[MOBY-l] An Ontology question

[Phillip Lord]

>>>>> "Frank" == Frank Gibbons <fgibbons at hms.harvard.edu> writes:

  Frank> So far, we've done this by hand, but clearly it would make
  Frank> much more sense to either automate it, or better yet, find a
  Frank> way to link from MOBY's object registry to a URI of
  Frank> PSI-MI's. Unfortunately, being an RDF newbie, I'm not really
  Frank> sure of the best way to do this, or even if it's possible. I
  Frank> know that Mark W. is meeting with the myGRID folks this
  Frank> summer, to discuss ways to re-use their massive ontology, so
  Frank> that may obviate any other plans.

  Frank> It strikes me that it might not be desirable to include any
  Frank> ontology in all of its gory detail, at least in the absence
  Frank> of appropriate tools to navigate and visualise MOBY's
  Frank> ontologies. Martin Senger has written some really useful
  Frank> tools for that, but they generate static images rather than
  Frank> interactive ones, so once below the spatial resolution of
  Frank> your printer/screen, it's not possible to read it, and it's
  Frank> close to that right now.


One of the issues with any complex ontology is how to visualise it,
and how to present it to the different kinds of user who might want to
use it. Within GO, for example, the annotators, GO curators and end
users all seem the same ontology. But its doesn't necessarily need to
be this way. (Indeed GO slims are an acknowledgement of this). I think
that mygrid ontology is too complex to present to the end user, per
se, but having a layer which removes most of the complexity for
display, well still leaving this complexity so that the underlying
machinery can use it seems like a good approach.

Personally I would like to realise the mygrid ontology in a number of
inter-related forms; the idea would be to have an underlying OWL-DL
based ontology (so that we can use reasoner support, which helps in
the building of the ontology). Then we use this to generate
representationally simpler ontologies. So a stripped down OBO style
ontology (modelled as a DAG). Moby-s would probably need something
half way in between--it needs data type information (sequence has-a
string) in the ontology which would probably be out of place in a OBO
ontology. So you'd have to add something back in. All of which makes
development of the ontology somewhat more complex, but will,
hopefully, make it easier for the end user. Which is probably the way
that it should be. 

I managed to talk with Steffen about this earlier in the month, and
should get a chance to talk with Mark at ISMB. I guess we should try
and advertise a meeting as a BOF, as it looks like others might be
interested. 

Incidentally, for those of you going to ISMB, who are interested in
ontologies, I have to say, in an entirely unbiased manner, that the
bio-ontologies SIG (http://bio-ontologies.man.ac.uk/) looks
wonderful. Everyone should attend!

Cheers

Phil