BioPAX-OBO Definitions - Cleaned up result of today's meeting

4 views
Skip to first unread message

alanrut...@gmail.com

unread,
May 6, 2008, 6:26:20 PM5/6/08
to biopa...@googlegroups.com
I've shared a document with you called "BioPAX-OBO Definitions":
http://docs.google.com/Doc?id=dthvctw_419fpr93zcv&invite=ff6fmb8

It's not an attachment -- it's stored online at Google Docs. To open this document, just click the link above.
---

Hi, I cleaned up our discussion notes from today. See below.

Group convenes: May 6, 2007, 4PM.

Part 1

Collecting definitions that Elgar or Oliver think are useful.

Elgar:
reaction is usually any response caused by some other event (http://en.wikipedia.org/wiki/Reaction )
chemical reaction is a process that always results in the interconversion of chemical substances (http://en.wikipedia.org/wiki/Chemical_reaction )

Alan asks:  What is a response? What is an event? Why? Hard to use the first definition if we don't understand what the words in the definition mean.

Second one is better, because if we believe "process" means "bfo:process" then we can define chemical_reaction is_a bfo:process and then try to add the additional conditions.

Issue in the second definition is what is "chemical substance"? What is interconversion?

We look at the definition of chemical substance and see: chemical substance is a material with a definite chemical composition

Not so good. Isn't anything something with a definite chemical composition? 

Looking at CheBI CHEBI:23367 molecular entities: A molecular entity is any constitutionally or isotopically distinct atom, molecule, ion, ion pair, radical, radical ion, complex, conformer etc., identifiable as a separately distinguishable entity.

This one seems better. The only problem is "identifiable".  But we could define a granularity and connectivity within it to be a connected to b if a bound to b. Then bfo:objects in this ontology are molecular entities.

Michel: I  also have some definitions in my ontologies - http://dumontierlab.com/index.php?page=ontologies
http://ontology.dumontierlab.com/ChemicalSubstance (chemical substance)
Any material with a definite chemical composition.</owl:Class>

http://ontology.dumontierlab.com/ChemicalReaction (
chemical reaction)
(some back and forth below as Alan and Michel joust)
A process in which at least one chemical
substance is ro:derived_from another. (alan fixes definition ;-) the definitions should be human readable ;-) The definitions should be possible to interpret unambiguously, then easy to understand, if possible. Matter of priority - Michel: well look at the definition, which involves, as its criteria:
Michel: "biologically significant portion"  - chemicals don't have biologically significant portion Alan: Fair criticism. 

Alan:
for definition of derivation, see "
Derivation as a relation between instances" in http://genomebiology.com/2005/6/5/R46
Michel: I think that definition is too vague and/or misleading. Alan: Give me a better one.

Alan says: This is the definition from wikipedia, see above. What does definite chemical composition mean? What doesn't have definite chemical composition?

Michel: The composition of a mixture may not be known precisely.  In some cases, we may make a test, and we know the substance has some properties, but we don't know the exact atomic composition and connectivity.

Alan: So if we are ignorant, then stuff can change from being a chemical substance to not being one? Or from not being one to being one, if we learn something?

michel: If you are ignorant, you would not start by calling it a chemical substance, but maybe an owl:Thing or bfo:Entity ;-)

Oliver's turn:
Chemical substance = group of bound atoms
Bound = separation costs energy
Chemical bond = relationship between two atoms that energy is needed to separate them
Chemical reaction = process that changes bonds between atoms

Elgar's turn:
Fusion/Fission/ and chemical reactions. 
Distinction between chemical and nuclear reactions.
Chemical reactions always contain bond changes, covalent bond changes.

Part 2

Alan, trying to be organized. Goal: Stop arguing over words. Give as many definitions as people use. Worry about what to attach the words to later.

Reaction #n is a subclass of bfo process.

