hi jim and lena,
great progress! this will be a nice tool.
a couple of comments.
1) i think ProtocolApplication is based seen as an individual instance of
the Protocol class. quite often there are arguments whether ontologies
should have individuals or be simply classes. to me, that doesn't apply
here where real world objects are being connected to ontologies. the
BioSource is realized as the 'Source Name' column in MAGE-TAB and those
entries represent real people in studies, mice or rats in non-clinical
studies, etc., and the characteristics values like age represent real
individual instances of age. in the same way, the values in the Protocol
REF column of MAGE-TAB are real wet-lab or analysis individual instances of
protocols, called protocol applications in MAGE-OM.
failure to make this distinction, to me, has obscured how much value
ontologies can have in the real world. too often i see ontologies seen in
and of themselves, which has its own value certainly, but not for the use
cases i have dealing with real biological data.
2) the usefulness, for this use case, of the information between the 'Source
Name' and its characteristics and the 'Derived Array Data Matrix File' or
'Derived Array Data File' has limited usefulness, error correction and
normalization can make some difference but if the provider of the MAGE-TAB
is trusted, all that is pretty routine these days. the above combined with
experimental factors and experiment design info is probably 95% to 99.9% the
worthwhile information from the MAGE-TAB. if one notices a difference in
the final gene set between two experiments that look the same, only then it
might be worthwhile going into more detail.
and has been noted the MAGE-TAB information needs to be supplemented with
the information on the final gene set, its expression values, and the
higher-level level analysis that was used, that is buried in the paper
usually.
3) i'm not sure if there was a desire to capture the raw data in the RDF.
that will be, for affymetrix, a million to six million probes in the CEL
file, even the processed data in the CHP file would have 20,000 to 60,000
probe sets. i'm not sure if that is the best way to represent that.
cheers,
michael
Michael Miller
mdmille...@comcast.net
----- Original Message -----
From: "Jim McCusker" <james.mccus...@yale.edu>
To: "Helena Deus" <helenad...@gmail.com>
Cc: "Kei Cheung" <kei.che...@yale.edu>; "mdmiller" <mdmille...@comcast.net>;
"HCLS" <public-semweb-lifesci@w3.org>
Sent: Monday, November 30, 2009 8:19 AM
Subject: Re: BioRDF Telcon
I'm following a similar strategy, but have been folowing the MGED
ontology where possible. I've finished aligning the IDF portion, and
have started on SDRF. MGED ontology is missing a property and class
for what is often termed as ProtocolApplication, which usually serves
as an edge between derived from and derived nodes, while linking to
the protocol used for the derivation. I am planning on creating this
link in a MAGE extensions ontology, but would like to vet the
structure here:
ProtocolApplication is a class.
New properties:
has_derivation_source
has_derivative
And then ProtocolApplication would have the restrictions:
has_protocol some Protocol
I don't put, domains, etc. on the derived properties to allow use in
directly describing derivations if people so choose. There is no
superclass for all nodes that can be derived or derived from, so I'm
not bothering with restrictions for those, although I could add a
union restriction to it.
If this structure us acceptable to people, I can publish the ontology
for general use pretty quickly, and let us work from the same data
structure. I would appreciate any feedback.
Jim
On Monday, November 30, 2009, Helena Deus <helenad...@gmail.com> wrote:
@Kei,
When you said data structure, did you mean the RDF structure
For now, all I have is the java object returned by parser. I've been using
Limpopo, which creates an object that I can then parse to RDF uing Jena.
The challenge, though, has been coming up with the predicates to formalize
the relationships between the various elements. I'm using the XML
structures fir IDF/SDRF etc. at http://magetab-om.sourceforge.net to
automatically generate the structure that will contain the data. My plan
is to then create the RDF triples that use the attributes described in
those documents and populate them with the data from the MAGE-TAB java
object created by Limpopo.
