Re: Problem with classifying the Human Phenotype Ontology

Hi Andrea,
Those of us who work on the HPO (OWL) use ELK for classification, and most other ontologies and it works quite well.
Cheers,
Melissa


On Aug 4, 2014, at 5:40 AM, Matthias Samwald <matthias.samwald@meduniwien.ac.at<mailto:matthias.samwald@meduniwien.ac.at>>
 wrote:

> TrOWL I have tried, but I have the impression it doesn't really make a classification upfront, but rather incrementally on demand. It's just an impression, but it classified HP in no time ;)

It is supposed to classify everything. Maybe you don't have a problem at all. ;)

 - Matthias


Am 04.08.2014 14:16, schrieb Andrea Splendiani:
Hi,
I didn't see the BioHackathon ML message. I have just realised my mail setup is a bit messed up...
TrOWL I have tried, but I have the impression it doesn't really make a classification upfront, but rather incrementally on demand. It's just an impression, but it classified HP in no time ;)

I will give a try to ELK and Konclude.
What I am bit puzzled with is: this is a largely used ontology. The issue of unfeasible classification should have come up already. Either I am doing something wrong, or nobody uses the OWL version (or I'm not good at googling).

best,
Andrea



On Mon, Aug 4, 2014 at 2:03 PM, Matthias Samwald <matthias.samwald@meduniwien.ac.at<mailto:matthias.samwald@meduniwien.ac.at>> wrote:
Hi Andrea,

I remember you got the recommendation to try ELK on the Biohackathon mailing list. Is ELK not working for you?
You might also want to give TrOWL a try if ELK is not working for you for some reason. Konclude might also be an option as it seems to outperform most other reasoners, but it does not have a Protege plugin (don't know if this matters to you). You can also have a look at the recent results of the OWL reasoner evaluation here:
http://vip.cs.man.ac.uk:8080/live.html

I have not worked with HPO yet, so those are just some general recommendations.

Best,
Matthias



Am 04.08.2014 13:53, schrieb Andrea Splendiani:

Hi all,

I have stumbled onto a problem for which I would like to hear from your experience.

In a project, I am using the Human Phenotype Ontology (http://www.human-phenotype-ontology.org/).
For the sake of the project, I really only need the is_a structure of the ontology, but as an OWL version was existing, and as we have anyway an RDF framework to integrate data, I was thinking of using this version.
The OWL version is not a simple representation of the is_a structure, as it is including axioms to map phenotypes to, from a quick inspection, anatomical parts and "qualities".

Now, as with any ontology, I was at first trying to classify it. This is an ontology (with imports) of around 20k classes (<200k axioms, ~60k logical axioms). It is big, but not huge.
I simply cannot classify it in any reasonable time.
I have tried a variety of reasoners and, in my longest wait, I have waited for days but we are under 1%).

Does anybody have experience in classifying it ?

If classification is unfeasible, than which use cases does the OWL representation cater to?

best,
Andrea Splendiani





Dr. Melissa Haendel

Assistant Professor
Ontology Development Group, OHSU Library
www.ohsu.edu/library/ontology<http://www.ohsu.edu/library/ontology>
Department of Medical Informatics and Epidemiology
Oregon Health & Science University
haendel@ohsu.edu<mailto:haendel@ohsu.edu>
skype: melissa.haendel
503-407-5970