Reaction #1: Process with Quality #2 or Quality #3 changes. aka Nuclear reaction (not for biopax)
Reaction #2: Process with Bond #6 changes   
Reaction #3: Process with Bond #1 changes. Aka: Chemical reaction. Subclass of reaction #2 - because Bond #1 subclass of Bond #6
Reaction #5: Process with Bond#7 change. Aka: Interaction. Example non-covalent protein binding.
Reaction #6: Conformational changes =def  ?? (example: Prion compared to wild type protein) Same chemical composition. Same Bond#1, expect change in Bond #7. Shape quality changes. 
Reaction #7: Process where an atom undergoes a change to its election configuration. Q: Which of the above reactions always have a part which is some Reaction #7? Reaction #3 (waiting for more...)

Bond #N are relational qualities.

Bond #1: Covalent
Bond #2: Hydrogen
Bond #3: ion pair/electrostatic
Bond #4: van der waals.
Bond #5: Hydrophobic bond (Oliver claims subclass of Bond #4)
Bond #6: Bond = Union of Bond #1, Bond #2, Bond #3, Bond #4, Bond
Bond #7: Non-covalent Bond = Bond #6 intersection complementOf Bond #1


Quality #1:  Electron configuration
    Q: What bears it? An atom or a molecule?
    Michel: Alternatively, it could be considered as part. electrons are part of atoms, atoms are part of molecules. 
Quality #2: Number of protons in an atom
Quality #3: Number of neutrons in an atom 

michel: In the absence of changes in covalent bonds, i don't think you have a "chemical reaction". Alan: This is Reaction #3

(Oliver gives examples)
- case: the absorption of an photon by an atom, bringing it to excited state -> chemical or nuclear reaction
- case: fission - the splitting of the nucleus of an atom into parts -> better as a nuclear reaction, rather than chemical reaction.


Michel: alternative definition for chemical reaction:
chemical reaction: A change in the electron configuration of one or more atoms. (
Alan:
or of a molecule?)
nuclear reaction: A change in the proton / neutron composition of an atom.


Group disperses.

Dan Corwin

unread,
May 7, 2008, 2:17:14 AM5/7/08
to biopa...@googlegroups.com
Alan: thanks for writing up these notes.  My $.02 added below.


alanrut...@gmail.com wrote:
I've shared a document with you called "BioPAX-OBO Definitions":
http://docs.google.com/Doc?id=dthvctw_419fpr93zcv&invite=ff6fmb8

It's not an attachment -- it's stored online at Google Docs. To open this document, just click the link above.
---

Hi, I cleaned up our discussion notes from today. See below.

Group convenes: May 6, 2007, 4PM.

Part 1

Collecting definitions that Elgar or Oliver think are useful.

Elgar:
reaction is usually any response caused by some other event (http://en.wikipedia.org/wiki/Reaction )
chemical reaction is a process that always results in the interconversion of chemical substances (http://en.wikipedia.org/wiki/Chemical_reaction )

Alan asks:  What is a response? What is an event? Why? Hard to use the first definition if we don't understand what the words in the definition mean.
Defining those words is a bigger job than BioPAX OBO can properly handle.


Second one is better, because if we believe "process" means "bfo:process" then we can define chemical_reaction is_a bfo:process and then try to add the additional conditions.

Issue in the second definition is what is "chemical substance"? What is interconversion?

We look at the definition of chemical substance and see: chemical substance is a material with a definite chemical composition

Not so good. Isn't anything something with a definite chemical composition? 
No.  It could have an indefinite chemical composition, like "dirt" or "cloth".

Looking at CheBI CHEBI:23367 molecular entities: A molecular entity is any constitutionally or isotopically distinct atom, molecule, ion, ion pair, radical, radical ion, complex, conformer etc., identifiable as a separately distinguishable entity.

