Relations to describe gene product - component annotation

[vanaukenk]

Following on from the Cambridge 2017 GO meeting:

We need to solidify what relations we will use going forward for GP - CC annotation.

Strongest relation: 'is active in' https://www.ebi.ac.uk/ols/ontologies/ro/properties?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FRO_0002432

Weaker relation: 'part of' https://www.ebi.ac.uk/ols/ontologies/ro/properties?iri=http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FBFO_0000050

New relation: 'located at'? What are the use cases?

Also see: https://github.com/oborel/obo-relations/issues/193 https://github.com/oborel/obo-relations/issues/215

[vanaukenk]

For discussion of colocalizes_with see: https://github.com/geneontology/go-annotation/issues/1500

[pgaudet]

At the GOC meeting it was suggested to replace 'part of' by something more explicitely about the location, for example 'located in' or even 'found in'.

@cmungall thoughts about a good relation label ?

[pfey03]

I second what Pascale said, to replace part_of as this is not really a loose relation. Maybe 'localizes to" would reflect the actual purpose well. Or "located at" as Kimberly stated above.

[bmeldal]

Please remember that there are 2 kinds of GP-CC relationships

1. GP is part_of some complex
2. GP is_located_at/localizes_to some cellular location

The two situations could get different qualifiers that would immediately distinguish which situation we are talking about.

The new Complex WG will look into the possibility of making the part_of complex relationships more expressive taking into account what we know about the function of the GP that's part of a specific complex.

[pfey03]

I thought we will have two in total only?

1. GP enables_activity_in
2. GP localizes_to OR is_located_at replacing part_of

More might be overkill (for me).

[pgaudet]

GOC meeting discussion (Oct 2017) identified the following tasks:

• [ ] Survey groups to see how many want to express BOTH active location and just ‘incidental’ locations
• [ ] Define exactly the relation name: is the relation ‘part of’ be used for CC? Suggestions: ‘localises in’; ‘found in’
• [ ] Legacy annotations: will default to ‘part of’, but groups can move to ‘enables_activity_in’. We’ll also use the same strategies as for BP to assign ‘enables_activity_in’ DEADLINE: Proposal: By the next GOC meeting every group will have added the qualifier (have to edit your GAF production pipeline) or a default value will be assigned by GOC (can change them in the future)
• [ ] Tell Tony what the qualifiers are to put in P2GO
• [ ] Each group needs to update their GAF-generation pipeline to integrate the qualifier. For each group that does not provide any information in the qualifier, these will be automatically filled in by the GOC to the more general qualifier (“part of” for CC/ “causally upstream of or within” for BP)

[bmeldal]

1. GP enables_activity_in

2. GP localizes_to OR is_located_at replacing part_of

doesn't allow annotations of non-catalytic subunits that are definitively part of a complex. Alternatively, these GPs simply cannot be annotated to the GO complex term as localizes_to OR is_located_at are too weak (IMHO) to describe defined complex membership.

I think Sylvain (GH ID?) and @ukemi brought up these issues.

[pgaudet]

@sylvainpoux

[cmungall]

See https://github.com/oborel/obo-relations/issues/215

[bmeldal]

We didn't get round to discussing this point yesterday (30/11/17) but can also be added to user survey.

[bmeldal]

I've tried to summarise the situation regarding the way we extract annotations from complexes and from individual GPs:

Along the black lines, the top relationships are those used currently, the ones added in red pen are the ones we are discussing.

Points:

• no changes needed for BP
• complex AC part_of GO:complex should be is_a as they are equivalent or the curated complex in the CP is a specific case of the GO term.
• there is an issue with how to annotate the case of the RPB4/7 (https://github.com/geneontology/go-ontology/issues/10931)
• GP[cat] - catalytic subunit
• GP[reg/assoc.] - regulatory/associated subunit
• I don't think we have yet got a solution for specifying the relationship between GP[reg] and GO:CC other than sticking to part_of.

Please review the picture and post your component class related comments here.

Please post MF class related comments in #1662

[vanaukenk]

While working on https://github.com/geneontology/go-ontology/issues/14832 I came across this paper:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3951269/

with a complex localization description for CPTP (from Cercopithecus aethiops, presumably).

Our annotation guidelines for CC annotation relations should use this paper as a benchmark.

[bmeldal]

@ValWood I guess the situation here is the same as for the MF annotations. If users ignore the qualifier there's no point distinguishing whether a protein is active in a complex or regulatory (licensing as Val calls it)?

