[MOBY-l] Queries in bioMOBY

[Phillip Lord]

>>>>> "Chris" == Chris Mungall <cjm at fruitfly.org> writes:

  Chris> On Fri, 8 Nov 2002, Lukas Mueller wrote:

  >> Hi, I recently heard at talk at a bay area bioinformatics meeting
  >> about a company that has implemented exactly what Chris is
  >> talking about. I was a bit skeptical of their implentation. What
  >> they do is develop wrappers around web-pages and query them as
  >> web-services, presenting to the user what looks like a big sql
  >> database. The company is at
  >> www.geneticXchange.com. Unfortunately, they don't have that much
  >> information on the site than the (I think) CEO presented in the
  >> talk. He also gave a demo and the performance seemed quite
  >> good. It must be a nightmare to maintain these wrappers.... as
  >> web-pages changes, these wrappers have to be adapted. They have
  >> about 60 wrappers developed. They really seem to be a poor man's
  >> ad-hoc moby, so I think that such a system would be better built
  >> on top of something like moby. Of course the system on top of the
  >> wrappers must be pretty neat.

  Chris> I agree the wrappers must be tricky to maintain - the idea
  Chris> here would be, say, you could wrap Entrez in a bunch of crazy
  Chris> scripts that ultimately allow you to query genbank as if it
  Chris> were a relational db?


A bunch of wrappers is how TAMBIS worked, with a ontology driven query
language on top of that. I'm sure Robert (who did a lot of the work on
TAMBIS) could tell you more details, but maintaining wrappers is a
pain. 

Of course, if you build your own wrappers on top of something else
(like SRS for instance), then you can isolate yourself from at least
some of the maintenance. 

  Chris> Given that the native form of the data is often a relational
  Chris> db, it seems to me that it would be easier having a few SQL
  Chris> views or materialised views to turn the native schema to an
  Chris> export schema...

You make the assumption that service provides would want to give SQL
access to their databases, which I think is questionable.

  Chris> Of course, if the site in question refuses to provide that
  Chris> then some kind of wrapper is necessary. A lot of
  Chris> bioinformatics would be so much easier if NCBI provided
  Chris> external access to a relational db containing genbank, dbsnp,
  Chris> ....

This is true, but it's unlikely to happen soon. Hence the eternal
problems, data warehouse, or data source wrappers?

Phil