Right now all I have is a very raw RDF/XML document describing the
relationships in the IDF structure:
http://magetab2rdf.googlecode.com/svn/trunk/magetabpredicates.rdf
The triples for that had to be encoded manually using Jena by reading the
model.
@Satya and Jun
I would very much like to be involved in that effort, do you already have
a URL that I can look at?
ThanksLena
On Tue, Nov 24, 2009 at 2:19 PM, Kei Cheung <kei.che...@yale.edu> wrote:
Hi Lena et al,
When you said data structure, did you mean the RDF structure. If so, is a
pointer to the structure that we can look at?
As discussed during yesterday's call, Jun and Satya will help create a
wiki page for listing some of the requirements for provenance/workflow in
the context of gene lists, perhaps we should also use it to help
coordinate some of the future activities (people also brought up Taverna
during the call yesterday). Please coordinate with Satya and Jun.
Cheers,
-Kei
Helena Deus wrote:
Hi all,
I apologize for missing the call yesterday! It seems you had a pretty
interesting discussion! :-)
If I understand Michael's statement, parsing the MAGE-TAB/MAGE-ML into RDF
would result in obtaining only the raw and processed data files but not
the mechanism used to process it nor the resulting gene list. That's also
what I concluded after looking at the data structure created by Tony
Burdett's Limpopo parser. However, having the raw data as linked data is
already a great start! Kei, should I be looking into Taverna in order to
reprocessed the raw files with a traceable analysis workflow?
Thanks!
Lena
On Tue, Nov 24, 2009 at 9:59 AM, mdmiller <mdmille...@comcast.net
<mailto:mdmille...@comcast.net>> wrote:
hi all,
(from the minutes)
"Yolanda/Kei/Scott: semantic annotation/description of workflow
would enable the retrieval of data relevant to that workflow (i.e.
data that could be used to populate that workflow for a different
experimental scenario)"
what is typically in a MAGE-TAB/MAGE-ML document are the protocols
for how the source was processed into the extract then how the
hybridization, feature extraction, error and normalization were
performed. these are interesting and different protocols can
cause differences at this level but it is pretty much a known art
and usually not of too much interest or variability.
what is usually missing from those documents, along with the final
gene list, is how that gene list was obtained, what higher level
analysis was used, that is generally only in the paper unfortunately.
cheers,
michael
.
----- Original Message ----- From: "Kei Cheung"
<kei.che...@yale.edu <mailto:kei.che...@yale.edu>>
To: "HCLS" <public-semweb-lifesci@w3.org
<mailto:public-semweb-lifesci@w3.org>>
Sent: Monday, November 23, 2009 1:27 PM
Subject: Re: BioRDF Telcon
Today's BioRDF minutes are available at the following:
http://esw.w3.org/topic/HCLSIG_BioRDF_Subgroup/Meetings/2009/11-23_Conference_Call
Thanks to Rob for scribing.
Cheers,
-Kei
Kei Cheung wrote:
This is a reminder that the next BioRDF telcon call will
be held at 11 am EDT (5 pm CET) on Monday, November 23
(see details below).
Cheers,
-Kei
== Conference Details ==
* Date of Call: Monday November 23, 2009
* Time of Call: 11:00 am Eastern Time
* Dial-In #: +1.617.761.6200 (Cambridge, MA)
* Dial-In #: +33.4.89.06.34.99 (Nice, France)
* Dial-In #: +44.117.370.6152 (Bristol, UK)
* Participant Access Code: 4257 ("HCLS")
* IRC Channel: irc.w3.org <http://irc.w3.org> port 6665
channel #
--
Jim
--
Jim McCusker
Programmer Analyst
Krauthammer Lab, Pathology Informatics
Yale School of Medicine
james.mccus...@yale.edu | (203) 785-6330
http://krauthammerlab.med.yale.edu
PhD Student
Tetherless World Constellation
Rensselaer Polytechnic Institute
mcc...@cs.rpi.edu
http://tw.rpi.edu