search

BFO-ontology / BFO

BFO repository including source code and latest documents
Creative Commons Attribution 4.0 International
250 stars 42 forks source link

request for reviewing ontology using BFO 2.0 OWL converted from BFO 1.x #180

Open zhengj2007 opened 8 years ago

zhengj2007 commented 8 years ago

From zhengj2...@gmail.com on June 03, 2013 13:27:54

Beta Cell Genomics Ontology (BCGO) is an application ontology built for the Beta Cell Genomics database ( http://genomics.betacell.org/gbco/ ). The ontology is developed using Basic Formal Ontology (BFO) as upper ontology, Ontology for Biomedical Investigations (OBI) as ontology framework and integrated subsets of multiple OBO Foundry (candidate) ontologies. BCGO contains 2384 classes referencing to 24 various OBO Foundry ontologies (BFO, 20 reference ontologies, and 3 application ontologies) and over thousands of logical axioms (not class A subClassOf class B) defined in OBI, UBERON, CL, and PATO. Details of BCGO ontology, please see https://code.google.com/p/bcgo-ontology/ BCGO was converted to BFO 2.0 (graz release) using BFO converter developed by He Group ( http://bfoconvert.hegroup.org/ ). The mapping file of BFO1.1, BFO pre-graze version, and OBO-REL to BFO 2.0 is: https://obi.svn.sourceforge.net/svnroot/obi/trunk/src/ontology/obi-view/obi-bfo2/TermsMapping-BFO2.xls For temporal relations, 'at some time' were used except 'has continuant part/part of continuant' which used 'at all time' for keeping transitive characteristics desired for most of OBO Foundry ontologies.

Details of conversion has been reported on 2013 BFO meeting, see http://ontology.buffalo.edu/bfo/2013/BFOconverter.pptx The BFO 2.0 graz release compatible BCGO OWL file is avaiable on: http://bcgo-ontology.googlecode.com/svn/trunk/release/20130214/bcgo_basis_bfo2.owl We would like BFO OWL developers can review this OWL file to see whether it is biological and logical correct. I think it would help us to see whether OBI, UBERON, CL and PATO can use BFO 2.0 and expand to other OBO Foundry ontologies.

Original issue: http://code.google.com/p/bfo/issues/detail?id=180

zhengj2007 commented 8 years ago

From alanruttenberg@gmail.com on June 05, 2013 11:01:33

Stefan will look. Barry will with some help from his friends. Earlier versions in release directory https://code.google.com/p/bcgo-ontology/source/browse/#svn%2Ftrunk%2Frelease%2F20130214

Status: Accepted
Owner: steschu@gmail.com
Labels: Milestone-BFO2-Release

zhengj2007 commented 8 years ago

From steschu@gmail.com on June 25, 2013 09:39:44

Here is the result of my review. Not all observations are related directly to temporalized. I do not distinguish between the sources of the axioms.

  1. There are numerous relations between continuants for which an all times / at some time distinction should be required:
    • connects and subclasses
    • continuous with
    • distal_to and subclasses
    • dorsal_to
    • ventral_to
    • posteriorly_connected_to and subclasses
    • secreted_by
    • secretes (should be at some time)
    • concretizes (could be at all times)
    • isAbout

  2. The role chain for develops from is problematic:

    'part of continuant at all times' o develops_from -> develops-from

This axiom is not true for anatomical entities which develop from distinct origins, e.g. the placenta:

3. There are axioms that anticipate of the existence of future entities

4. subrelations of has_part / part_of not classified as such and not temporally qualified

5. "lacks" relations as object properties (between individuals) are not meaningful:

(145 uses of lacks* )

  1. Dispositional meaning implicitly embedded in the object property, produces wrong existential assertions:
    • Mammary gland equivalentTo Gland and secretes some Milk
    • 'erythroid lineage cell' subClassOf (capable_of some 'erythrocyte differentiation')… (108 uses of capable_of,
  2. Category problems: Inceased_nucleus_size subClassOf Nucleus
  3. General observation about anatomical entities: Most anatomical entities can maintain their identity even severed from the whole they are part of in vivo. In most cases they are no longer functional, but may be subject to lab investigations and manipulations. As a consequence, one could argue that nearly all part-of relations between anatomical entities are of the type "continuant_part_of_at_some_time", unless the meaning is restricted to entities in vivo. Even in this case "kindney subClassOf 'part of continuant at all times' some abdomen" would be incorrect due to the possibility of transplants (generic permanent parthood).

One could consider axioms like

A and bearer_of some in_vivo subClassOf part_of_at_all_times some (B and bearer_of some in_vivo)

to express anatomical hierarchies (however, with the exception of anatomical entities that can be transplanted)

zhengj2007 commented 8 years ago

From zhengj2...@gmail.com on June 25, 2013 14:48:30

Stefan, thanks for your review.

The issue #1 is the problem of automatic BFO version convertor. Need to be taken care in the future.

The other issues are more related to UBERON and CL ontologies. The reason we use 'at-all-time' because UBERON and CL would like to keep transitive property of relations (e.g. part of/has part) that defined in BFO 1.1. And it is the main issue for them to transit to BFO 2. It might be worth to let Chris look at it.

zhengj2007 commented 8 years ago

From cmung...@gmail.com on June 25, 2013 15:36:47

uberon tracker: http://purl.obolibrary.org/obo/uberon/tracker cl tracker: http://purl.obolibrary.org/obo/cl/tracker thanks for the review stefan. I transferred some issues that were not to do with TRs to the relevant trackers https://code.google.com/p/cell-ontology/issues/detail?id=76 - increased nucleus size https://code.google.com/p/cell-ontology/issues/detail?id=77 - lacks https://code.google.com/p/cell-ontology/issues/detail?id=78 - Dispositional meaning implicitly embedded in the object property, produces wrong existential assertions

zhengj2007 commented 8 years ago

From cmung...@gmail.com on June 25, 2013 15:57:07

http://purl.obolibrary.org/obo/uberon/tracker/282 Truth of assertions concerning non-canonical anatomy

zhengj2007 commented 8 years ago

From alanruttenberg@gmail.com on June 26, 2013 11:57:28

Wierd issue that one. Recorded on the uberon tracker, but closed to comment?

My comment: if we need a generic part of true when "in vivo", this could be defined as a different relation. BFO defines some but not all relations that ontologies need.

However, I doubt even that is always true.

FWIW. I'd need to check, but I believe that the approach I gave for generic parthood could be adapted to say this.