[Bioperl-l] GO categories and load_ontology.pl

[Hilmar Lapp]

Hi Annie,

you do need to look at the path in order to find out which terms are 
descendants of which of the three GO categories.

You can do this either

	- in bioperl by populating an ontology instance with all relationships 
from the database (see t/ontology.t in the bioperl-db test suite for an 
example) and then query the ontology object for all ancestors of your 
query term and see which category term is among them, or

	- like before, but instead of high-level querying for all ancestors 
use the graph that backs the engine and query the graph ($graph = 
$ontology->engine->graph for GO) directly using its path methods. The 
backing graph implementation is Graph::Directed; Nat Goodman said there 
are issues with some of the flow algorithms, but I don't remember the 
specifics (if you email Nat I'm sure he can tell you). Or,

	- query in SQL the term_path table, joining it to term in the same way 
as you would join term_relationship. You can automatically populate the 
transitive closure when you load or update GO by supplying --computetc 
to load_ontology.pl. See the POD for details.

Hth,

	-hilmar

On Tuesday, April 27, 2004, at 07:41  AM, Law, Annie wrote:

> Hi Hilmar and other bioperl enthusiasts,
>
> Thanks for your response.  I now have a better understanding of the 
> reason
> for the 200 entries
> Which are not in GO.  My focus now are the items that are in GO and 
> are not
> obsolete.
>
> My main concern now is given a GO id how do I find out what GO 
> category it
> belongs to.
> I want to know if it belongs to Molecular function, biological 
> process, or
> cellular component by the simplest method
> Available with bioperl.
> I don't see any obvious way of doing this since in the table called 
> term all
> of the entries in the term table are of
> Ontology Id = 1 (Gene ontology).
>
> It was suggested to me by another helpful bioperler that I could 
> somehow
> look to see if my term is a child of the
> GO id for molecular function, biological process, or cellular 
> component.
> This could be done by looking at the path
> Table if transitive closure was used.  I am not too sure about how to 
> go
> about doing this and would be open
> To other solutions.
>
> The other alternative I can think of is to load a text file from GO 
> ontology
> in my database which is the
> GO.terms_and_ids file which is provided by Gene Ontology.  This is 
> basically
> a tab delimited text file which
> includes the GO id and a letter such F, P, or C next to the ID to 
> indicate
> F = molecular function, P = biological process, C = cellular component.
> This is not really my first choice as a solution if I can access this
> information and it is somehow made available by the load_ontology.pl 
> script
> which I have already used.
>
> Thanks very much,
> Annie.
>
>
>
>
> -----------------------
> To: Law, Annie
> Cc: 'bioperl-l at bioperl.org'
> Subject: Re: [Bioperl-l] GO categories and load_ontology.pl
>
>
> Annie, I still owe you an answer for your earlier email. I haven't
> managed to get to that yet. See below for my response to this one.
>
> On Wednesday, March 17, 2004, at 08:50  AM, Law, Annie wrote:
>
>> It seems that most of the Entries in the term table are of Ontoloy Id
>> = 1
>> (Gene ontology) and only around 200 entries molecular function,
>> biological
>> process, and cellular component put together when there are about 
>> 16000
>> entries in the term table.
>> This is only true if I load locuslink into the database.
>
> This is because LocusLink lags behind the latest version of GO in terms
> of the release that they use for annotating sequences. I.e., LocusLink
> uses some terms which have meanwhile been retired or obsoleted from GO.
> Depending on whether they are still in GO's .defs file, they won't be
> in your database if you chose to ignore obsoleted entries (which is not
> a bad choice at all per se), or they aren't part of GO anymore at all.
>
> LocusLink doesn't give the ontology of GO terms (which would be 'Gene
> Ontology'); rather it gives the category. Because a term must have an
> ontology associated, the SeqIO LL parser interprets as the ontology
> what NCBI really meant to be the category.
>
> You'd have the following choices to proceed.
>
> 	- Ignore the 200 entries which aren't in Gene Ontology. You're not
> going to miss a significant amount of your annotation, and it's
> annotation with obsoleted terms anyway.
>
> 	- Load GO including obsoleted terms, and see with how many non-Gene
> Ontology terms that would leave you. If it's a lot less than 200, you
> may just want to ignore the rest.
>
> 	- Build a SeqProcessor module (see Bio::Factory::SeqProcessorI and
> Bio::Seq::BaseSeqProcessor) which takes the seq objects as the LL
> parser returns them, goes in and retrieves all GO term annotations, and
> replaces the ontology for those with 'Gene Ontology.' Then you pass
> your SeqProcessor to load_seqdatabase.pl using the --pipeline
> command-line option (see the script's POD).
>
> The last option may sound like but is really not a lot of work if you
> can program perl. Note, however, that then you still wouldn't have any
> relationships for those terms - they simply have been retired.
>
> Depending on what your project is, just ignoring those 200 may be the
> most reasonable way to go.
>
> 	-hilmar
> -- 
> -------------------------------------------------------------
> Hilmar Lapp                            email: lapp at gnf.org
> GNF, San Diego, Ca. 92121              phone: +1-858-812-1757
> -------------------------------------------------------------
>
>
-- 
-------------------------------------------------------------
Hilmar Lapp                            email: lapp at gnf.org
GNF, San Diego, Ca. 92121              phone: +1-858-812-1757
-------------------------------------------------------------