loading MeSH and GO .owl files onto protege3.1

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

loading MeSH and GO .owl files onto protege3.1

Sira-2
Hi all,

this has become a big trouble. I don't know how to investigate into
this problem. I've tried creating a protege project from an existing
source (mesh.owl from http://www.fruitfly.org/~cjm/obo-download/ and
go.owl from http://www.geneontology.org/GO.downloads.shtml#ont). Both
of these files were loaded successfully. However, when i tried to save
the project, it ran into infinite loop of failure in saving and it
would not save .pprj file for me (neither mesh or go). Now I am stuck,
I've consulted with Chris Mungall and we sort of suspected a bug in
protege (SWOOP, another ontology editor, let me load and save
successfully). Both MeSH and GO are major ontologies in biomedical
field that a bug like this in protege would create such trouble. Is
there a way to investigate into the matter of how to fix this issue?
Thank you very much.

Sincerely,
Sira

-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: loading MeSH and GO .owl files onto protege3.1

Christine Golbreich
Hi Sira

>both MeSH and GO are major terminologies/ontologies in biomedical field

Yes! but ...
owl files that may be semantically incorrect have not a major interest !

there is a warning on the download site:
---
a.. Generated from obo-xml using oboxml_to_owl.xsl. See Mapping OBO to OWL
for details of mapping.
Note:certain assumptions are made in the mapping that may be *incorrect*
---

I am not sure of  the "ontological" quality of the mesh.owl file
from http://www.fruitfly.org/~cjm/obo-download/ ...

You may have noted for example that the original MeSH taxonomy is not
organized according an "is-a"(subClassOf) relation but a "father-child"
(broader-narrower) relation which is a mixture of different viewpoints.
It is the reason why for example a first step in our migration of MeSH to
OWL was to "clean" the MeSH taxonomy, in distinguishing between the
'part-of' and the 'is-a' relationships (the MeSH Anatomy, Biological
Sciences and Geographic Locations hierarchies are processed separately ,
etc.).
You may have a look at the paper Representing the MeSH in OWL: Towards a
Semi-Automatic Migration presented at KR-MED 2004
http://www.med.univ-rennes1.fr/lim/doc_91.pdf, (although it was a
preliminary work)

I did not look very carefully at the stylesheet, nor at the
mesh.owl file, but if  is is obtained from a straightforward translation of
the original MeSH, for example based on the IDs, the resulting file
may have quite strange features and semantics which may generate unxepected
tool behavior ...
So it would be perhaps worth to investigate what would be related to a buggy
mesh.owl file, which would be much more troubling than a Protégé bug, if any
:-)

PS. it raises a more general question about the usefulness of the owl files
issued of MAPPING OBO TO OWL  based on the proposed XSLT, and of the input
OBO file

> I've consulted with Chris Mungall and we sort of suspected a bug in
> protege

could you precise the bug you suspect ?

best

Christine


----- Original Message -----
From: "sira" <[hidden email]>
To: <[hidden email]>
Sent: Friday, January 20, 2006 8:57 PM
Subject: [protege-owl] loading MeSH and GO .owl files onto protege3.1


> Hi all,
>
> this has become a big trouble. I don't know how to investigate into
> this problem. I've tried creating a protege project from an existing
> source (mesh.owl from http://www.fruitfly.org/~cjm/obo-download/ and
> go.owl from http://www.geneontology.org/GO.downloads.shtml#ont). Both
> of these files were loaded successfully. However, when i tried to save
> the project, it ran into infinite loop of failure in saving and it
> would not save .pprj file for me (neither mesh or go). Now I am stuck,
> I've consulted with Chris Mungall and we sort of suspected a bug in
> protege (SWOOP, another ontology editor, let me load and save
> successfully). Both MeSH and GO are major ontologies in biomedical
> field that a bug like this in protege would create such trouble. Is
> there a way to investigate into the matter of how to fix this issue?
> Thank you very much.
>
> Sincerely,
> Sira
>
> -------------------------------------------------------------------------
> To unsubscribe go to http://protege.stanford.edu/community/subscribe.html
>


