[maker-devel] Mapping gene names

Carson Holt carsonhh at gmail.com
Wed May 14 15:18:52 MDT 2014 

Thanks.  Looks interesting. Also since output is already GFF3, you could probably just use it with gff passthrough.  It doesn't appear to support eukaryotes though.

--Carson


Sent from my iPhone

> On May 14, 2014, at 3:07 PM, Shaun Jackman <sjackman at gmail.com> wrote:
> 
> Hi, Carson. Perhaps MAKER could integrate Barrnap to predict rRNA.
> 
> Cheers,
> Shaun
> 
> 
>> On 4 March 2014 18:33, Carson Holt <carsonhh at gmail.com> wrote:
>> Trying to call non-coding RNA from ESTs or even sequence homology is extremely messy (non-trivial problem in most organisms with high false positive rate), so MAKER for the most part doesn’t even try to do that.  It focuses only on the coding genes.  You can now use tRNAscan and snoscan in the newest version for some non-coding RNA support (those features were only added a couple of months ago).  So just like other prediction tools (snap, augustus etc.), the primary focus has always been the coding genes.  We’ve only started adding non-coding RNA support recently for iPlant, so it’s still relatively immature.
>> 
>> Thanks,
>> Carson
>> 
>> 
>> From: Shaun Jackman <sjackman at gmail.com>
>> Reply-To: Shaun Jackman <sjackman at gmail.com>
>> Date: Tuesday, March 4, 2014 at 7:10 PM
>> 
>> To: Carson Holt <carsonhh at gmail.com>
>> Cc: "maker-devel at yandell-lab.org" <maker-devel at yandell-lab.org>
>> Subject: Re: [maker-devel] Mapping gene names
>> 
>> Hi, Carson. I set single_length=50, and it worked like a charm. Thanks for the tip.
>> 
>> The rRNA genes that are found with est2genome have the feature type set to mRNA and have corresponding five_prime_UTR, CDS and three_prime_UTR features. Ideally the feature type would be set to rRNA or tRNA as appropriate, and would omit the UTR and CDS features. Is that a feature that you would be interested in adding to MAKER? The rRNA gene names all start with “rrn” and the tRNA gene names with “trn”, as is standard, so determining the appropriate type should be straight forward.
>> 
>> Thanks again for your help with this. Cheers,
>> Shaun
>> 
>> 
>> 
>>> On 27 February 2014 17:13, Carson Holt <carsonhh at gmail.com> wrote:
>>> Set single_exon=1, and the minimum size to a smaller value.  I think it's set to 250 right now.  Also est2genome is looking for ORF, so if there is none (as with tRNAs) they probably won't get picked up.
>>> 
>>> --Carson 
>>> 
>>> Sent from my iPhone
>>> 
>>>> On Feb 27, 2014, at 5:27 PM, Shaun Jackman <sjackman at gmail.com> wrote:
>>>> 
>>>> Sorry, ignore my previous question. est_forward also carries forward the names of protein evidence and works like a charm. Thank you!
>>>> 
>>>> The larger rrn16 and rrn23 genes annotated perfectly, but the smaller rrn4.5 and rrn5 and tRNA genes didn’t make it into the all.gff file. They are in the blastn output, and in the evidence_0.gff. rrn5 has perfect identity, sufficient bits (242 > bit_blastn=40) and sufficient E Value (2e-66 < eval_blastn=1e-10). How should I debug which filter is removing these hits?
>>>> 
>>>> organism_type=prokaryotic
>>>> est2genome=1
>>>> protein2genome=1
>>>> est_forward=1
>>>> Cheers,
>>>> Shaun
>>>> 
>>>> 
>>>> 
>>>>> On 27 February 2014 15:17, Shaun Jackman <sjackman at gmail.com> wrote:
>>>>> Is there a corresponding protein_forward=1 option to map forward protein names from protein2genome?
>>>>> 
>>>>> Cheers,
>>>>> Shaun
>>>>> 
>>>>>> On 2014-February-26 at 15:45:39 , Carson Holt (carsonhh at gmail.com) wrote:
>>>>>> 
>>>>>> Sorry I meant to say prefilter on the score in the mRNA column before passing the gff3 to model_gff.
>>>>>> 
>>>>>> --Carson 
>>>>>> 
>>>>>> Sent from my iPhone
>>>>>> 
>>>>>>> On Feb 26, 2014, at 3:50 PM, Carson Holt <carsonhh at gmail.com> wrote:
>>>>>>> 
