Re: blog: semantic dissonance in uniprot

2009/3/22 Oliver Ruebenacker <curoli@gmail.com>:
>     Hello Peter, All,
>
> On Sat, Mar 21, 2009 at 5:48 PM, Peter Ansell <ansell.peter@gmail.com> wrote:
>> 2009/3/22 Egon Willighagen <egon.willighagen@gmail.com>:
>>> Chemists are not interested in single molecules (well, most are not,
>>> but with increasing nanotechnology...). I was told recently that upper
>>> ontologies have proper mechanisms to point out the difference between
>>> (in Java terminology) objects and classes, or instances and concepts.
>>
>> There is that possibility.
>
>  There is an infinite number of possibilities. What is the criteria
> for being relevant?

The possibility I was referring to was the option for going through
and defining the properties of the class of molecules as distinct from
a random instance with instance specific variables. I would define the
relevance of that method to be relative to the usefulness of the extra
complexity involved in making sure that in each case the properties
assigned to the random instance of the class were specific/precise
enough to be more useful to scientists than the case where everything
is placed as a property on the class of molecules. As you can't likely
define all the configurations for a given instance of a complex
molecular structure with respect to its environment, this would be
difficult, but if you found an application that was grounded in a
given environment it would be a possibility.

>> Having different identities might be the rational scientific way to do
>> things. They might be caused by different perspectives on the one
>> item, or they might be caused by an actual duality of theory based on
>> an actual inability to describe something in a single theory. Making
>> god-like decisions about which class particular records actually
>> belong to as ontologists might sound fun but in the world case it
>> seems counterintuitive because it doesn't promote progress in both
>> areas concurrently.
>
>  Can you give an example?

For example... How do you define a gene? Tough question I know, but do
you define everything that can be transcribed as being a gene? How do
you really associate genes across organisms considering mutations? If
a gene acts as a promoter for its transcription in one gene but
doesn't in another due to a mutation in the relevant upstream region
can you really say the same sequence is the same gene? Even though its
position in the intracellular space could still be the same, and its
protein 3D conformation is still the same, there would be a duality
where you would then have to redefine the gene as being a regulator
for itself if you wanted to include its molecular functions inside the
cell as part of its property set. It sounds counterintuitive to have
to refer to the gene in two different ways just because in one
organism its function didn't bind to its environment.

One non-biological example I was referring to was the wave-particle
duality for light, where if you wanted to be really in depth and
define the light "thing", then you would either contradict yourself
per current theory or you would create a dual distinct model that
violated someones previous god-like decision that ":wave
owl:distinctWith :particle"

Cheers,

Peter

Received on Sunday, 22 March 2009 21:06:57 UTC