Thanks Michael for sending on Jim's figure. I'll look into making the
server public.
James
On 30/11/2011 20:13, Michael Miller wrote:
hi james,
thanks for the example, it looks very promising.
i believe this is jim's link to his illustration of his work (in no way
reflects any effort on my part!):
http://dl.dropbox.com/u/9752413/IRWG/carpt-00469.png
not seeing the formal owl definition but just looking at the png, it looks
like if the owl ontology was a realization of the MAGE-TAB object model
(http://wwwdev.ebi.ac.uk/microarray-srv/magetab/generated-doc/objectmodel.
html or ftp://smd-ftp.stanford.edu/smd/transfers/magetab/magetab.xml --
see attachments) a lot of the attributes and relations can be in a super
term like Material and the realizable terms like Source, Sample, ..., can
have 'is_a' or similar to OBI and EFO.
in terms of making the site public, that would be great so that the W3C
note could reference it and it could joined into LODD
cheers,
michael
-----Original Message-----
From: James Malone [mailto:mal...@ebi.ac.uk]
Sent: Wednesday, November 30, 2011 9:23 AM
To: HCLS
Cc: Marco Brandizi
Subject: MAGE-TAB in RDF
All,
I had promised to send something on regarding the work our student
(Drashtti Vasant) had been doing on producing RDF from our MAGE-TAB
experiments in ArrayExpress. So here are some bits. We started this a
couple of months back before a model was sent around by Jim or Michael
(apologies, not sure who it was exactly - for whatever reason I can not
find the email and diagram that was sent round - could someone resend
it
please!). Interesting to see the things that align/different, I seem to
recall they aren't a million miles off.
Anyway, I've attached the graphs we used as our baseline and the table
of nodes we included as rdf and the triples we formed in this rdf -
note
we do not include everything in the graphs attached. For factor values,
we ran our Zooma term matching across the values to try and mine these
against matches in EFO and for those that we matched we formed triples
(e.g. liver, cancer). For those we couldn't match automatically we
added
some default high level triples (which are less useful) just to say
this
has some factor values and we captured it as a literal text string. I'm
simplifying a bit as it was a bit more complicated but more or less
that
is what we did. As well as the triples in the doc I've attached we also
added some "convenience" triples to each of these nodes directly to the
experiment using part_of relation so you can do simple queries on the
experiment without having to traverse the whole graph.
The store is sitting on an internal server at the moment but we may
open
it up if there is any interest in using it. Otherwise, we have some
internal (interesting) stuff we're doing with it as well expanding and
refining what we have. We focused on things we wanted to ask questions
about explicitly based on a set of competency questions we had formed.
It's not perfect but I'm of the school of release early, often, refine.
Cheers,
James
PS credit for work is to Drashtti Vasant who did most of this
implementation, myself and Tony Burdett supervised with some guidance
on
RDF matters from Marco Brandizi.
--
European Bioinformatics Institute,
Wellcome Trust Genome Campus,
Cambridge, CB10 1SD,
United Kingdom
Tel: + 44 (0) 1223 494 676
Fax: + 44 (0) 1223 494 468
--
European Bioinformatics Institute,
Wellcome Trust Genome Campus,
Hinxton,
Cambridge,
CB10 1SD,
United Kingdom
Tel: + 44 (0) 1223 494 676
Fax: + 44 (0) 1223 492 468
