geneontology / go-ontology

Source ontology files for the Gene Ontology
http://geneontology.org/page/download-ontology
Creative Commons Attribution 4.0 International
223 stars 40 forks source link

question /localization #6035

Closed gocentral closed 9 years ago

gocentral commented 15 years ago

Not sure about these:

  1. Should the nucleosome assembly terms be a child of cellular protein complex localization?

  2. should cellular protein complex localization have parent cellular protein localization?

Reported by: ValWood

Original Ticket: geneontology/ontology-requests/6053

gocentral commented 15 years ago

responding out of order ...

  1. Good question. At the moment, protein complex localization is a sibling of protein localization for both cellular and generic, and the same holds for macromolecule loc and macromolecular complex loc, so at least it's internally consistent. If we change one we should change 'em all. I can raise the question with David, Jane, etc. if you think it's important.

  2. I'm inclined not to put nucleosome assembly under protein complex localization, but my reason is that nucleosomes are protein-DNA complexes, so the entire process of forming a nucleosome isn't just localizing a protein complex.

m

Original comment by: mah11

gocentral commented 15 years ago

For 1. I agree. I realised this later whan I realised why the "deposition wasn 't relevent", as the nucleosome encompasses the DNA too.

  1. Could you ask. I will think about this sime more.

Val

Original comment by: ValWood

gocentral commented 15 years ago

OK, mail sent ...

Original comment by: mah11

gocentral commented 15 years ago

The underlying question for making

protein localization --[i]protein complex localization

is this:

TRUE or FALSE: protein complex is_a protein

I lean towards the FALSE because the definition of 'protein complex' in GO is this:

Any macromolecular complex composed of two or more polypeptide subunits, which may or may not be identical. Protein complexes may have other associated NON-PROTEIN [emphasis mine] prosthetic groups, such as nucleotides, metal ions or carbohydrate groups.

This line of reasoning points to NOT making the relationship proposed between the localization terms.

My 2 cents,

Tanya

Original comment by: tberardini

gocentral commented 15 years ago

OK, I agree with your logic.

The reason I wondered about this was because I often make annotations to "cellular protein localization" when gene/protein A is shown to be required for the correct localization of protein B.

Sometimes, it later becomes apparent that gene/protein A is, in fact, localizing a complex C containing protein B, so there is a sort of inadvertant incorrectness in the original annotation, which has to be changed to the other branch.

I wondered whether it was more definitional, than what a protein actually is. For intance if the protein localization terms were defined any process by which individual proteins or complexes are transported to.... instead of Any protein by which a protein is transported to.... This would work.

I suspect that a lot of the use of protein localization will turn out to be protein complex localization which means essentially we have been using protein localization term in the more general way I described above.

Doe that make sense? I'd be interested to know how the term has been used.

Original comment by: ValWood

gocentral commented 15 years ago

We do include 'protein complex binding' as a child of 'protein binding' though (protein binding: Interacting selectively with any protein or protein complex (a complex of two or more proteins that may include other nonprotein molecules)).

I couldn't find anywhere we define 'protein' in GO! ChEBI have "A biological macromolecule minimally consisting of one polypeptide chain synthesized at the ribosome." so that would include protein complexes wouldn't it?

Original comment by: jl242

gocentral commented 15 years ago

Given the definition we use for protein complex, it would make sense to move protein complex binding.

Original comment by: mah11

gocentral commented 15 years ago

We'd hope PRO to have the definitive answer:

PRO:000000001 ! protein [DEF: "A biological macromolecule that is composed of amino acids linked in a linear sequence (a polypeptide chain) and is genetically encoded. Proteins descended from a common ancestor can be classified into families and superfamilies composed of products of evolutionarily-related genes. The domain architecture of a protein is described by the order of its constituent domains. Proteins with the same domains in the same order are defined as homeomorphic."]

If this is refined to "composed entirely of amino acids linked..." then we reject protein_complex is_a protein

note we can reject this assumption and yet still have protein complex localization is_a protein localization, if we define protein localization sufficiently broadly such that it covers the parts of the thing localized.

non-bio example: transport of me entails the transport of the change in my pocket, my teeth...

of course you could get relativistic and say the location of my teeth has remained exactly the same, in my mouth, even though I have moved across the room. Before long we're asking questions like "can a hole rotate?"

From Val's question it sounds like it might be desirable to have "protein complex localization" is_a "protein localization". If so, then we can justify this by treating location in a non-relative fashion. We can add an axiom to the bp_xp_cc file such that results_in_localization_of transitive_over has_part. In cc_xp_pro we have protein_complex = complex that has_part >=2 protein. Then the reasoner will recapitulate the asserted is_a between "PC loc" and "P loc". Ta-ra!

We could do the same to recapitulate the existing binding links Jane mentions. But is this justifiable? Could we not have a protein complex binding in which a carboydrate group not part of either protein be the binding participant?

Original comment by: cmungall

gocentral commented 15 years ago

> We can add an axiom to the bp_xp_cc file such that results_in_localization_of > transitive_over has_part. In cc_xp_pro we have protein_complex = complex > that has_part >=2 protein.

I think this would work well for localization (tho I'm inclined to hold off on making "protein complex localization" is_a "protein localization" until we can actually implement the above cleverness in the live GO).

I don't think it fits as well for binding, because binding may not affect the whole complex the way localization does, making Chris' carbohydrate example relevant.

Original comment by: mah11

gocentral commented 14 years ago

This stalled quite a while ago, but a new SF item has made me look at it again.

Just today it occurred to me that the relevant question to ask is this:

Does every instance of protein complex localization involve protein localization?

I think the answer is yes -- you can't move a protein complex without moving some proteins. This is essentially what Chris' comment is getting at, but maybe in plainer English.

And referring to Tanya's early comment, although I agree that 'protein complex is_a protein' is false, it doesn't necessarily follow that 'protein complex localization is_a protein localization' is also false.

I'm now in favor of adding the link (and parallel links e.g. for the corresponding cellular terms); I don't even need to wait for xp implementation.

I do still think binding is different. As long as we have protein complex defined as including non-protein other stuff as well as 2 or more proteins, something can bind to a protein complex without binding to the protein portion, so I think protein complex binding should not be is_a protein binding.

m

Original comment by: mah11

gocentral commented 14 years ago

I came to the same conclusion when making these annotations.

The other problem, if these were disjoint is that the experiments do not always distinguish between proteins and complexes so I would have to re-evaluate >500 annotations.....

Original comment by: ValWood

gocentral commented 14 years ago

Original comment by: mah11

gocentral commented 14 years ago

Hi midori,

ok, I buy your reasoning for the protein complex/protein

full steam ahead.

T

Original comment by: tberardini

gocentral commented 14 years ago

DONE (hurrah!)

Original comment by: mah11

gocentral commented 14 years ago

Original comment by: mah11

gocentral commented 13 years ago

Original comment by: mah11