W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > May 2011

RE: Relations in RDF

From: Michel_Dumontier <Michel_Dumontier@carleton.ca>
Date: Wed, 4 May 2011 13:51:47 -0400
To: James Malone <malone@ebi.ac.uk>, HCLS <public-semweb-lifesci@w3.org>
CC: "M. Scott Marshall" <mscottmarshall@gmail.com>
Message-ID: <E1784B0107E5634C8997868083EDE780612001EF28@CCSMBX10.CUNET.CARLETON.CA>
Hi James,

In Bio2RDF, and linked data in general, it's common to have a predicate for nearly every relation in the dataset - this at least solves the syntactic interoperability problem, by at least having the data in RDF and available to query using SPARQL. However, it makes dataset integration an outstanding challenge. With respect to semantic interoperability, the idea is that we'll decompose the meaning of these by mapping them to ontologies that do provide meaningful, but still largely domain independent differentiation of relations. Given that some relations actually map to expressions (in part as we showed in http://www.biomedcentral.com/1471-2105/11/441), there's quite a bit of flexibility, although more is required. 

 While I'm not a big fan of predicate proliferation, having a diverse array is important when trying to provide axiomatic descriptions for *every* class. In doing so, we've been thinking about informational relations, and we have a few now that are more general and different from "is_about". Currently, the most general relation is 'refers to'


from which new predicate specializations emerge around reference, description and representation. For us at least, the "is about" relation provides a description (some information) about it.  

We're working on mapping Bio2RDF's types and predicates to the SIO ontology (
http://code.google.com/p/semanticscience/wiki/SIO), that way we have a better understanding of the nature of data in play, and the kinds of relations that are required to support the semantic integration effort. Discussions about these and other relations are most welcome on our mailing list:



> -----Original Message-----
> From: public-semweb-lifesci-request@w3.org [mailto:public-semweb-lifesci-
> request@w3.org] On Behalf Of James Malone
> Sent: Wednesday, May 04, 2011 11:42 AM
> To: HCLS
> Cc: M. Scott Marshall
> Subject: Relations in RDF
> Hi Scott, All,
> I was wondering what the general take is on predicates in RDF
> representations used by the HCLS group. I've been looking at our RDF
> model for Gene Expression Atlas at EBI and presently I'm using the same
> "is_about" relation for a lot of the predicates as this is the lowest
> level of constraint from the OBO Foundry folks for some of these
> information relations. Alan Ruttenberg tells me that empirical evidence
> suggests that using a larger number of relationships correlates to
> poorer ontologies. However, I've also been told from various RDF
> advocates that having more granular level predicates is useful for
> querying. Are there any thoughts from the group on this? I have no
> preconceptions here (I have no reason to disbelieve Alan or the RDF
> folks) so open to thoughts and suggestions.
> Cheers,
> James
> --
> European Bioinformatics Institute,
> Wellcome Trust Genome Campus,
> Cambridge, CB10 1SD,
> United Kingdom
> Tel: + 44 (0) 1223 494 676
> Fax: + 44 (0) 1223 494 468
> -----
> No virus found in this message.
> Checked by AVG - www.avg.com
> Version: 10.0.1325 / Virus Database: 1500/3614 - Release Date: 05/04/11
> -----
> No virus found in this message.
> Checked by AVG - www.avg.com
> Version: 10.0.1325 / Virus Database: 1500/3614 - Release Date: 05/04/11
Received on Wednesday, 4 May 2011 17:52:15 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:52:46 UTC