Re: [Chebi-ontology] Oxygen-18 to isotope: instance or subclass?

[Emw <emw...@gm...>]

Alan, Stefan,

Apologies for the delayed reply.

I think minimal requirements for querying and inference from a
representation with isotopes would include these criteria:

A) Retrieve all entities like oxygen-16, oxygen-17 and oxygen-18 as
isotopes of oxygen.

B) Retrieve all entities like helium-1, germanium-73, and the entities in
(A) as isotopes.

C) State axioms about isotopes, e.g. "oxygen-18 *neutrons* 10",
"oxygen-18 *natural
abundance* 0.2%", "oxygen-18 *spin* 0".

 D) Avoid any potential subclasses of the isotopes in (A) and (B) without
relying on those subclasses being leaves in a subsumption hierarchy.

E) In subclasses of a given isotope, e.g. carbon-14 in fossil fuels,
carbon-14 in human bodies, infer facts using statements made for the
superclass, e.g. carbon-14.

F) Retain syntactic and philosophical compatibility with BFO, considering
the definition of "instance" in Smith and Rosse 2004 [1]: "Instances are
individuals (particulars, tokens) of special sorts... bound to a
specific... location in space and time."

G) Retain compatibility with ChEBI.

Statements like the ones Alan writes about:

oxygen-16 isIsotopeOf oxygen-18
oxygen-18 isIsotopeOf oxygen-16

Might technically establish disjointness, but they would need to be hidden
away from the UI, as they are intuitively incorrect.  (The statement
"oxygen-16 is an isotope of oxygen-18" would quite reasonably strike a
reader as false if seen in the Wikidata page for oxygen-18,
https://www.wikidata.org/wiki/Q662269.)  To be honest, I don't know how the
statements above establish disjointness -- could you point me to where I
could read more on that technique?

It is still not clear to me whether the statement (i.e., Option B)

    oxygen-18
*        subclass of* oxygen
        *instance of* isotope

      (where *instance of* is rdf:type and *subclass of* is rdfs:subClassOf)

is compatible with ChEBI and BFO.  It seems that everyone who has commented
on the matter here thinks that modeling things like isotopes like oxygen-18
as a class is preferable to modeling it as an instance, but "not
preferable" does not necessarily mean "not compatible".

If there is one take-away I would like from this discussion, it would be a
clear answer to the question:

Is the model in Option B above syntactically and semantically compatible
with ChEBI and BFO?

Stefan's most recent comment addresses this directly.  His paper *The
Ontology of Biological *
<http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2718636/>*Taxa*
<http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2718636/> is helpful (and
fascinating) background to consider.

His attached file TwoLevelOntology.owl file in Protege 4.3.0 reveals
proposes a novel approach not discussed in that paper.  I am struck by a
few things in that OWL file:

   1. It replaces key built-in OWL 2 DL vocabulary with
   application-specific analogs.  Unlike all OWL exports of OBO ontologies I
   have seen, OWL's rdfs:subClassOf is not used in place of OBO's* is_a*,
   and rdf:type is not used in place of *instance_of*.

   2. It introduces a new basic membership property -- isClassifiedAs --
   adding to the familiar isA and instanceOf properties.

   3. The class hierarchy is entirely inferred, and not declared explicitly.

TwoLevelOntology.owl is interesting, but seems idiosyncratic in its
approach to classification.  Is there a way to account for metaclasses and
maintain compatibility with BFO and ChEBI, while also not abandoning the
foundational equivalences *instance of = rdf:type* and *subclass of =
rdfs:subClassOf*, not introducing a third basic membership property, and
not relying entirely on inference to construct a class hierarchy?

The alternate approach in TwoLevelOntology.owl seems to be, as Alan
suggests, probably more mechanism than Wikidata needs, and, in my opinion,
too complex for that community to wrap its head around.

Really, the core unresolved question in this thread as I see it is whether
the statement in Option B above is syntactically and philosophically
compatible with BFO and ChEBI.

Can Barry Smith's interpretation of the metaclass "chimpanzee" as the
instance "total population of all chimpanzees" (read: object aggregate [2]
of all chimpanzees) be applied to Option B, making "oxygen-18" construable
as the object aggregate of all oxygen-18 atoms?  Section 3.3 of *The
Ontology of Biological Taxa* explores that theme further.
Thanks,
Eric

[1] Barry Smith, Cornelius Rosse (2004). *The Role of Foundational
Relations in the Alignment of Biomedical Ontologies*.
http://ontology.buffalo.edu/medo/isa.pdf
[2] Object aggregate, BFO 2 (Graz) draft reference.
http://bfo.googlecode.com/svn/releases/2012-07-20-graz/BFO2-Reference-for-html.htm#_Toc313270697
[3] Stefan Schulz et al. (2008).  *The Ontology of Biological Taxa*,
Section 3.3: Biological taxa as populations*.  *
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2718636/#SEC3.3