[maker-devel] Why are some complete gene predictions not present in the final results?

[Carson Holt]

I've released 2.28 on the website.  This is one of the bugs that was fixed.
It happens under a very specific set of circumstances.  You need to run
maker with the -a command line flag to get it to recalculate upstream
variables after upgrading.  Alternatively you can also just give maker your
old GFF3 file (make all other options blank exempt for the *_pass= options),
and maker will just rebuild it.

Thanks,
Carson


From:  Hung-Wei Hsu <ares711122 at gmail.com>
Date:  Monday, 20 May, 2013 9:19 PM
To:  Carson Holt <carsonhh at gmail.com>
Cc:  <maker-devel at yandell-lab.org>
Subject:  Re: [maker-devel] Why are some complete gene predictions not
present in the final results?

Thanks a lot for your helps.
Your suggestions will be greatly helpful for our analysis.

I've tried to add EST sequences to improve gene predictions.
The EST sequences I used were CDS sequences of the same organism.
But I got an error as below.

substr outside of string at .../TranslationMachine.pm line 162
ERROR: Failed while polishig ESTs
ERROR: Chunk failed at level:2, tier_type:3

What's wrong with my analysis?
The EST sequences I used are wrong?
Thank you.

Hung-Wei



2013/5/21 Carson Holt <carsonhh at gmail.com>
> On default settings MAKER will only put ab initio predictions that have some
> sort of evidence support (EST or protein) in the final gene set.  The rejected
> predictions are still in the GFF3 for reference purposes as match/match_part
> features, but not as gene/mRNA/exon/CDS features.  So a lack of evidence might
> be why it is not there.  You can add all rejected models that don't overlap an
> accepted model by setting keep_preds=1 (this usually brings a lot more into
> the final gene set than you really want though (lots of false positives).  But
> for some organisms like fungi, which have high gene densities, this approach
> is relatively safe.
> 
> Alternatively the gene is missing because it overlaps another gene model that
> was accepted.  MAKER won't allow overlapping models on the same strand in
> eukaryotes.  The only way to force that kind of overlap is to give MAKER the
> reference models in model_gff and not let it call it's own models (then maker
> is really just aligning evidence and scoring the reference models).
> 
> One final note.  If there is no evidence supporting the model, and that is why
> it is rejected, you can also try adding more evidence to the maker run or you
> can consider the possibility that the gene model in the reference is not real
> to being with (i.e. a false positive gene model called during the initial
> annotation process and not supported by protein or expression data from any
> source).
> 
> Thanks,
> Carson
> 
> 
> 
> From:  Hung-Wei Hsu <ares711122 at gmail.com>
> Date:  Monday, 20 May, 2013 12:16 AM
> To:  <maker-devel at yandell-lab.org>
> Subject:  [maker-devel] Why are some complete gene predictions not present in
> the final results?
> 
> Hi MAKER developers,
> 
> I was exploiting MAKER to perform gene prediction and annotation on my
> contigs.
> I used Artemis to examine gff and found some CDS with complete structure were
> absent in the final results.
> They are really predicted and annotated on the ref genome.
> I'm wondering if they were discarded due to overlapping with another CDS.
> How can I preserve these CDS?
> Thanks a lot in advance.
> 
> Hung-Wei
> _______________________________________________ maker-devel mailing list
> maker-devel at box290.bluehost.comhttp://box290.bluehost.com/mailman/listinfo/mak
> er-devel_yandell-lab.org



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20130522/d15002b7/attachment-0003.html>