[maker-devel] *maker.proteins and*non_overlapping_ab_initio.proteins files

[Carson Holt]

correct_est_fusion is not guaranteed to never merge a gene.  If you are
giving maker imperfect evidence, there is only so much it can do.  Also you
should be using protein evidence in combination with EST evidence,
especially when using the correct_est_fusion option or you are limiting it's
effectiveness.  MAKER does not work as well on ESTs alone, especially for
organisms with few introns as internal logic is relying on the combination
of evidence support.

--Carson


From:  刘慧泉 <liuhuiquan at nwsuaf.edu.cn>
Date:  Tuesday, 16 April, 2013 9:49 PM
To:  <maker-devel at yandell-lab.org>
Subject:  Re: [maker-devel] *maker.proteins
and*non_overlapping_ab_initio.proteins  files



Hi Carson and Daniel,
 
Thank you very much for your quick responses!
 
By multiple tries, I have known the reason why only a few genes were
annotated by maker. This is due to turn on of the “correct_est_fusion”
option. I got about 8000 transcripts from PASA assembly. Because the gene
density of my fungus is very high, many of the assembled transcripts merged
adjacent genes even if the trinity and PASA were used with relevant
parameter. Maker may not use the merged transcripts as evidence, it the
“correct_est_fusion” option is turn on. However, even though the
“correct_est_fusion” option is used, I also found many genes produced by
maker have merged more than one gene.
 
I’m now using the ORFs (trainingSetCandidates.cds) extracted from the
transcripts by PASA as the EST evidence supplied to maker. I found most of
the extracted ORF can accurate match the gene model predicted by augustus
and genemark. This can better resolve the “merged gene” issues for fungi
with high gene density.
 
For the 'non-overlapping' file, if only using genemark, its predictions can
be found in the 'non-overlapping' file. Is previously issue due to the gene
mode generated by augustus is better that genemark, so only augustus gene
was putted into the 'non-overlapping' file? Will the genes predicted only by
one program not found in the 'non-overlapping' file? how to get these genes?
 
Thank you
Huiquan


发件人: Carson Holt <carsonhh at gmail.com>
发送时间: 2013-04-16 24:01
收件人: 刘慧泉 <liuhuiquan at nwsuaf.edu.cn>;maker-devel at yandell-lab.org
主　题: Re:Re: [maker-devel] *maker.proteins
and*non_overlapping_ab_initio.proteins files

 1. by view the gff file produced by maker2, I have found most of the
predicted gene loci have est matches. but why only 254 gene annotations got
by maker2 ?

>> I'd really have to see the results to tell you why.


2. in the “non-overlapping ab initio”file, I found sequences are all from
augustus_masked prediction. Does the non-overlapping file only include the
best gene modes from predicted by both augustus and genemark?  Does it
include genemark- or augustus-specific genes ?
 
>> The 'non-overlapping' file should have the one with best consensus if there
>> are 3 or more predictors, and the longest one otherwise.  It should be able
>> to have augustus and genemark genes.  Try it with only genemark and let me
>> know if the file is empty.

Thanks,
Carson


_______________________________________________ maker-devel mailing list
maker-devel at box290.bluehost.com
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20130418/8400f99f/attachment-0003.html>