Generative Methodology thread

Sunday, June 11, 2006

On ontological modeling of expression

The metaphor between gene, cell and social expression à [217]

On Formal verses Natural systems à [206]

To allow this line of analysis to continue, I would like to suggest that natural category formation within "levels of organization" is a key to making distinctions such as:

"This is a general concept: once a molecule comes into existence in a
subcellular location, its possible fates are determined by its location,
not its history

This statement "has truth" because (I conjecture) the levels of organization "slower" than the substrate somehow reproduce "exactness" for purposes of "function". This need not be "gene" driven, but could be. As Cristina points out, this exactness is absolutely necessary to biology; and thus one might assume a mechanism.

Atoms and molecules exhibit this same phenomenon, and there is no "gene". The hypothesis I am putting forward is radical, that replicator mechanisms exists at all level of physical organization.

One might suppose that replicator mechanisms also have different properties depending on the phenomenon, but in all cases emergence of a function occurs through an aggregation process. I also feel that the "memetic science" has about 3/4 of the theory wrong, and has hurt the case for stratification, formal modeling of emergence, and clear minded investigations on the role that natural category formation plays in the expression of social discourse (another type of signal pathway expression - with (conjectured) replicator mechanisms).

http://www.bcngroup.org/area3/pprueitt/kmbook/Chapter6.htm

The water molecule does not "care: which Oxygen atom it has, it just wants one.. i.e. something that fits the natural category of Oxygen. The same is true of reactions that need H2O as part of the participation of the reaction.

I say "has truth", because there are fine detail where history (the historical process that produce a specific ATP, for example) may have a role. This might likely be atypical, due to the replicator mechanism associated with treating all "elements" of a natural category as "functional the same". But we, as a science community, have not "looked", and so we do not know. We are troubled with things like intelligent design, and the mythology we as scientists live in. These are open questions, not final science; and we must treat them as such.

Now I know that this is theory, and I know that others agree that this is interesting theory and can be grounded in a literature, and I cite G Edelman because I know this work well. I also cite the works by Nicolas Rashevsky and by his student Robert Rosen (on mathematical biology and category theory). The work I cite on Soviet QSAR and quasi axiomatic theory is simply not known, and no one cares - but it is there.

I would like to develop this theory more fully, in light of the clash that I have seen coming for BioPAX. (I will travel anywhere there is a position).

Others seen these issues also as illustrated by

Communication to BioPAX: from Cristina Thalhammer-Reyero

I think Paul is completely right. In order to model biological processes and pathways, the two important structures or nodes to be modeled are the "interactions/reactions" and the "pools" of entities that participate in them. The class "physicalEntity" seems to represent the "description" of the general characteristics of a particular class of entities, such as ATP, as it may be found in a database. Therefore, it is inappropriate to say that a "pool" is an instance of a "description". The problem of the whole discussion is the origin of BioPAX, focusing on databases rather than biology taking central stage.

The "pool" should be a class in its own right, and subclasses of "pool" could be defined to describe different types of pools, such as pools of soluble (concentration), membrane-bound or complex-bound (quantity) molecules; or single or attached cells. The entities comprised in a particular "pool" may be available to participate in different "interactions/reactions", which would be competing for the variable "number of entities" in the pool. You still need another structure on the interactions/reaction side to accommodate the role [and the rate of production or consumption] that a given "number of entities" play in the particular "interaction/reaction". That is, you need connectors between each "pool" and the various "interactions/reactions" in which its entities may be participants, and between each "interaction/reaction" and its various participants' "pool". For example, "reaction A" could provide "x ATP molecules[/sec]" to "Pool 1", in which case the "x ATP molecules" would be a "product" of "reaction A", while at the same time "y ATP molecules[/sec]" from "Pool 1" could participate in "reaction B" as a substrate, "z ATP molecules[/sec]" from "Pool 1" could participate in "reaction C" as an substrate, and "w ATP molecules[/sec]" from "Pool 1" could participate in "reaction D" as a cofactor. Actually, the "pool" does not need to know much about the characteristics of its contents, other than the name associated with the "description" of its contents, in BioPAX the "physicalEntity".

This may fit well with your use of SKOS, but in a way that is the reversal of what you propose. The "pool" is the real player (node) in one or more pathways, and the "pool" would be annotated with the name of the "physicalEntity", for anyone who wants to use it as a link to learn more about the characteristics of the components of such "pool".

Cristina's suggestion that The problem of the whole discussion is the origin of BioPAX, focusing on databases rather than biology taking central stage.
is a critical issue.

