Monday, January 16, 2006
New
discussion about signal pathways
and complex ontology
[350] ß parallel discussion in “national debate” bead thread
This is part of a discussion
that will be moved to a Wiki page soon.
Ontological modeling with
category theory and convolutions
[347] ß Judith Rosen’s recent communication
I note about a side discussion:
Andrea Splendiani suggested I read a paper on the usefulness of UML (to describe and specify biological systems and processes)
<quote>
One
of the main issues in Systems Biology is to deal with semantic data
integration. Previously, we examined the requirements for a reference
conceptual model to guide semantic integration based on the systemic
principles. In the present paper, we examine the usefulness of the Unified
Modelling Language (UML) to describe and specify biological systems and
processes. This makes unambiguous representations of biological systems, which
would be suitable for translation into mathematical and computational
formalisms, enabling analysis, simulation and prediction of these systems
behaviours.
<end quote>
I did not want to pay to read the article, but the abstract has the essential information.
The issue is one shared by OWL with UML; in that IF the objective is to provide a standardization around data integration and coding (inserts into information bases) then SOME usefulness is found. Of course. In face of the now seemly intractability of full scale AND completely satisfactory data integration and coding using UML, this usefulness is interesting.
However, one has to consider down stream issues.
1) does the structural nature of any standarization based on UML or OWL
cause specific irreversible limitations in the usefulness of cell and gene
expression data, specific to issues such as non-locality of causes (field
constraints) and/or non-measured differences between functional categories (so
that data is incorrectly placed into a structural form where OWL logics are
used assuming the functional placement is "truth".
2) does the structural nature provide the best known encoding and
retrieval mechanisms to support eventual "translation into mathematical
and computational formalisms, enabling analysis, simulation and prediction of
these systems behaviours"
3) are mathematics and computational formalism sufficient to key inquiries regarding cell and gene expressiveness.
I have the opinion the #1 has the answer yes, but this answer can be accommodated by the type of reminders that biologists make to each other all the time regarding formalism. On #2, this depends on how BioPAX evolves, and I am very interested in tracking this evolution, working with some expression data, and making myself available to the discussion. #3 is the subject of the Rosen interpretation of natural complexity, as being something more than formalism can capture. So we work within the existing mathematical and computational formalism, and see if there is a way to extend mathematics (for example as did the Soviet school of applied semiotics), and - in a different way rough sets.
These are opinions that I simply state, and would like not to discussion them very much given other constraints on our time. I note that several others in the working group share these opinions, and am happy about that, but it is not something that I am interesting in debating.
Now a note about Phillip's comments and discussion points.
I am an outsider to the DERI group, but see how they are shaping interesting OASIS standards - in a way that is currently already available... so there is work already done on modeling language and modeling ontology - which is consistent with how I see the development of the BioPAX environment.
But I am still conjecturing about DERI's future role in data integration efforts.
I note that Web Services Execution Environments (called WSMX by W3C and SEE by OASIS) come in the W3C and Oasis flavors.
I have learned to not trust the W3C standards because the W3C (and Semantic Web) community are overly reductionist in practice philosophy and underlying philosophy. If the natural world were perfectly modeled by scientific reductionism, then there would be not valid criticism. However, there is controversy about this.
OASIS addresses the issues quite differently. Even though the Topic Map paradigm was terminated (in the minds of the two authors of version 1.0 Newcomb and Biezunski (spelled wrong) and others), there was once the notion that TM and RDF were completely different in basic use philosophy. OASIS supported TM and W3C supported RDF.
The approach being made by BioPAX is a practical approach that is not consistent with the mainstream use philosophy of OWL according to the standard's body, the W3C. For this I am grateful. BioPAX development of a data integration process might be one of the very first examples of mature natural scientists disregarding the Emperors of AI/SW and attending to the issues of natural science rather than the research projects of computer science.
I might have been confused but WSML (Web service modeling language) and WSMO (web service modeling ontology) are both OASIS standards, not in any way the "same as" the W3C’s MSMX.
Is this how you see it?
Phil, your last statement
"My own feeling is that the next step will come from improved
best practices, making better use of the formalisms we have, rather
than
starting with new ones. The recent debate on the biopax is perhaps
a sign of this."
is NOT consistent with my view of the world and of the best strategy. I have argued long and in many forums that the current formalism are fundamentally flawed in specific ways, I will not repeat the analysis of the flaws once again.
I do not see the debate within this working group as signalling a strategy to “merely” use existing formalism better. What I see is the biologist's understanding that formalism should be more open to the natural categorization and re categorization common in everyday human experience of one's own awareness. This means using that part of set theory that is proper and introducing more elements of category theory.
I do not think that the mature understanding about the limitations of computation and mathematical foundations requires that we throw away set theory. I do think that this mature understanding is the next step, and that arguments based on conjectures about the ultimate success of OWL or UML in providing general purpose ontological modeling of biological expression is not in good faith.
It is the politically correct argument in some but not all circles, but it is not the correct one. (This is my opinion).
Dr Paul Prueitt (mathematics and quantum cognitive neuroscience)
The Taos Research Institute
Taos New Mexico