[MOBY-l] Re: BIoinformatics Protocols Ontological KnowledgeBase?

[Phillip Lord]

>>>>> "Elia" == Elia Stupka <elia at tll.org.sg> writes:

  >> I'm not sure that this makes a huge amount of difference.

  Elia> See my later mail, I agree now.

  Elia> As you say, we all need to describe data going in, data coming
  Elia> out and the process (if one analysis) or processes (if a
  Elia> pipeline) that are applied to it, so the ontology we need is
  Elia> identical, I agree.


  Elia> So I think the only slight difference is in the focus, not in
  Elia> substance, i.e. while we obviously need to work on the
  Elia> ontology and the plumbing what we would like to see developed
  Elia> is an actual resource that would collate and curate
  Elia> established protocols (workflows,etc.)  to extract biological
  Elia> information from raw data as well as possibly allow execution
  Elia> of those protocols on a local server and commenting/ranking of
  Elia> protocols by bioinformaticians.


Do you mean curate established protocols, or descriptions of these
protocols? Within mygrid we have generated a relatively large ontology
describing many tools within bioinformatics. The paper on this is
available at...

http://www.cs.man.ac.uk/~wroec/mygrid/mygrid_service_ontology_02.pdf

The ontology is probably available on the web somewhere as well. If
you want it, I can find it for you. It may be at the level you need,
it may not. 

The issue of ranking and commenting on protocols, is also I think very
interesting, and important, although whether its possible to do this
in structured way, I don't know. 



  >> We would like to extract good descriptions of a workflow from its
  >> component

  >> parts, although its not clear how far this is possible at the
  >> moment.

  Elia> We hope that BioPipe and its XML will make it a step closer
  Elia> though as yet it is not as abstract as you would probably like
  Elia> it to be.

It may do. But I think that there are some fundamentally difficult
questions here. 

To extract a description should be easy. To extract a good one is much
harder. 

So for instance, if we describe our individual services in terms of
inputs and outputs, then for a pipeline the input is just the input of
the first service, and the output is just the output of the last
service. No problems. 

However, within mygrid we also describe what services do. So, for
instance, Blast performs an alignment. Now say we have a two step
pipeline, the first of which reformats a file, and the second does a
blast. The biologist will know that the main idea of the pipeline is
to do a blast, and the reformatting is secondary, but it can't be
deduced from descriptions of the services. 


  Elia> Do you have a meeting/workshop scheduled soon? What is the
  Elia> status of Moby?

Other people are better placed to comment on this I think....

Cheers

Phil