Because of the funding and the momentum towards W3C standards, these issues will get resolved by young computer scientists who also have a good education in neuroscience or signal pathways, and who believe that W3C standards for web-ontology is to biology what mathematics is to mechanical engineers. This is simply a mistake.

I can not criticize this new generation, because they are following what the professions are doing and where the funding has been. I do blame NSF, NIST and the other funding mechanisms.

But BioPAX started out as a way to integrate databases. Now we have forgotten this and wish to declare that the ontological structure of biology is well specified by the constructions that can be achieved by OWL Full - some even pretend that one can get ontological models of signal pathways using DL. We mislead ourselves if this happens.

-----Original Message-----
From: biopax-discuss-bounces@cbio.mskcc.org [mailto:biopax-discuss-bounces@cbio.mskcc.org]On Behalf Of Peter D'Eustachio
Sent: Monday, June 12, 2006 9:19 AM
To: biopax-discuss@cbio.mskcc.org
Subject: Re: [BioPAX-discuss] Three proposals for BioPAX

I would like to use the ATP example to go back to the biological processes
that we are trying to capture, to state a kind of user requirement to go
with the discussion.

Almost always, all ATP molecules found in the cytosol of a mammalian cell
are functionally indistinguishable. That is, whether they arose as the
output /product/right_side of a substrate level phosphorylation event, some
sort of exchange of phosphate groups among nucleotides, or a transport event
that delivered them to the cytosol from someplace else (typically, the
mitochondrial matrix), they are all equally at risk for use as the
input/substrate/left_side of a a reaction that consumes ATP. That's because
the aqueous content of a cellular compartment is almost always treated as
though it were homogeneous, like the contents of a well-mixed test tube.
Therefore, distinguishing ATP[synthesized by phosphoglycerate kinase] from
ATP[synthesized by pyruvate kinase] is not just unnecessary, it actually
misrepresents most available experimental data.

On the other hand, the subcellular localization of that ATP is a critically
important functional attribute. The mitochondrial membrane, for example, is
completely impervious to ATP - ATP can cross only via a channel formed by a
specific, highly regulated, membrane-bound transport protein, and if that
transport process fails, ATP-dependent processes stop.

This is a general concept: one a molecule comes into existence in a
subcellular location, its possible fates are determined by its location,
not its history. (Sometimes, the only way a molecule can get to a particular
location is by having a particular history so the distinction doesn't
matter, but that's not a counterexample.)

Peter D'Eustachio
Reactome

> Nicolas Le Novere wrote:
>> On Mon, 12 Jun 2006, Matthias Samwald wrote:
>>
>>> No problem. If I understand you correctly, in your view physicalEntity
>>> acts as a kind of 'class' of molecules (say, the class of ATP
>>> molecules), while physicalEntityParticipant is an 'instance' of this
>>> class that exists at a certain spatio-temporal location.
>>
>> Yes. Hence the attribute "cellular-location" of
>> physicalEntityParticipant. Although I do not think that was intended
>> to at the origin (I wasn't involved at the time). A broad guess based
>> on the name would be that physicalEntityParticipant was meant to be
>> the equivalent of SBML speciesReference, that is a link to a pool,
>> used in a specific interaction. This is why they have the attribute
>> "stoichiometric-coefficient". But the attribute "cellular-location"
>> screwed that. Actually physicalEntityParticipant is very sensible. One
>> needs it. But "cellular-location" should be move to something higher,
>> that represent all the physicalEntityParticipant in the same
>> place. This something higher would be the pool. Since in the
>> definition of PhysicalEntity, we find "A pool of entities, not a
>> specific molecular instance of an entity in a cell", I would be
>> inclined to move cellular-location in there. But then one needs
>> something even higher, that defines a class of physcical entity,
>> regardless of their location (the equivalent of SBML L2V2 SpeciesType).
>>
>>> First of all, if physicalEntityParticipant is actually more 'concrete'
>>> than physicalEntity, why is it a mere utilityClass?
>>
>> Entirely agree. It is very much a core element of BioPAX descriptions.
>>
>>> Furthermore, the current ontology allows the use of both physicalEntity
>>> and physicalEntityParticipant as the value of some properties. This can
>>> lead to some ambiguity and makes it a bit harder to write tools -- but
>>> not only that, it also leads to problems of ontological consistency.
>>> When a physicalEntityParticipant is something different from a
>>> physicalEntity, how can they take the same role?
>>
>> Indeed.
>>
> _______________________________________________
> BioPAX-discuss mailing list
> BioPAX-discuss@cbio.mskcc.org
> http://cbio.mskcc.org/mailman/listinfo/biopax-discuss