[Ontology-editors] CC, PRO and CORUM

[Darren Natale]

Judith Blake wrote:
> I concur with Midori here.  The purpose of the generation of this set was to expand the representation in GO for those complexes missing that are important in the mammalian system.  For this CORUM provided a great portal for my summer intern to identify and work with generating the intersection of the GO and CORUM sets.
> 
> However, as Midori notes, the definitions provided need to be generalized for the GO context.  And, as noted, there are more complexes to be represented as well.  I think the use of the CORUM set as a starting point is proving productive however, and I'm glad to see the extension of complex representation in GO.
> 
> The work of the moment, then, is to establish how we will represent the relationship between the context-dependent (species/celltype/space/time) representation in PRO and the generic representation in GO.

My initial thought was that is_a would be appropriate.  I think Chris 
expressed the same idea.  However, I'm no longer certain that this is 
the case.  For example, the inactive form of a complex--which would also 
be represented in PRO--would not fit the function-based definitions that 
are sometimes used in GO.  Technically speaking I suppose part_of could 
be used (Chris, correct me if I am wrong but I think the definition of 
part_of allows for saying some whole part_of itself).  I don't really 
like that, however.  Needs more thought.

> This does not address the other issue as to whether 'complexes', because of their context dependence should be classified differently than the cellular structures such as the mitochondria or golgi.
> 
> Judy
> 
> -----Original Message-----
> From: Midori Harris [mailto:midori at ebi.ac.uk] 
> Sent: Wednesday, November 26, 2008 9:29 AM
> To: Chris Mungall
> Cc: Ontology Editors; Darren Natale; Judith Blake
> Subject: Re: [Ontology-editors] CC, PRO and CORUM
> 
> Hi Chris,
> 
> As noted briefly yesterday, I'm gradually working my way through the many 
> suggested new complex terms in SF 2049652, and Jane has taken on a subset 
> in SF 2125996. It's very slow going, because I'm finding that I have to 
> check each term manually, and that often requires reading (or at least 
> skimming) much of the full text of a paper.
> 
> Based on what I'm finding, I would not incorporate anything into GO based 
> on importing or mapping anything from CORUM or MIPS. The Sin3 complex 
> example is fine, but there are problems with enough of the CORUM entries 
> I've looked at to make me very leery of trying to speed up adding the 
> complex terms. For example, CORUM cites PMID:11713266 for more different 
> complexes than the paper shows; CORUM:3284 is a separate entry based on a 
> paper that actually shows that a particular protein is a member of a 
> previously identified complex; for CORUM:832 the paper cited actually says 
> that two of the listed subunits are found in *different* complexes.
> 
> Even for the accurately described complexes, the localization (and maybe 
> also function and process) may be species-specific in some cases, and 
> those wouldn't meet our all-some criteria. Like the composition (and 
> existence!) of the complexes themselves, localizations might have to be 
> individually vetted.
> 
> For those CORUM complexes that do pass the curator vetting procedure, I'll 
> be able to pull GO IDs and CORUM IDs out of a spreadsheet if we want to 
> add dbxrefs.
> 
> midori
> 
> [snip PRO stuff]
> 
>> * Mapping CORUM to GO.
>>
>> I see we have 18 CC terms with synonyms with CORUM as provenance. It may be 
>> an idea to have a separate xref too (e.g. if the CORUM name is identical). So 
>> there's another ~1200 to be added? I couldn't find the tracker item for this.
>>
>> There are quite a few that could be trivially mapped. E.g:
>>
>> [Term]
>> id: GO:0016580
>> name: Sin3 complex
>> namespace: cellular_component
>> def: "A multiprotein complex that functions broadly in eukaryotic organisms 
>> as a transcriptional repressor of protein-coding genes, through the 
>> gene-specific deacetylation of histones. Amongst its subunits, the Sin3 
>> complex contains Sin3-like proteins, and a number of core proteins that are 
>> shared with the NuRD complex (including histone deacetylases and histone 
>> binding proteins). The Sin3 complex does not directly bind DNA itself, but is 
>> targeted to specific genes through protein-protein interactions with 
>> DNA-binding proteins." [PMID:10589671, PMID:11743021, PMID:12865422]
>> is_a: GO:0000118 ! histone deacetylase complex
>>
>> To:
>>
>> http://mips.gsf.de/genre/proj/corum/complexdetails.html?id=54
>>
>> Note that MIPS localizes this to the nucleus. Can we just import MIPS 
>> localizations, or are they not as rigorous as us in forcing an all-some 
>> relation?
>>
>> Once we have these we can also use mips2go and the mips funcat entry for each 
>> complex to get CC->MF/MP links
>>
>>
>> _______________________________________________
>> Ontology-editors mailing list
>> Ontology-editors at geneontology.org
>> http://fafner.stanford.edu/mailman/listinfo/ontology-editors