This one seems better. The only problem is "identifiable".  But we could define a granularity and connectivity within it to be a connected to b if a bound to b. Then bfo:objects in this ontology are molecular entities.
Identifiable roughly means "named".  If it isn't, who would care about it?  That limitation helps define the scope of what BioPAX can properly address.
Michel: I  also have some definitions in my ontologies - http://dumontierlab.com/index.php?page=ontologies
http://ontology.dumontierlab.com/ChemicalSubstance (chemical substance)
Any material with a definite chemical composition.</owl:Class>

http://ontology.dumontierlab.com/ChemicalReaction (
chemical reaction)
(some back and forth below as Alan and Michel joust)
A process in which at least one chemical
substance is ro:derived_from another. (alan fixes definition ;-) the definitions should be human readable ;-) The definitions should be possible to interpret unambiguously, then easy to understand, if possible. Matter of priority - Michel: well look at the definition, which involves, as its criteria:
Michel: "biologically significant portion"  - chemicals don't have biologically significant portion Alan: Fair criticism. 

Alan: for definition of derivation, see "Derivation as a relation between instances" in http://genomebiology.com/2005/6/5/R46 
Michel: I think that definition is too vague and/or misleading. Alan: Give me a better one.

Alan says: This is the definition from wikipedia, see above. What does definite chemical composition mean? What doesn't have definite chemical composition?

Michel: The composition of a mixture may not be known precisely.  In some cases, we may make a test, and we know the substance has some properties, but we don't know the exact atomic composition and connectivity.

Alan: So if we are ignorant, then stuff can change from being a chemical substance to not being one? Or from not being one to being one, if we learn something?

michel: If you are ignorant, you would not start by calling it a chemical substance, but maybe an owl:Thing or bfo:Entity ;-)

Agreed.  The Wikipedia definition of "chemical substance" is really ALL ABOUT what you call it, not about what it is.  The scope of chemical models is named things only.

Wikipedia says that concept was "firmly established" in the 18th century.  (For models more physically grounded, you might explore 17th century literature, written before all those abstract chemical models and naming issues began confusing matters.)
Oliver's turn:
Chemical substance = group of bound atoms
Bound = separation costs energy
Chemical bond = relationship between two atoms that energy is needed to separate them
Chemical reaction = process that changes bonds between atoms

Elgar's turn:
Fusion/Fission/ and chemical reactions. 
Distinction between chemical and nuclear reactions.
Chemical reactions always contain bond changes, covalent bond changes.
Dan's turn:
Chemical reactions are changes in the nature of chemical substances
Chemical equations are symbolic models of chemical reactions
Pathways are chained structures of chemical equations

Things not expressible in chemical equations are irrelevant to pathway models, as (like unnamed things) they lie outside the scope of what pathways can represent.

The stuff below directly helps BioPAX only to the extent that it leads to improved formal notations for chained chemical equations.

regards,
Dan

Jonathan Rees

unread,
May 7, 2008, 8:47:09 AM5/7/08
to biopa...@googlegroups.com

On May 7, 2008, at 2:17 AM, Dan Corwin wrote:

Identifiable roughly means "named".  If it isn't, who would care about it?  That limitation helps define the scope of what BioPAX can properly address. 

Maybe no one cares about it (names it) today, but they might tomorrow. Maybe something is not identifiable by us, but it might be identifiable by someone else. Maybe something is not named, or not identifiable, or we don't know whether it's named or identifiable, but we still something want to say something about (perhaps as part of a categorical statement). For example, we cannot identify the catalysts of 40% of metabolic reactions, yet we know they exist and it is useful to talk about them. Maybe Etc.

I think Alan's point is that it is desirable to have class definitions that are objective - that is, membership of something in a class should not depend on human behavior (our state of knowledge, whether something's named, etc.), because then something could change its class through no fault of its own. If today we say X is not in A, and then tomorrow change our mind and say X is in A, it should only be because we have fixed (or made) a mistake, not because the definition of A takes our knowledge state into account and our knowledge state changes. If we don't know whether something's in a class, that's fine; this uncertainty just becomes part of the colossal pool of our ignorance. Saying we don't know whether X is an A is different from saying X isn't an A *because* X isn't named (identified, known, etc.).