-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: loading MeSH and GO .owl files onto protege3.1

Chris Mungall

On Jan 20, 2006, at 5:05 PM, Christine Golbreich wrote:

> Hi Sira
>
>> both MeSH and GO are major terminologies/ontologies in biomedical  
>> field
>
> Yes! but ...
> owl files that may be semantically incorrect have not a major interest  
> !
>
> there is a warning on the download site:
> ---
> a.. Generated from obo-xml using oboxml_to_owl.xsl. See Mapping OBO to  
> OWL
> for details of mapping.
> Note:certain assumptions are made in the mapping that may be  
> *incorrect*
> ---

Hi Christine

This in particular refers to the need to decide whether a restriction  
should be universally or existentially qualified when making the  
conversion from a format in which this is not specified (existential is  
the default, which is consistent the definitions in the OBO relations  
ontology). In this case I don't believe that any of the assumptions  
made by the default mapping were incorrect.

> I am not sure of  the "ontological" quality of the mesh.owl file
> from http://www.fruitfly.org/~cjm/obo-download/ ...
>
> You may have noted for example that the original MeSH taxonomy is not
> organized according an "is-a"(subClassOf) relation but a "father-child"
> (broader-narrower) relation which is a mixture of different viewpoints.

Note that the MeSH ontology that is submitted to OBO (using the obo  
file format) uses is_a relations, so the error is not in the conversion  
to OWL but actually upstream in the original submission (ie in the  
conversion from MeSH to OBO format)

I will contact the submitters, although I'm not sure if much can be  
done until they clean it up and use well defined relations.

> It is the reason why for example a first step in our migration of MeSH  
> to
> OWL was to "clean" the MeSH taxonomy, in distinguishing between the
> 'part-of' and the 'is-a' relationships (the MeSH Anatomy, Biological
> Sciences and Geographic Locations hierarchies are processed separately  
> ,
> etc.).
> You may have a look at the paper Representing the MeSH in OWL: Towards  
> a
> Semi-Automatic Migration presented at KR-MED 2004
> http://www.med.univ-rennes1.fr/lim/doc_91.pdf, (although it was a
> preliminary work)

This seems very useful. If you wish to place your semi-automated OWL  
translation somewhere it can downloaded I can make a link to it from  
the download page.

>
> I did not look very carefully at the stylesheet, nor at the
> mesh.owl file, but if  is is obtained from a straightforward  
> translation of
> the original MeSH, for example based on the IDs, the resulting file
> may have quite strange features and semantics which may generate  
> unxepected
> tool behavior ...
> So it would be perhaps worth to investigate what would be related to a  
> buggy
> mesh.owl file, which would be much more troubling than a Protégé bug,  
> if any
> :-)
>
> PS. it raises a more general question about the usefulness of the owl  
> files
> issued of MAPPING OBO TO OWL  based on the proposed XSLT, and of the  
> input
> OBO file

This particular issue is in the incorrect specification of MeSH in OBO  
format; the mapping from OBO format to OWL is fairly straightforward,  
apart from the existential vs universal commitment.

>> I've consulted with Chris Mungall and we sort of suspected a bug in
>> protege
>
> could you precise the bug you suspect ?

I didn't suspect a specific bug - Sira reported that both files (not  
just the MeSH one - the Gene Ontology OWL file should be free of the  
problems you point out) gave the same results with Protege. Both worked  
fine in SWOOP (but slow!), and I'd previously used them in SWOOP and  
Sesame. Also the fact the problem only manifested when saving seemed to  
justify me passing the buck.

