- From: kei cheung <kei.cheung@yale.edu>
- Date: Mon, 12 Jun 2006 10:59:56 -0400
- To: donald.doherty@brainstage.com
- Cc: public-semweb-lifesci@w3.org
Hi Don, Thanks for drafting the task. I agree with you that establishing semantic correspondences between databases can be a manually intensive process, although id cross-references (e.g., protein ids and gene ids) may sometimes be available from existing sources (e.g., GeneBank and UniProt), which can be used to aid rule construction automatically. After John Barkley's presentation, I’ve been thinking about the way "owl:sameAs" can used to establish correspondences at the class level or instance level (how to draw a distinction between the two levels?). For example, suppose there is a patient related ontology, which has a class called “Patient”. To me, a patient named “John Doe” would be considered as an instance of the class Patient. Since different hospitals may use different types of identifiers to identify their patients (it is possible the same patient has been admitted to different hospitals). For example, the patient “John Doe” may be identified by his soc. sec. number in one hospital, whereas he may be identified by a system-generated id in another hospital. To use owl:sameAs construct to establish correspondences between different patient ids, does it mean that we have to treat each individual patient as a class? Is there a thing called owl:sameIndividualAs? The complexity between pairwise mapping between databases increases with the number of the databases involved (N^2 , where N is the number of databases), whereas mapping the schemas of individual databases to a global schema tends to have a lower complexity (N). However, it may be more difficult to reach a consensus for constructing or choosing a global schema. For your Parkinson's Disease (PD) use case (neurondb and cocodat mapping), we might also want to consider whether it's possible to incorporate microarray data related to parkinson disease. For example, the following microarray experiment: http://arrayconsortium.tgen.org/np2/viewProject.do?action=viewProject&projectId=61 compares gene expression patterns dopaminergic neurons between two brain regions (SNc) and (VTA). Such a differential expression pattern might be due to the difference in neuronal properties (e.g., neuronal circuitry) between these brain regions. NeuronDB and CocoDat contains this type of information. This example also provides a nice high level description of the experiment (e.g., hypothesis, experimental procedure and design, etc). It also demonstrates the need to incorporate other ontologies such as UMLS disease ontology and NeuroNames (different brain regions including SNc and VTA are involved). In addition, it exports data in MAGE-ML format. We can use such an example as a use case to explore how to extract some high-level experimental description from MAGE-ML and map it to FUGE or EXPO (ie., we just map a small subset of MAGE-ML to FUGE or EXPO). This seems to fit into the current discussion on FUGE and EXPO. We might also be able to find more PD-related microarray data in other public microarray databases (e.g., GEO and ArrayExpress) to broaden our microarray use case. Finally I would like to bring your attention to the DopaNet ontology (http://www.dopanet.org/), which might be of interest to our PD use case. I am interested in participating in your task as I see a potential cross-fertilization between the senselab task and your task. Cheers, -Kei Donald Doherty, Ph.D. wrote: >I've added a page for the Bridging Ontology at: > >http://esw.w3.org/topic/HCLS/OntologyTaskForce/Create_Bridging_Ontology_between_NeuronDB_and_CoCoDat_databases_and_UMLS_Common_Vocabulary#preview > >You may also get there from the BIONT page: > >http://esw.w3.org/topic/HCLS/OntologyTaskForce/ > >Please provide feedback! > >Thank you, >Don > >Donald Doherty, Ph.D. >Brainstage Research, Inc. >412-478-4552 > > > >
Received on Monday, 12 June 2006 15:00:16 UTC