>>>>>>> What you can do is run it once with just est_forward=1 and est2genome/protein2genome set to 1.  Then take those results, pass them in as model_gff and use the map_forward option to then filter the results based on mRNA score and that would copy names onto new gene under the standard MAKER pipeline.  Eventually it’s really supposed to go into a separate tool that will map genes onto new assemblies (but under the hood the tool will just be calling MAKER with certain parameters restricted).  I do this because if people commonly use it mixed with things like SNAP I can start to get some very weird behaviors. 
>>>>>>> 
>>>>>>> Thanks,
>>>>>>> Carson
>>>>>>> 
>>>>>>> From: Mikael Brandström Durling <mikael.durling at slu.se>
>>>>>>> Date: Wednesday, February 26, 2014 at 3:04 PM
>>>>>>> To: Carson Holt <carsonhh at gmail.com>
>>>>>>> Cc: "maker-devel at yandell-lab.org" <maker-devel at yandell-lab.org>
>>>>>>> Subject: Re: [maker-devel] Mapping gene names
>>>>>>> 
>>>>>>> It seems that this could be a very useful option in those cases where you have firm a priori knowledge of the placement of ESTs. However, while trying it I note that est_forward implies that the est2genome predictor is turned on, implicitly. Is this necessary for this to work? I’m after the behavior you describe below where exonerate is made to try really hard within a limited region to align an est, but I would not like maker to produce est2genome predictions.
>>>>>>> 
>>>>>>> In general, I think this maker_coor and est_forward is a feature set that is worthy to be promoted into a documented feature.
>>>>>>> 
>>>>>>> THanks,
>>>>>>> Mikael
>>>>>>> 
>>>>>>>> 26 feb 2014 kl. 17:09 skrev Carson Holt <carsonhh at gmail.com>:
>>>>>>>> 
>>>>>>>> It will still work without est_forward.  It just works a little differently.  Keep in mind this was a hidden feature I used to find stubborn or hard to find missing genes after reassembly of a genome.
>>>>>>>> 
>>>>>>>> If est_forward is provided, MAKER will parse the database to look for the maker_coor tags early in the pipeline.  Then it will create a list of locations to search, and it will search them even if there are no BLAST results to seed the search (normally MAKER gets a BLAST result first and then polishes it with exonerate).  So maker_coor=chr1 will cause MAKER to look for a match using all of chr1 as the input to exonerate even when BLAST finds nothing (this is a very very slow search, but can help pick up one or two stubborn genes that don’t remap well).  To allow this, MAKER gives exonerate looser matching parameters (i.e. allows for single base pair introns perhaps caused by assembly errors).  The logic here is that given the fact that I already told MAKER that with some degree of confidence I expect sequence A to map to to location X, it will try its hardest to make it match. 
>>>>>>>> 
>>>>>>>> Without est_forward set, the maker_coor= flag still gets read in GI.pm at line 1563, but only after a BLAST alignment has already seeded it to the region (that BLAST result has the information in its description parameter).  MAKER will then ignore seeds completely outside of maker_coor. In addition any BLAST seeds that overlap maker_coor will get the search space for alignment polishing adjusted to match maker_coor exactly.  Also match parameters for exonerate will not be relaxed as they were with est_forward.
>>>>>>>> 
>>>>>>>> As you can see the behavior, is slightly different (because it’s an accidental feature).
>>>>>>>> 
>>>>>>>> Thanks,
>>>>>>>> Carson
>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>> From: Mikael Brandström Durling <mikael.durling at slu.se>
>>>>>>>> Date: Wednesday, February 26, 2014 at 6:37 AM
>>>>>>>> To: Carson Holt <carsonhh at gmail.com>
>>>>>>>> Cc: "maker-devel at yandell-lab.org" <maker-devel at yandell-lab.org>
