Template based relations - call 12/4

[Bjoern Peters]

Hi all, 

It has come up repeatedly that we have a conflict in RO between the desire to keep RO to a small set of well-defined relationships, and the need to create a lot of specific relationships that are useful to make direct statements in a specific domain (last example was 'specimen collected during organism life stage', but there are plenty of others). 

I previously suggested that we create a 2-tier set of relationships, where the RO 'core/base/foundational' relationships are kept to a minimum, and only use range/domain elements from COB. 

In addition we allow for 'template based relationships', which have to follow a documented design pattern (implemented in e.g. a ROBOT template or the like), which we would freely give out with limited discussion. 

This has been talked about in past RO calls, and there was no massive pushback, so I would like to give this a try in the RO call I am chairing on Dec 4th. I am much less familiar with RO than I should be, so I would like to ask for feedback beforehand, specifically if some such efforts already implicitly exist (I believe GO does, but I couldn't immediately find this). 

Also it would be great to pick an example. Looking at RO, many of the 'grouping relationships' are actually very different from each other and are not good examples. The one that kind of work as a true parent seems to be 'has participant'. Any better suggestions? 

- Bjoern

--

Bjoern Peters
Professor
La Jolla Institute for Immunology
9420 Athena Circle
La Jolla, CA 92037, USA
Tel: 858/752-6914
Fax: 858/752-6987
http://www.liai.org/pages/faculty-peters

[Bill Duncan]

Hi Bjeorn,

This has been sitting in my inbox for a bit. Just now getting a chance to respond.

I agree in general with the approach you advocate. I don't know the best way to approach this given that we volunteer our time.

I like the idea of establishing high-level design patterns. I am agnostic about whether new terms should be added using robot templates, submissions the ro-edit.ow file, or an ontology import file. All three approaches have their own issues. 

One thing I think would help is that we formally create subgroups to govern different domains (see ticket). That would allow it to be more efficient getting term requests filled and also targeting the right people to review NTRs, and I think the GO process for adding terms should be more transparent. Sometimes it seems like new GO terms get added w/o any reviews from folks.

It would also be great if we could hammer out RO's relation with COB. Although the ticket about COB vs native IDs needs to be settled (I know ... I'm part of the controversy going on with this. I apologize). Also, Damion's NTR collected during needs a range of occurrent, which is not in COB (at the moment).

In the past, I was pushing for all domain and range classes to be in RO. But, now I am thinking this just isn't really practical given our limited resources. We can have high-level domains and ranges, but, if so, I think we have to allow other ontologies to specify more restrictive domains and ranges. I realized this runs counter to folks who do not want other ontologies injecting axioms. However, without resources to manage this, I don't see how we can maintain such a position.

Anyway, I hope my thoughts were helpful. I may not be able to make the Dec. 4th call.

Bill

--
You received this message because you are subscribed to the Google Groups "obo-relations" group.
To unsubscribe from this group and stop receiving emails from it, send an email to obo-relation...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/obo-relations/CAFc%2BKWhFZyQ5OYVZGN%3DFeF%2B5k3dEKFWB-%3D7d4Rqd3znBYf4OhA%40mail.gmail.com.

[Bjoern Peters]

Thanks for your thoughts Bill! You are touching on a bunch of topics - which are all important, but definitely to much to cover in this one call. I would like to narrow it to an attempt of 'how do we define relationship design patterns', and it sounds like you are very much in agreement with that goal. I also think it works well with your other point of subdividing RO responsibilities into different interest groups (trying to avoid 'domain' here because overload...), as different group could be in charge of different templates. 

All: I would still very much appreciate inputs on a good topic for a relation (or maybe 2 or 3) that we can go to. Preferably something uncontroversial. 

Some ideas, just to get this started, and illustrate better what I am looking for: 

1) process 'has participant that realizes function' some material entity; which would be the parent of relationship like 'utilizes device', 'utilizes reagent', 'utilizes fluorescent label', 'utilizes sequencing instrument'

2) material entity 'has part with function' some material entity; which would be parent of 'has DNA binding site', 'has power supply', 'has packaging'

  

To view this discussion on the web visit https://groups.google.com/d/msgid/obo-relations/CAM2Eo-ec_tNc1LbkLSXjA0mfDYkV-UqynbLJ9BCDSePtwtT58A%40mail.gmail.com.