RE: Minutes from Monday's Linked Life Data (BioRDF / LODD) teleconference

hi all,

sorry for jumping in late, i've been traveling and then catching up at
work.

characteristics has its origins from the original MAGE model and, in this
context, is meant, because it is associated with a BioSource, to be
absolutely anything that can be associated with any living or dead
organism.  so it is not surprising that we're having this issue!

in the MAGE model the OntologyEntry was used to represent both relations
and objects from desired ontologies and this was preserved in MAGE-ML.
here's a rough sketch of the elements, attributes and associations
(->)from the DTD:

BioSource -> OntologyEntry -> DatabaseEntry        -> Database
                           category:                accession:
version:
                           value:                       accessionVersion:
URI:
                                                               URI:

roughly, the Database element referred to the Ontology, the DatabaseEntry
would be the URI for the OntologyEntry in the Ontology, and the
OntologyEntry would be the label where if it were a class, category would
equal value and if it were an instance, category would be the class of the
instance.  then the BioSource, probably naively, would be an instance of
the class the OntologyEntry referred to.  if OntologyEntry were a
relation/restriction in the Ontology, then the BioSource would be
considered an instance of a class in the Domain of the restriction [1][2]


now for MAGE-TAB this is reduced to be even more simple (the guiding
principle of MAGE-TAB) where the OntologyEntry becomes the Characteristics
header and the [...] after 'Characteristics' holds the category while the
column value is the value, often the label of the class .  the following
TermSourceRef indicates the Ontology and a TermAccession column after
that, the URI of the class in the Ontology with the same implied instance
relationship from above.

is this underspecified?  probably yes.  was it useful?  for someone
reading the document it did give the gist but, yes, for automation i can
see the problem.

cheers,
michael

[1] http://mged.sourceforge.net/ontologies/MOMAGEpolicy.txt
[2]
http://mged.sourceforge.net/ontologies/GilbertoTestAnnotationValidated.xml

Michael Miller
Software Engineer
Institute for Systems Biology

> -----Original Message-----
> From: Michel Dumontier [mailto:michel.dumontier@gmail.com]
> Sent: Thursday, April 26, 2012 2:31 PM
> To: M. Scott Marshall
> Cc: Mark; Jim McCusker; HCLS
> Subject: Re: Minutes from Monday's Linked Life Data (BioRDF / LODD)
> teleconference
>
> Hi from a rest stop in Vermont with tasty coffee and wifi :)
>
>   So SIO features a few information relations of interest.  The most
basic is
> 'refers to',
>
> http://semanticscience.org/resource/SIO_000628
>
>  which is to say that an entity is mentioned.  If there is an actual
description,
> then we can use 'describes' otherwise, it might only be a reference
without
> further elaboration - 'references'.
>
> Does that help?
>
> m
>
>
> On 2012-04-26, at 4:03 PM, "M. Scott Marshall"
<mscottmarshall@gmail.com>
> wrote:
>
> > Hmm. Michel suggested that "set" of relations in November:
> > Michel wrote: " there are some basic relations that we can look at,
> > for instance, i have defined a wider set of relations as part of my
> > SIO ontology: "
> >
> > What we were looking for was a way to describe that a given dataset
> > might have to do with a number of specific subjects, such as the
> > taxonomic identifier of the model organism from which the data was
> > collected, but possibly also, the disease or phenotype that was being
> > modelled. So, we needed a predicate that was vague enough to express
> > such things, yet still be useful in other domains.
> >
> > -Scott
> >
> > On Thu, Apr 26, 2012 at 9:52 PM, Mark <markw@illuminae.com> wrote:
> >> I think Michel deprecated "is related to" DURING the Hackathon! (out
of
> >> disgust, if I recall correctly...) "not enough semantics"!
> >>
> >> So... He's forcing us to be more clever :-)
> >>
> >>
> >> M
> >>
> >>
> >> "M. Scott Marshall" <mscottmarshall@gmail.com> wrote:
> >>>
> >>> On Monday, Jim's description reminded me about a recent discussion
in
> >>> the biohackathon dbcatalog group about the best predicate for
> >>> indicating that a dataset has 'something to do with', for example,
mus
> >>> musculus (not unlike Mark's example). I remembered some is_about
> >>> predicate from SIO. The predicate that I couldn't quite remember at
> >>> that time but has remained the strongest contender for the job was
> >>> http://semanticscience.org/resource/is-related-to .
> >>>
> >>> -Scott
> >>>
> >>> P.S. My sympathy goes out to you Mark as you wait for your WIFI
> >>> hookup. I guess that you are forced to go to the pub and have a
drink
> >>> in order to get a bit of normal net speed (from their WIFI). :)
> >>>
> >>> On Thu, Apr 26, 2012 at 9:04 PM, Jim McCusker
> <james.mccusker@yale.edu>
> >>> wrote:
> >>>> Actually, we've switched over to generating subproperties of
> >>>> sio:has_predicate, so I'm glad you suggested the same thing.
> >>>>
> >>>> Jim
> >>>>
> >>>> On Thu, Apr 26, 2012 at 2:07 PM, Mark <markw@illuminae.com> wrote:
> >>>>> Hmmmm... what's the range of that predicate? It's a bit like the
SIO
> >>>>> "has
> >>>>> attribute", where you need to be sure to explicitly type what's at
the
> >>>>> other
> >>>>> end to do "useful" reasoning. We're just concluding a project
where
> we
> >>>>> use
> >>>>> species-of-origin to restrict the choices the SHARE query resolver
can
> >>>>> make
> >>>>> when it queries the SADI registry. (the query requires the
discovery of
> >>>>> a
> >>>>> BLAST service, and it will chose the right one based on the
species
> >>>>> that the
> >>>>> sequence is derived from...)
> >>>>>
> >>>>> Similar problem? if so, let me know and I'll post the model we
came-
> up
> >>>>> with
> >>>>> a few weeks ago... Maybe we can converge on a common solution?
> >>>>>
> >>>>> Cheers all!
> >>>>>
> >>>>> M
> >>>>>
> >>>>>
> >>>>>
> >>>>> "M. Scott Marshall" <mscottmarshall@gmail.com> wrote:
> >>>>>>
> >>>>>> We had an interesting discussion with Jim McCusker about Linked
> Data /
> >>>>>> RDF representations of MAGE-TAB and some issues, such as choices
> of
> >>>>>> predicates to make describe a human sample.
> >>>>>>
> >>>>>> how to encode "is a sample originating from human"
> >>>>>> closest now is "has characteristic"
> >>>>>>
> >>>>>> Jim showed us some linked data versions of MAGE-TAB containing
> >>>>>> descriptions and data of studies. Very nice!
> >>>>>>
> >>>>>> The minutes are here:
> >>>>>> http://www.w3.org/2012/04/23-HCLS-minutes.html
> >>>>>>
> >>>>>> Cheers,
> >>>>>> Scott
> >>>>>>
> >>>>>
> >>>>
> >>>>
> >>>>
> >>>> --
> >>>> Jim McCusker
> >>>> Programmer Analyst
> >>>> Krauthammer Lab, Pathology Informatics
> >>>> Yale School of Medicine
> >>>> james.mccusker@yale.edu | (203) 785-6330
> >>>> http://krauthammerlab.med.yale.edu
> >>>>
> >>>> PhD Student
> >>>> Tetherless World Constellation
> >>>> Rensselaer Polytechnic Institute
> >>>> mccusj@cs.rpi.edu
> >>>> http://tw.rpi.edu

Received on Friday, 4 May 2012 16:20:45 UTC