I don't think that objectivity is an absolute requirement, but I think our ontologies will be a lot more complicated epistemologically if the truth of statements we make about something depend on what we know we know about it (roughly speaking). Nonobjective classes introduce a time and speaker dependence into our assertions that is inconvenient and confusing. For this reason it is worth investing some effort in nailing down objective definitions so that we can avoid this kind of complication.

Jonathan

Dan Corwin

unread,
May 7, 2008, 12:05:51 PM5/7/08
to biopa...@googlegroups.com
Hi Jonathon -

Debating topics are in-line below.

Best regards,
Dan


Jonathan Rees wrote:

On May 7, 2008, at 2:17 AM, Dan Corwin wrote:

Identifiable roughly means "named".  If it isn't, who would care about it?  That limitation helps define the scope of what BioPAX can properly address. 
Maybe no one cares about it (names it) today, but they might tomorrow. Maybe something is not identifiable by us, but it might be identifiable by someone else. Maybe something is not named, or not identifiable, or we don't know whether it's named or identifiable, but we still something want to say something about (perhaps as part of a categorical statement). For example, we cannot identify the catalysts of 40% of metabolic reactions, yet we know they exist and it is useful to talk about them. Maybe Etc.
You did just talk about them, as "catalysts of .. metabolic reactions".  Most of what humans talk about are identified only by their class names, in exactly the same way.

I agree with Alan's point as you interpret it below, but that is a totally different topic that should not be conflated.  MY point was that until and unless a class is named, it is beyond the scope of BioPAX.

My related second point (made later) was that until and unless such identifications find their way into chemical equations, they will ALSO lie beyond the scope of BioPAX.

Both limits were meant to suggest practical boundaries for future discussions, to help the tiny group called BioPAX-OBO better allocate its extremely limited resources, and as Alan said."to get organized".

I think Alan's point is that it is desirable to have class definitions that are objective - that is, membership of something in a class should not depend on human behavior (our state of knowledge, whether something's named, etc.), because then something could change its class through no fault of its own. If today we say X is not in A, and then tomorrow change our mind and say X is in A, it should only be because we have fixed (or made) a mistake, not because the definition of A takes our knowledge state into account and our knowledge state changes. If we don't know whether something's in a class, that's fine; this uncertainty just becomes part of the colossal pool of our ignorance. Saying we don't know whether X is an A is different from saying X isn't an A *because* X isn't named (identified, known, etc.).
"Objective" class definitions are good, but not magical.  ALL class definitions are artificial objects of information, created by some specific human(s) to meet some specific goal(s).  That is widely understood and accepted by other ontology groups.

An interesting case in point is discussed by this week's NOVA program, which explains that
"the plant equivalent of the Human Genome Project is under way, forging a new family tree of flowers—with many surprises, such as the news that strawberries and marijuana are closely related."
The new family tree (based on DNA) and the old one (based on macroscopic features analyzed extensively  since Darwin's day at least) are both "objective".  Neither one is incorrect, yet they often have very little in common.  Which one you should use depends on your goals.

I don't think that objectivity is an absolute requirement, but I think our ontologies will be a lot more complicated epistemologically if the truth of statements we make about something depend on what we know we know about it (roughly speaking). Nonobjective classes introduce a time and speaker dependence into our assertions that is inconvenient and confusing. For this reason it is worth investing some effort in nailing down objective definitions so that we can avoid this kind of complication.
BioPAX-OBO would do well to acknowledge that all classes are irrevocably linked to human behavior, knowledge, and especially intent.  Its classes (if any ever actually emerge) will also have a time and a speaker and an agenda firmly attached.  Trying to pretend otherwise is counterproductive.

The group should spend more time on clarifying its intentions and agenda, which seem increasingly vague.  At one point they included compatibility with OBO and its guidelines.  Compatibility with unchanging "truth" is a whole lot more slippery, and contradictory to the one above.

