gmod-ajax

[Gmod-ajax] trouble with GFF3 parsing

[James C. <ca...@gm...>]

I have the following file:
"""
##gff-version 3
##sequence-region CYP2C8 1 36725
CYP2C8	src	Feature	4100	4291	.	-	.	ID=37
CYP2C8	src	Feature	6125	6213	.	-	.	ID=38
CYP2C8	src	Feature	8093	8330	.	-	.	ID=39
CYP2C8	src	Feature	11013	11204	.	-	.	ID=40
CYP2C8	src	Feature	23538	23764	.	-	.	ID=41
CYP2C8	src	Feature	20003	20213	.	-	.	ID=42
CYP2C8	src	Feature	32411	32944	.	-	.	ID=43
CYP2C8	src	Feature	34400	35725	.	-	.	ID=44
"""

Running "flatfile-to-json.pl --gff /tmp/e.gff --tracklabel 'feature'
--cssclass feature --out browsers/human-gene-test --type Feature"
produces the following track:

"featureNCList":[[4099,4291,-1,"37"],[6124,6213,-1,"38"],[8092,8330,-1,"39"],[11012,11204,-1,"40"],[23537,23764,-1,"41"]]

You can see that everything above 30,000 bases is ignored.

Changing the sequence-region pragma to be a larger region fixes the problem:
##sequence-region CYP2C8 1 46725

What is going on here?  Why do I not get all the features I declared?

thanks,
James

Re: [Gmod-ajax] trouble with GFF3 parsing

[Mitch S. <mit...@be...>]

On 01/12/2010 02:29 AM, James Casbon wrote:
> Changing the sequence-region pragma to be a larger region fixes the problem:
> ##sequence-region CYP2C8 1 46725
>
> What is going on here?  Why do I not get all the features I declared?
>    

Thanks for providing such a nice clean test case.

This sounds like it might be related to an issue in 
Bio::DB::SeqFeature::Store::memory that was fixed last year after 
bioperl 1.6 was released.  The fix is in bioperl 1.6.1, though--what 
version are you using?

Also, JBrowse used to need the sequence-region line, but in the current 
code it's no longer required.  So one work-around if you can't upgrade 
bioperl would be to leave out the sequence-region line, I think.

Regards,
Mitch

Re: [Gmod-ajax] trouble with GFF3 parsing

[James C. <ca...@gm...>]

2010/1/12 Mitch Skinner <mit...@be...>:
> On 01/12/2010 02:29 AM, James Casbon wrote:
>>
>> Changing the sequence-region pragma to be a larger region fixes the
>> problem:
>> ##sequence-region CYP2C8 1 46725
>>
>> What is going on here?  Why do I not get all the features I declared?
>>
>
> Thanks for providing such a nice clean test case.
>
> This sounds like it might be related to an issue in
> Bio::DB::SeqFeature::Store::memory that was fixed last year after bioperl
> 1.6 was released.  The fix is in bioperl 1.6.1, though--what version are you
> using?
>
> Also, JBrowse used to need the sequence-region line, but in the current code
> it's no longer required.  So one work-around if you can't upgrade bioperl
> would be to leave out the sequence-region line, I think.

I think you are right that it may be a problem with bioperl, but I am
fairly certain I am on 1.61, although this is confusing:
$ perl -MBio::Root::Version -e 'print $Bio::Root::Version::VERSION,"\n"'
1.006001

Is there any way to load data that doesn't go via bioperl?  My
experience getting GFF3 to parse has not been joyful.

cheers,
James

Re: [Gmod-ajax] trouble with GFF3 parsing

[Mitch S. <mit...@be...>]

On 01/13/2010 01:49 AM, James Casbon wrote:
>
> I think you are right that it may be a problem with bioperl, but I am
> fairly certain I am on 1.61, although this is confusing:
> $ perl -MBio::Root::Version -e 'print $Bio::Root::Version::VERSION,"\n"'
> 1.006001
>    

Yeah, that's 1.6.1; I think it's formatted that way so that textual 
comparisons of version strings give the right answer (e.g., when 
comparing 1.6.2 to 1.6.12).

> Is there any way to load data that doesn't go via bioperl?  My
> experience getting GFF3 to parse has not been joyful.
>    

Well, the idea was to use bioperl to deal with parsing GFF3 so that we 
wouldn't have to write our own GFF3 parser.  The problem was in bioperl 
in this particular case, but (if you haven't already) you could also see 
if your GFF3 validates:
http://dev.wormbase.org/db/validate_gff3/validate_gff3_online

The bug where some features at the end of the refseq wouldn't show up 
has been fixed in bioperl, although it'll be a little while before the 
fix gets released.  I've attached a patch that you could apply to your 
bioperl installation, or you could get this file:

http://code.open-bio.org/svnweb/index.cgi/bioperl/checkout/bioperl-live/trunk/Bio/DB/SeqFeature/Store/memory.pm?rev=16695

and replace the version you currently have installed.

That said, you can also use BED, or a database (although the database 
would have to be one that has a bioperl Bio::DB::(whatever) interface, 
like chado, Bio::DB::SeqFeature::Store, or Bio::DB::GFF).

I've been thinking about writing something to import data into JBrowse 
that doesn't use bioperl, but some of the interfaces in the JBrowse code 
have to be reworked a bit before that can happen.

Regards,
Mitch

Re: [Gmod-ajax] trouble with GFF3 parsing

[James C. <ca...@gm...>]

2010/1/15 Mitch Skinner <mit...@be...>:
> Well, the idea was to use bioperl to deal with parsing GFF3 so that we
> wouldn't have to write our own GFF3 parser.  The problem was in bioperl in
> this particular case, but (if you haven't already) you could also see if
> your GFF3 validates:
> http://dev.wormbase.org/db/validate_gff3/validate_gff3_online
>
> The bug where some features at the end of the refseq wouldn't show up has
> been fixed in bioperl, although it'll be a little while before the fix gets
> released.  I've attached a patch that you could apply to your bioperl
> installation, or you could get this file:
>
> http://code.open-bio.org/svnweb/index.cgi/bioperl/checkout/bioperl-live/trunk/Bio/DB/SeqFeature/Store/memory.pm?rev=16695
>
> and replace the version you currently have installed.

Thanks, very much, Mitch.  I'll try that out and let you know.   To be
honest, the parser has given me a lot of hassle with other stuff (line
endings and memory use) so I might try out BED and see if that is any
better.


James