Re: [Gusdev-gusdev] Re: Phenotype tables - Interactions

[<cra...@pc...>]

Arnaud-

> I didn't include anything about genetic interactions even if in the
> future we will want to store them. I've reviewed the interaction table
> and I've got some thoughts about this table.
> 
> Genetic interactions may involve more than one effector/target. If we
> want to make the Interaction table generic, we need to store more than
> one effector and more than one target. I don't have any use cases yet,
> but I can ask around one if needed.

This makes sense to me, since not all groups of effectors or targets will 
necessarily be Complexes.

> An extra controlled vocabulary is needed. This controlled vocabulary
> will be used to classify the behaviour of the effector for a given
> interaction.
> e.g. Allele1 "inhibits" the expression of Allele2.

This also sounds like a good idea; I don't see anywhere that we represent
this currently.  

> Regarding physical interactions, there are two cases in which it will be
> useful to annotate them:
>     * Transient interactions associated with a function, e.g. a protein
> regulating the transcription will be interacting with DNA.
>     * Structural interactions involved in the formation of a complex. In
> that case, we can associate component interactions with the complex they
> are involved in. Some of these interactions are experimentally
> characterized, others are hypothetical.

I don't think we have to make any changes to the Interaction table in order
to use it for representing physical interactions, right?  Presumably the 
physical interactions will simply be a subset of the interactions enumerated
in the controlled vocabulary that we're going to create.  The only thing I
noticed when looking at the DoTS::Interaction table is that there's no way 
of representing interactions for which "direction" is not a meaningful 
concept.  (I assume there will be such interactions, particularly when we 
consider physical interactions.)  So maybe we could add a "has_direction" 
column, which would only be non-NULL in the case that direction_is_known > 0
(or which should indicate that the direction_is_known column should be 
ignored, if set to 1.)

> Currently in GUS, a complex is a set of components. Would it be possible
> to associate a complex with a set of interactions as well ?

It should be, since the DoTS::Interaction table can reference any other table
in the database (including Complex.)  Or are you asking about an explicit
representation for an "InteractionSet" (whatever that would mean)?   I'm not
sure I completely understand this question.

> The other point I didn't mention in my previous email was the review of
> the phenotype table. Would it be possible to associate phenotypic data
> with GO terms ?

I believe we should be able to use the DoTS::GOTermAssociation table to 
associate GO terms with the appropriate rows in the Phenotype table.  In
gusdev right now we're using a special-purpose table called ECGOFunctionMap
to represent the mapping between the enzyme classification numbers and the
GO Function terms, but I believe that the only reason we did this is because
we didn't have the GOTermAssociation table to work with in GUSdev.

Jonathan

-- 
Jonathan Crabtree
Center for Bioinformatics, University of Pennsylvania
1406 Blockley Hall, 423 Guardian Drive Philadelphia, PA 19104-6021
215-573-3115