Wednesday, February 15, 2006
Generative Methodology Glass Bead Games
n-articulated ontological framework
On using RDF to model biological expression
Link back to part of the “solution” to translations between
RDF / OWL and Models of Information - [167]
Concurrent Anticipatory web discussion à [30]
Comments from Tom to some questions put forth by Paul
One way in which they might
be useful is if there are a lot of database investments and people are
unwilling or not informed about the use of ontology as a step up. Like the horse, when cars were being
introduced. Horses are useful, although having to put up with them I have never
quite felt that they were THAT useful.
My own perspective, as software developer, is that we are attempting to create dynamic models of biological systems. In that context, bioPAX in particular and semantic web approaches in general promise to capture the information needed to generate the classes, methods, data, and everything else that goes into getting the *dynamics* of those models right. That's the good news. As I see it, the bad news is that ontologies -- as I understand the current state of the art -- are relatively bad at capturing time-dependent behavior. Whoops.
Relational databases are good at managing large volumes of data whose structure changes infrequently. They are relatively bad at performing operations on that data. For example, they're very good at recording the transactions against your bank account. They're relatively bad at calculating the interest rate, service charges, and average daily balance.
I would expect some sort of database schema to be part of a model generated from a bioPAX description of a signaling pathway, but I think that's one of the last things to be determined, rather than one of the first. I suppose you *might* be able to generate spreadsheets that, when executed by Excel, do what you wanted -- and those spreadsheets might have tables and columns in addition to functions. I think even that, however, is a LONG way away from relational database schema.
Can we, with BioPAX ontology, represent the nonlocal aspects of
signal pathways, remembering that "knowing what "signal
pathways"" means will simply go over the head of most computer
scientists. Even if the biologist is able to convince a database
engineer that there is something to understand about how "the interior of
a cell performs "communications"" as part of the complex process
of assembling proteins etc... one is
likely to find that the engineer
training of the database engineer will simple
ignore aspects that cannot be encoded into a relational schema.
If we ask a "database engineer" to do this modeling for us, we have larger problems than we're talking about here.
So, is this useful to create persistence and efficiency by encoding the bits and bytes of a ontological models with a massive number of instances. Perhaps. But clarity over the issues is perhaps more important. Yes?
It seems to me that we want to ask our computer scientists to help create simulations of biological behavior, where the active elements of the simulation are generated in deterministic ways from the ontologies we record with tools like bioPAX. As I see these ontologies being used, the ontology will emit javascript, java, python, ruby, perl, or whatever that in turn models desired behavior *when executed*. In that context, I think relational databases play a role comparable to gui toolkits, operating system calls, and file systems. Necessary, yes. Central? I doubt it.
Thx,Tom