I've just tried loading go.owl, made a random modification, tried  
saving, and got an out of memory error:

CONFIG: Protege 3.1.1 Build 216, JVM 1.4.2_05-141.4, memory=201M, Mac  
OS X, encoding=UTF-8, language=en, country=US
CONFIG: Loaded plugin edu.stanford.smi.protegex.owl - OWL Plugin
CONFIG: Loaded plugin ca.uvic.cs.chisel.jambalaya - Jambalaya
CONFIG: Loaded plugin edu.stanford.smi.protegex.prompt - PROMPT tab
CONFIG: Loaded plugin edu.stanford.smi.protegex.standard_extensions -  
Graph Widget, Table Widget
CONFIG: Loaded plugin uk.ac.ecs.iam.akt.tgviztab - TGVizTab
CONFIG: Loaded plugin uk.ac.man.cs.mig.coode.owldoc
CONFIG: Loaded plugin uk.ac.man.cs.mig.coode.owlviz
CONFIG: Loaded plugin uk.ac.man.cs.mig.coode.protege.wizard - Basic  
Wizards
CONFIG: Loaded plugin uk.ac.man.cs.mig.coode.protege.wizard.owl - OWL  
Wizards
[ProtegeOWLParser] Triple 10000
[ProtegeOWLParser] Triple 20000
[ProtegeOWLParser] Triple 30000
[ProtegeOWLParser] Triple 40000
[ProtegeOWLParser] Triple 50000
[ProtegeOWLParser] Triple 60000
[ProtegeOWLParser] Triple 70000
[ProtegeOWLParser] Triple 80000
[ProtegeOWLParser] Triple 90000
[ProtegeOWLParser] Triple 100000
[ProtegeOWLParser] Completed triple loading after 109576 ms
[TripleChangePostProcessor] Completed lists after 3 ms
[TripleChangePostProcessor] Completed anonymous classes after 205 ms
[TripleChangePostProcessor] Completed deprecated classes after 29 ms
[TripleChangePostProcessor] Completed properties after 7 ms
[TripleChangePostProcessor] Completed named classes after 1498 ms
... Loading completed after 178255 ms
WARNING: java.lang.NumberFormatException: multiple points --  
ProjectPluginManager.afterShow()

Jambalaya, Release: 2.2.0, Build: 15, 2005/07/07 15:04
CHISEL Group, University of Victoria, [hidden email]

Starting Wizards v1.0 beta build 05 ...
Loading standard Protege wizards library
Loading OWL wizards library
Jan 21 00:51:25 adsl-68-126-147-90 last message repeated 20 times
apple.awt.EventQueueExceptionHandler Caught Throwable :  
java.lang.OutOfMemoryError
java.lang.OutOfMemoryError
Jan 21 01:01:25 adsl-68-126-147-90 last message repeated 20 times

---

perhaps this is related to Sira's problem. Seems odd it should only  
manifest when saving.

I can investigate more after the weekend.


>
> best
>
> Christine
>
>
> ----- Original Message -----
> From: "sira" <[hidden email]>
> To: <[hidden email]>
> Sent: Friday, January 20, 2006 8:57 PM
> Subject: [protege-owl] loading MeSH and GO .owl files onto protege3.1
>
>
>> Hi all,
>>
>> this has become a big trouble. I don't know how to investigate into
>> this problem. I've tried creating a protege project from an existing
>> source (mesh.owl from http://www.fruitfly.org/~cjm/obo-download/ and
>> go.owl from http://www.geneontology.org/GO.downloads.shtml#ont). Both
>> of these files were loaded successfully. However, when i tried to save
>> the project, it ran into infinite loop of failure in saving and it
>> would not save .pprj file for me (neither mesh or go). Now I am stuck,
>> I've consulted with Chris Mungall and we sort of suspected a bug in
>> protege (SWOOP, another ontology editor, let me load and save
>> successfully). Both MeSH and GO are major ontologies in biomedical
>> field that a bug like this in protege would create such trouble. Is
>> there a way to investigate into the matter of how to fix this issue?
>> Thank you very much.
>>
>> Sincerely,
>> Sira
>>
>> ----------------------------------------------------------------------
>> ---
>> To unsubscribe go to  
>> http://protege.stanford.edu/community/subscribe.html
>>
>
>
> -----------------------------------------------------------------------
> --
> To unsubscribe go to  
> http://protege.stanford.edu/community/subscribe.html
>

-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html