>>>>>>>> Subject: Re: [maker-devel] Mapping gene names
>>>>>>>> 
>>>>>>>> That might be a useful and time saving accidental feature. But, reading the code, it seems that I need to supply maker_coor but not gene_id, as well as the configuration option est_forward for this to work. Any occurrences of maker_coor in GI.pm seems to be conditioned on set_forward=1 right?
>>>>>>>> 
>>>>>>>> Mikael
>>>>>>>> 
>>>>>>>>> 26 feb 2014 kl. 14:22 skrev Carson Holt <carsonhh at gmail.com>:
>>>>>>>>> 
>>>>>>>>> Yes.  That should work as well as an accidental feature.
>>>>>>>>> 
>>>>>>>>> --Carson 
>>>>>>>>> 
>>>>>>>>> Sent from my iPhone
>>>>>>>>> 
>>>>>>>>>> On Feb 26, 2014, at 5:30 AM, Mikael Brandström Durling <mikael.durling at slu.se> wrote:
>>>>>>>>>> 
>>>>>>>>>> Can this use of maker_coor be used only to hint about the placement of the ests, without affecting the naming of the final genes? Ie if I have a database of EST where I have a priori knowledge of their rough placement, can this placement be given to maker without providing est_forward=1?
>>>>>>>>>> 
>>>>>>>>>> Thanks,
>>>>>>>>>> Mikael
>>>>>>>>>> 
>>>>>>>>>>> 26 feb 2014 kl. 01:58 skrev Carson Holt <carsonhh at gmail.com>:
>>>>>>>>>>> 
>>>>>>>>>>> There is a way.  It’s not a standard option and it’s undocumented, but if you add est_forward=1 to the maker_opts.ctl file, then it will do just that.  The option won’t already be there so you’ll have to type it in.
>>>>>>>>>>> 
>>>>>>>>>>> There is also a feature designed to work with this option.  If you add tags to your fasta headers, those can be used to guide the mapping and naming.  For example, gene_id=<some_gene>  will ensure different isoforms that share a common gene_id get clustered into the same gene, and maker_coor=chr1:1-10000 in the fasta header will force a particular sequence to only be mapped against chr1 within the range of 1-10000 bp  and just using maker_coor=chr1 will force it to only be mapped against chr1.
>>>>>>>>>>> 
>>>>>>>>>>> This is an undocumented way to remap genes onto new assemblies using blast alignments of earlier transcript or protein annotations as a guide.
>>>>>>>>>>> 
>>>>>>>>>>> —Carson
>>>>>>>>>>> 
>>>>>>>>>>> 
>>>>>>>>>>> 
>>>>>>>>>>> 
>>>>>>>>>>> From: Shaun Jackman <sjackman at gmail.com>
>>>>>>>>>>> Reply-To: Shaun Jackman <sjackman at gmail.com>
>>>>>>>>>>> Date: Tuesday, February 25, 2014 at 5:06 PM
>>>>>>>>>>> To: <maker-devel at yandell-lab.org>
>>>>>>>>>>> Subject: [maker-devel] Mapping gene names
>>>>>>>>>>> 
>>>>>>>>>>> Hi,
>>>>>>>>>>> 
>>>>>>>>>>> I’m annotating a genome using a closely related genome from Genbank, using the .frn (RNA) and .faa (protein) files from Genbank as evidence to annotate my genome. I’ve run Maker, and the annotation seems to have worked well. Is it possible to map the names of the genes from the related species to my annotation? I see the map_forward option, which applies to the model_gff parameter. Is there a similar option for est and protein?
>>>>>>>>>>> 
>>>>>>>>>>> maker_opts.ctl
>>>>>>>>>>> 
>>>>>>>>>>> 
>>>>>>>>>>> est=NC_123456.frn
>>>>>>>>>>> protein=NC_123456.faa
>>>>>>>>>>> est2genome=1
>>>>>>>>>>> protein2genome=1
>>>>>>>>>>> Thanks,
>>>>>>>>>>> Shaun
>>>>>>>>>>> 
>>>>>>>>>>> _______________________________________________ maker-devel mailing list maker-devel at box290.bluehost.com http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
>>>>>>>>>>> _______________________________________________
>>>>>>>>>>> maker-devel mailing list
>>>>>>>>>>> maker-devel at box290.bluehost.com
>>>>>>>>>>> http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
>>>>>> _______________________________________________ 
>>>>>> 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/20140514/ff7f1e74/attachment-0003.html>

More information about the maker-devel mailing list