[Bioperl-l] another use of Perl for biology

[Guillaume Rousse]

Le Mardi 25 Février 2003 19:36, vous avez écrit :
> Guilleme,
>
>
> I had a quick look at your site.
>
> Coming in totally unprepared is was quite difficult to get a picture of
> what you are working on in practical terms. The whole system is there
> but it has not been broken into manageable chunks. I had to dig into one
> of the posters to find out what what OSIS stands for (Object-oriented
> Systematics Information System).
My fault, the url i gave was uncomplete
http://lis.snv.jussieu.fr/~rousse/recherche/osis/en

>  It looks really ambitious and complex.
> You are using graphical interfaces to model taxonomic problems by
> reusing basic classes, right?
No, basic modeling tool is UML + OCL. 

> I can not get the feel of how stable are
> your basic classes? In other words, is the development now focused on
> connecting and using these basic classes rather than trying out
> different ways of modelling basic units in systematics?
The first way. Current focus is to import legacy databases with complety 
different models, which was quite easy for shared classes (scientific names 
), but is more difficult for classes (pathologic vectors) to a given model. 
Anyway, specific classes are usually created as specialisation from basic 
ones.

> How about the
> ontology?
I feel uneasy to set the limit between a data model and an ontology. Too much 
hype, and too many definitions. Some people even consider a simple list of 
words ("controled vocabulary") to be an ontology.

> Do you have a graphical representation of the current class
> model(s)?
Yes, see http://lis.snv.jussieu.fr/~rousse/recherche/osis/en/model.html

> Is there now a consensus in the field  how these things should
> be modelled?
No, parlty because most people in the field usually publish only 
implementation models, which are really difficult to compare. And even before 
any modeling concern, some key concepts (taxon, for instance) still lack also 
formal definition.

 Are you confident you model can tackle most real life
> problems?
Depends of what you consider "real life" :-). There are currently 4 real 
systematics database using this framework, so i'd say it can tackle at least 
systematician needs.

> We do not have practically nothing related to taxonomy/biodiversity in
> bioperl, but I'd love to see something. If your basic atomic classes are
> in stable stage, it'd be grand if could consider putting them into
> bioperl with some sample scripts. That would be a quick way for other
> people to get feel of your approach and give you feedback.
Well, i've asked some times ago about a namespace to use, to avoid threading 
on anyone's else toes, and i finally use Bio::Systematics::OSIS (and 
Bio::Systematics::whatever for any other Perl developement in my lab), so i 
should at least be BioPerl-non-conflicting.
Right now i miss time to write sample scripts beyond the test case and current 
applications, moreover it's more practical for me to use our own CVS as long 
as only local people are involved in development. But joing BioPerl later is 
really an interesting proposal :-)

Thanks for your interest.
-- 
The speed with which components become obsolete is directly proportional to 
the price of the component. 
	-- Murphy's Computer Laws n°9