[ome-users] What did OME ever do for us?

[Graham Klyne]

[Resend... the original copy of this message collected an invalid address 
and got bounced.  My apologies if any duplicates are received.]

Jason,

thanks for the various pointers you provided...

At 09:52 18/07/05 -0400, Jason Swedlow wrote:
>As I understand it, you have missed completely the CategoryGroup/ Category 
>functions.

I did.  I'm playing with them now...

>   You will see these on the page where you view a
>specific dataset-- note the text ""To cluster thumbnails by
>category....".  We think of this as an alternative set of user- definable 
>hierarchies.
>
>This is a user defined CategoryGroup, including a set of Categories.
>A screenshot of my own customised annotations, for one of my
>Datasets, is attached.  The bottom line-- OME defines a group of user- 
>defined phenotypes as Categories, and the grouping of these as a
>CategoryGroup.  An image can belong to only one Category **within** a
>CategoryGroup, but can belong to multiple Categories within different
>CategoryGroups.  Both the Java and web UI's support this functionality.

On this basis, I'm thinking that I can create category groups "Gene" and 
"Strain", with categories like "Grip128", "Rpn6", "CG10151" that are gene 
names or flybase identifiers for the gene group, and names like "WT", 
"aly", "can", "comr" for the strain group, which are used by researchers to 
identify wild type and mutant strains of drosophila under 
examination.  Does this sound consistent with the intended use?  I assume 
new catagories (e.g. new gene names) can be added as needed?

Is there any way to edit a category group after it has been created?  I 
just created one and mis-typed the description.

I guess that a category is assigned to an image by applying a custom 
annotation of type "Classification", which asks for a category to be 
selected.  Additional such annotations being added for each axis of 
categorization?  (The documentation you mentioned isn't clear about how one 
applies classification information).

Can an image belong to more than one category in a group?  I just noticed 
thew documentation says this is not possible.  (E.g. we may generate images 
that show in situ expression for multiple strain, for comparative 
purposes.  If I use Strain as a category group, I may wish to associate 
multiple strains with an image.)   Hmmm... what does OME do if multiple 
classification annotations are added to an image with different categories 
within the same group?

[...]

>One last point.  A  MAJOR (in my mind, at least...) limitation of
>this functionality is the limit that there are really on two levels--
>CategoryGroups and Categories.  We are still working on this,
>especially from the UI standpoint, but expect this to be fleshed out
>in later versions.

Hmmm... maybe.  So far, I think the restriction to a single category in a 
group is a greater problem.

If one views Category Groups and Categories as two levels in a class 
hierarchy (in the logical sense rather than the programming-language sense 
of "class"), then the restriction to two levels does seem severe.  If I 
were mapping this into a logic-based ontology, such as RDF/OWL, then I 
think treatment as classes would be appropriate.  Another approach would be 
to treat Category Groups as classes, and Categories as members (instances) 
of those classes that are related to the corresponding images in some 
way.  Under the latter view, the two-level hierarchy seems reasonable.

One way to explore this is to see how they map into logical expressions, 
such as RDF/OWL (which we intend to do):
(a) for the former view, treat Category Groups and Categories as classes, 
with corresponding subclass constraints.  Ultimately, the members of these 
classes would likely be the images.
(b) treat Category Groups as defining both classes and binary relations 
between images and members of those classes (in RDF, these would be classes 
and properties)

(I note that neither formulation requires restricting an image to at most 
one category in a category group -- I think the reason for that constraint 
may be to avoid complications in presentation.)

I see practical rather than fundamental advantages and disadvantages to 
each approach:
Formulation (a) means that the structure of category groups and categories 
would be amenable to reasoning using existing Description Logic based 
reasoning tools.
Formulation (b) would make it easier for the categories in a group to be 
open-ended (e.g., following the gene example from above, it would be easier 
to add new genes as they are introduced into the experimental environment).

#g


---
Graham Klyne
Image Bioinformatics Research Group (http://www.bioimage.org/)
Department of Zoology, University of Oxford
South Parks Road, Oxford OX1 3PS, UK
E-mail: <Graham.Klyne at zoo.ox.ac.uk>
Direct phone: +44-(0)1865-281991
Departmental fax: +44-(0)1865-310447