This discussion is linked to the use/misuse of colocalises_with, as linked above.

Draft survey: https://docs.google.com/document/d/1P_VLM9g13kj9lu3CRAgotAI3cUmWkS1yWaVg95u2Vbk/edit?usp=sharing

[bmeldal]

Summary from call on 9th May 2018:

@vanaukenk @deustp01 @sandraorchard @pgaudet @hdrabkin @krchristie @ValWood @tberardini

1. Inferring complex membership from complex to GP no issue

2. Inferring location of GP from complex tricky as experiments usually only show location of a selection of GPs, and complex location is inferred based on other experiments that indicate these GP(s) are found in the given complex. This was already discussed in relation to Noctua use and will be further debated on NYC mtg.

Conclusion:

• More complex modelling might be possible
• Parked til NYC mtg
• NB: We didn't get into the debate of more specific qualifiers at this point

[bmeldal]

From call on 31/1/19:

linked to https://github.com/geneontology/go-annotation/issues/1650

What is the difference btw: GP1 enables MF | AE has_direct_input "CP AC" and GP1 is_active_in CP AC | ?some AE with the MF?

The latter is less expressive as the MF gets lost if the AE and/or relationship is not captured by the tools.

To be discussed in relationship with the RO Hackathon. @vanaukenk

[bmeldal]

From call on 13/2/19:

Present: Birgit, Harold, Ruth, Kimberly, Darren, Peter, Leonore, Suzie, Judy, Edith, Helen

RO group suggested:

part_of [localisation or complex]

localizes_to [subcellular_location] - when not clear if active here is_active_in [subcellular_location] - when it's clear that it's active in this location

located_in = localizes_to

Action:

• Groups to look at current curation practice and discuss in next call.

[bmeldal]

From call on 25th Feb 2019:

Present: Birgit, Harold, Ruth, Thomas Hayman, Kimberly, Darren, Peter, Suzie, Judy, Edith, Helen, Val

NB: qualifier in GAF = relationship on GPAD

• We agreed that the above proposal from the RO WG is good, but the following cases need defining:

1. Many proteins are recycled (e.g. membrane receptors in plasma and vesicle membranes) or functional in several compartments (e.g. nuclear receptors in plasma membrane and nucleus) --> Should they be annotated with is_active_in in all those compartments?

2. So far GO only annotated the functional location of a GP if it was known but proteomics studies use compartment markers to check for contamination --> GPs are often found transiently in a location which is a true location but not their functional site (e.g. path from the ER to its final destination, receptor recycled from vesicle back to plasma membrane) --> can we use the localized_to relationship here to make GO more useful for HT proteomics users?

--> We thought both cases were sensible options.

3. Potential for annotation redundancy:

GP1 enables MF | AE occurs "CC" and GP1 is_active_in CC | ?some AE with the MF?

--> We are happy for this double annotation as tools will ignore the AE and which removes the redundancy - otherwise we'd either loose the CC or MF annotation.

Action:

• bring proposal to GOC meeting

[bmeldal]

Will be discussed at the Cambridge meeting

[pgaudet]

Added to the agenda.

I have a question about the proposal: Usually when some data provides imprecise information, we use a parent rather than a sibling. What is the difference between 'part_of ' and 'localizes_to' ? It seems to me like these two relations are redundant.

Thanks, Pascale

[vanaukenk]

Discussing with @ukemi and @pgaudet

The proposal from the Berkeley hackathon (not RO, per se) was to have this structure:

part_of _or_located_in -part_of -located_in —is_active_in

For GO annotation, we would propose to use 'located in' and 'is active in'.
What we need to clarify still is how we think about, and use, 'part of'. If a physical entity (continuant) is 'part of' another physical entity (continuant), is it necessarily located there?

@balhoff @cmungall

[vanaukenk]

@ukemi @pgaudet @cmungall

Do we still want to present a proposal for this at the Cambridge meeting? If so, we need to review the meaning, and use of, part_of for CC annotations.

Thx.

[ukemi]

I think so. Should this get rolled into a bigger spatial relations versus functional relations project?

[pgaudet]

I'm available to help if we decide to do it.

Pascale

[bmeldal]

Are we discussing this in the context of the Complex WG or in a separate slot? (I'm starting on my slides ;-) )

[pgaudet]

This is resolved , see https://wiki.geneontology.org/Annotation_Relations#Standard_Annotation:_Gene_Product_to_Term_(gp2term)_Relations