Alan Ruttenberg

unread,
May 7, 2008, 12:32:45 PM5/7/08
to biopa...@googlegroups.com
I will repeat my standard Barry quote here.

For more see: http://ontology.buffalo.edu/medo/biodynamic.pdf

> The methodology presupposed in what follows and developed in [10,
> 33] is realist,
> fallibilist, perspectivalist, and adequatist:
>
> (i) Realism holds that reality and its constituents exist
> independently of our
> (linguistic, conceptual, theoretical, cultural) representations
> thereof.
>
> (ii) Fallibilism accepts that our theories and classifications can
> be subject to revision.
>
> (iii) Perspectivalism maintains that there exists a plurality of
> alternative, equally
> legitimate perspectives on reality.
>
> (iv) Adequatism maintains that these alternative views are not
> reducible to any single
> basic view. Thus adequatism is opposed to reductionism, i.e. to the
> thesis that
> there is some one privileged perspective to which all other
> representations of
> reality can be reduced.
>
> These four axes of our methodology are not independent of each
> other. Thus
> perspectivalism is constrained by realism, which means that it does
> not amount to the thesis
> that just any view of reality is legitimate. To establish which
> views are legitimate we must
> weigh them against their ability to survive critical tests above all
> when confronted with
> reality in scientific experiments.
>
> Those perspectives which survive are deemed in the spirit of realism
> to be transparent to
> reality: it is however a fact that apparent scientific certainties
> are sometimes abandoned
> over time, and so each given perspective is accepted always in a way
> which leaves open the
> possibility of future revision. The biomedical sciences are
> themselves in considerable flux,
> a fact that is well illustrated by the recent radical reorganization
> of science departments in
> medical schools [21], reflecting an increasing emphasis in the
> biomedical sciences on
> subcellular processes and modes of organization.
>
> Perspectivalism and realism combined with adequatism generate the
> view that we need
> (and do not merely have as an option) a plurality of alternative
> theories to reflect the
> different perspectives which cover complex domains of reality like
> that of biomedicine.
> Reality is like cheese: it can be cut in many ways. Our methodology
> endorses the need for
> views of entities belonging to different domains (neurology,
> cardiology, urology) all of
> which coexist within a single organism. The progress of science
> often involves appeal to
> the reductionistic methodology for instrumental reasons: scientific
> explanations often take
> the form of a demonstration of how coarse-grained phenomena can be
> reduced to finer-
> grained phenomena, for example at the level of microphysical
> particles. But the purposes of
> building ontologies are distinct from those of empirical science,
> and experience has shown
> that an adequate representation of reality of the sort needed for
> purposes of biomedical
> ontology must take account of a plurality of different views, all of
> which are equally
> veridical. This is because the central purpose of ontology lies
> precisely in its ability to assist
> in the communication between the perspectives associated with
> different scientific
> disciplines.
>
> A perspectivalist approach to biomedical ontology with ambitions to
> remain consistent
> with science and to cope with its reductionistic tendencies will
> need to find ways to do
> justice above all to a plurality of perspectives on different levels
> of granularity. Granularity
> is indeed here understood as reflecting those specific ways of
> carving up domains of reality
> we associate with different scientific theories. One perspective
> might focus on whole
> organisms, another on cellular assemblies. Yet another might seek to
> do justice to the very
> same reality in terms of complexes of atoms or molecules. A fourth
> might talk in terms of
> changes and invariants in an associated continuum of metabolic
> pathways, or of behavior
> (of walking, eating, drinking, sweating) on a whole-organism scale.
> Our approach allows
> that all of these views can be tenable within their respective
> boundaries, and that there need
> be no privileged approach which could justify the reduction of one
> to another. It allows us
> also simultaneously to embrace both commonsensical and scientific
> perspectives on reality;
> that is, it allows us to endorse the view that both common sense and
> science at different
> levels of detail and granularity can give us genuine knowledge of
> the world.

-Alan

Reply all
Reply to author
Forward
0 new messages