better unitig construction, but rarely can end up declaring that similar
unique sequence is a repeat and breaking the unitig.

doFragmentCorrection=0

This is an expensive step, that can improve results.  For now, leave it off.

b



On 11/13/13 11:07 AM, "mic...@ip..."
<mic...@ip...> wrote:

> Dear WGS developers
> 
> I am trying to assemble a subset of 2 Gbp error corrected PacBio reads with
> CA8.0 using SGE to get a feeling about speed and  resources needed for the
> assembly. The full dataset will be about 13,8 Gbp of error-corrected PacBio
> reads. 
> Read lengths range from 790 up to 19'000 bp.
> I assume its curcial to have optimal parameter settings in the spec-file.
> Unfortunately, i cant find appropriate examples for my dataset.
> 
> Currently i am stuck with an error in the obt store generation:
> Failure message:
> 
> failed to build the obt store
> 
> Would it be possible provide a PacBio spec template and give me some hint
> about how to overcome the obt store failure?
> Attached are error-files from the runCA.sge.out , the std.err and the
> obtStore.err along with the used spec file.
> 
> Thank you very much,
> Michel
> 
> 
> 
> 
> ________________________________________
> Von: wgs...@li...
> [wgs...@li...]
> Gesendet: Mittwoch, 13. November 2013 16:31
> An: Moser, Michel (IPS)
> Betreff: Mailman privacy alert
> 
> An attempt was made to subscribe your address to the mailing list
> wgs...@li....  You are already subscribed to this
> mailing list.
> 
> Note that the list membership is not public, so it is possible that a bad
> person was trying to probe the list for its membership.  This would be a
> privacy violation if we let them do this, but we didn't.
> 
> If you submitted the subscription request and forgot that you were already
> subscribed to the list, then you can ignore this message.  If you suspect that
> an attempt is being made to covertly discover whether you are a member of this
> list, and you are worried about your privacy, then feel free to send a message
> to the list administrator at wgs...@li....