[Biodevelopers] RDBMS vers new data model

Dan Bolser dmb at mrc-dunn.cam.ac.uk
Thu Apr 22 16:14:38 EDT 2004


On Thu, 22 Apr 2004, Michel Dumontier wrote:

> Dan,
> > IMHO, I think no one datamodel can / should / would or could be set in
> > place, rather every biologist makes their own data model either from
> > scratch, by combining other datamodels in to larger models and/or with
> > available ontologies.
> Your statement hardly helps the state of things... which is a disaster from
> an integrative and service perspective.  Aside from issues of data quality,

I agree ;)

> it is current very difficult to make full and accurate database cross
> references, nevermind doing proper updates where everybody has their own
> accession space.  If each person has their own data model, then it makes it

Can I ask what is accession space?

> extremely difficult to share information without having to write additional
> code to map your object model to mine.  However, given a standard way of

True, this is the 'under the hood' aspect which means that all the models
I describe must be linked by some machine readable standard - which is
extremely difficult to get accepted by the 1000's of people making
databases and publishing data on the web. That isn't to say that such
standards don't exist and are not efficient and 'seamless' (to an extent).

Specifically I am thinking about semantic layers above data model layers
(hence ontologies). But we need a flexible way to allow ontologies to
develop - which isn't too controversial a thing to say.

I think that this development should be a community effort, not

> representing a gene and all associated annotation,  then I could use your
> annotation immediately without having to do the tedious mapping  If you only

True, and learning the details of a particular model can be a drag
too. However, this is the task of a researcher researching data in an
unfamiliar field - we do this all the time in the process of doing 'real'

> need a particular aspect of a grander data model for your purposes, then
> just use that part.  If you need to extend a data model, for whatever
> reason, then there should be a process (and XML/ASN.1 based data models are
> fully extensible).  However, experience says that once you start building a

That is what I mean (I think). Having a well defined model for your source
data allows you to take what you need and build what you want. Then
someone else starts to take something of what you provide and so on.
Sooner or later a core standard will emerge. I hope that anyone who has
read this far can somehow see this vague point, because this is what
science is, the development of ideas and the formation of a consensus.

> 'new'/'better suited' model, you'll end up re-inventing a well-thought-out
> model.  Having standardized, even peer-reviewed, data models promotes a
> better computing environment and makes it easier to make advances in
> science.  For most of us, Science, not technology or programming, is the
> ultimate pursuit.

I absolutely agree. It boggles my brain to think what we (me and my peers)
will be able to do scientifically in the next 50 years with the integrated
biological universe at our fingertips!

> -=Michel=-


> _______________________________________________
> Biodevelopers mailing list
> Biodevelopers at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/biodevelopers

More information about the Biodevelopers mailing list