Just to address one point from David Decraene: (01)
[DC]
> The problem I have however with keeping role-like concepts
completely distinct
> (different hierarchy) from what many refer to as natural kinds is
that you
> often have to make an ugly choice (I'll take proteins as an example
again):
>
> Either you maintain a huge flat list of proteins (tens of thousands)
> all directly related to protein (isa protein) because you are not
allowed
> to organise it's children into role-like hierarchies.
>
> Or you create artificial hierarchies that completely mimic & point
to
> the role-hierarchy in question (e.g. protein as an enzyme, protein
as a
> transferase, versus the role enzyme & transferase) in order to
structure
> the ontology a bit more and make it more manageable.
>
> None of these solutions strike me as elegant/aesthetic nor do they
> aid ontology development.
>
> Internally we do integrate the role-like concepts into the regular
hierarchies,
> but also (try to) define them through necessary and sufficient
relations
> (enzyme: a protein with function-realisation catalysis) and (try to)
establish
> guidelines that aid this as (try to) to avoid ontological
misinterpretations to
> often rise from these situations. (02)
My advice is to boldly go and do what makes sense in your domain,
whether or not you think there may be quibbles about 'Role' hierarchies
because the seem 'artificial'. It is unlikely you will run into
logical contradictions, and in engineering the information
infrastructure, aesthetics should take a second place to
comprehensibility and efficiency. (03)
There are at least three perspectives which make it perfectly sensible
to classify enzymes by their function:
(1) any set of entities with a particular property will logically form
a class (type, sort) of things that have that property. If the
property is of a transient nature, then the entity will be an instance
only while it has that property. The need to handle time intervals is
only one case of the need to handle different contexts in which
assertions (e.g. instanceship) may be true or false. There are several
ways to handle the time-indexing of properties and roles. If you try
hard of course, you can create problems by, for example, adopting a
particular form of four-dimensionalism
(2) For an Enzyme, it is less of a problem than for other things that
look like Roles, because an Enzyme is an Enzyme from the time it is
created till the time it ceases to be capable of performing that
function. If it is of the kind that requires activation, then it may
have multiple states, or these states, if they are the result of
chemical modification, can be optionally treated as subtypes of the
primary class of enzymes that have a particular essential polypeptide
sequence.
(3) things that have evolved by evolution to perform certain functions
can be treated similarly to artifacts that are created by intelligent
agents to perform certain functions. In both cases there is a
teleological property that is intended to be present at the creation of
each instance of that category, and that "design purpose" (intelligent
or evolutionary, whether or not the instance actually has that
property) is a legitimate character by which the instances of that
class can be classified. Thus a folding chair is still a chair, even
if it is folded and stacked where no one can sit on it, because it got
that classification by being created for the purpose of sitting, and it
will stay a chair until it is busted up and can no longer be sat on.
For that matter, a blob of matter is a "person" because it was born (or
conceived) as a result of an evolutionary and biological process that
created organisms with distinctive genetic characteristics; a Person
can be viewed as a Role played by organisms whose cells have some
necessary DNA sequences. What is and isn't a Role, and why it matters,
is far from settled. I find Roles useful, when carefully handled (I
don't classify 'Person' as a Role, but I do have 'Student' as a subtype
of both 'Person' and 'Role'). (04)
I would hesitate to create broad enzyme categories that collect
multiple functions, such as 'LigandAddingEnzyme' - to me that would be
ugly, but if it can be precisely defined, would still be a legitimate
Type (class, category, sort). (05)
I don't foresee any problems in creating functional hierarchies, as
long as the inheritance of properties is strictly adhered to. If it
appears useful and doesn't create any logical contradictions, I would
not overly worry about aesthetic qualms. (06)
But:
> (enzyme: a protein with function-realisation catalysis) (07)
What?? No RNA enzymes? Guess you'll need a supertype of macromolecule
catalysts. ;-) (08)
Pat (09)
Patrick Cassidy
MITRE
260 Industrial Way West
Eatontown NJ 07724
Eatontown: 732-578-6340
Cell: 908-565-4053
pcassidy@xxxxxxxxx
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (010)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (011)
|