Re: post-comps and compositional descriptions

[Chris Mungall]

On Sep 18, 2008, at 9:20 AM, Jim Balhoff wrote:

> Hi Chris and Rob,
>
> I'm trying to get data back out of our OBD instance. In our
> OBDModelBridge class, we create phenotypes as
> CompositionalDescriptions before adding them to OBD. I'm trying to
> figure out if we're handling post-compositions correctly. If the
> phenotype references a post-composed entity term, such as TAO:
> 0000622^OBO_REL:part_of(TAO:0001114), right now that term ID is just
> being added to the phenotype CD, like this:
>
> cd.addArgument(relationVocabulary.inheres_in(),e.getID());
>
> Should the entity be made into a CompositionalDescription first,
> before being linked the quality like that?

You're right, it looks like the CD bearer entity needs to be added.
This can happen at any time, it doesn't have to be before the
enclosing phenotype CD is added.

Adding this now...

> I'm asking because when I try to get the entity out of OBD as a CD,
> I can't work with it in the same way as the phenotype CD. For
> example, in this code:
>
> CompositionalDescription cd =
> this.shard.getCompositionalDescription(node.getId(), false);
> String genusID = "";
> for (Statement statement : cd.getDescriptionAsStatements()) {
> genusID = statement.getNodeId();
> break;
> }
>
> The CD has no statements. I expected at least one saying that TAO:
> 0000622 is part of TAO:0001114.
>
> I don't see however how to add a CD to another CD with a particular
> relationship in the bridge.
>
> Thanks,
> Jim
>
>

[Chris Mungall]

This is fixed, and there is also a junit to pull back a nested
compositional description from the db

[Chris Mungall]

Yep, I filled in your method

We could probably do with more convenience methods for building CDs,
particularly those of a genus-differentia pattern. CDs can be any OWL
Description, but we only need a subset of this the majority of the time

On Sep 19, 2008, at 11:17 AM, Jim Balhoff wrote:

> Thanks Chris. I wrote something like "public
> CompositionalDescription translateOBOClass(OBOClass c)" yesterday,
> so it will be helpful to compare how you did it.
>
> - Jim