Ontology Driven Dynamic Linking of Biology Resources

1. Ontology Driven Dynamic Linking of Biology Resources S.K. Bechhofer, R.D. Stevens, and P.W. Lord 2005 PSB(Pacific Symposium on Biocomputing) Junwon Jung SNU OOPSLA Lab. Nov 10. 2005

2. Contents • Introduction • Static link • Dynamic linking service • GO(Gene Ontology) • COHSE(The Conceptual Open Hypermedia Service) • GOHSE(COHSE+GO) • Discussion • Conclusion

3. Introduction(1/5) • Bioinformatics relies on the knowledge held within its documents • web pages, database entries, books or articles • Query by navigation is still fundamental to practical bioinformatics • It is the links between biology documents that provide the utility to both humans and machines • Common usage of the Web involves embedding links within documents

4. Introduction(2/5) • Static link • A number of limitations(1/2) • Hard Coding • Links are hand-crafted and hard coded in the HTML encoding of a page • Format Restrictions • Documents need to be written in a particular format • Ownership • Ownership of the page is required in order to place an anchor in a page

5. Introduction(3/5) • A number of limitations(2/2) • Legacy resources • It can be difficult to deal with legacy material • Maintenance • There is a weight of maintenance in creating and updating links in pages • Link targets • Current Web links are restricted to point to point linking

6. Introduction(4/5) • Dynamic linking service • Dynamic linking services, supported by ontologies, offer a mechanism to overcome restrictions of static link • links are derived through the use of an ontology along with mapping from concepts to possible link target • Dynamic linking service provides • glossary functionality • building dynamic hypertext structures • linking to multiple resources

7. Introduction(5/5) • Example • A biologist is reading aWeb page about cellular structure • e.g. http://www.rcn.com/BiologyPages/CellularRespiration.html • Static link • She will see static links (inserted by the author) • Dynamic link • Static links (inserted by the author) • Associated terms and synonyms are highlighted as link sources • Offering her a range of resources related to the term • She can get more general or more specific terms

8. GO(GeneOntology) (1/2) • Consistent descriptions of the major attributes of gene products in different databases • Each term has • associated textual definition describing the term • relationships with other terms • associated synonyms • mappings between other vocabularies or classification systems and GO • GO offers a huge resource of community knowledge

9. GO(GeneOntology) (2/2) • GO Ontology fragment

10. COHSE(The Conceptual Open Hypermedia Service) (1/4) • Open Hypermedia Systems seek to solve some of the problems in static link • Links are stored and managed separately from the documents • Stored, transported, shared and searched separately from the document • The Distributed Link Service (DLS) • Developed by the University of Southampton • Links are taken from a link base • COHSE extends the DLS with ontological services, providing information relating to an ontology

11. COHSE(The Conceptual Open Hypermedia Service) (2/4) • These services include mappings between concepts and lexical labels(synonyms) • The services also provide information about relationships, such as sub- and super-classes • The use of an ontology helps to bridge gaps • COHSE thus extends the notion of generic linking • There is no longer a need to own the page in order to make the link from the source to the target

12. COHSE(The Conceptual Open Hypermedia Service) (3/4) • The system is implemented as a COHSE agent, along with two supporting services • Ontology Service (OS) • Ontology Service delivers ontological information in a dynamic fashion to the DLS • Annotation Service (AS) • Annotation Service associates concepts with resources

13. COHSE(The Conceptual Open Hypermedia Service) (4/4) • Process 1. The agent contacts the OS to obtain a collection of relevant lexical items 2. As documents come through the proxy, the agent then looks for these items. 3. Any that are found in the documents provide potential link sources 4. For each source, a link is then added that includes • (1) The concept to which the lexical item resolves • (2) A description of the term • (3) A collection of link targets associated with that term

14. GOHSE(COHSE+GO) (1/3) • Both the ontology and the annotations are obtained dynamically from a copy of the GO database • The ontology is produced via an on the fly translation to the recently developed OWL • The links can help in providing both explanations of relevant terms and links to relevant materials • A key aspect of the system is its open-ness

15. GOHSE(COHSE+GO) (2/3) • Before Proxy Linking

16. GOHSE(COHSE+GO) (3/3) • After Proxy Linking

17. Discussion(1/2) • Linking is consistent between documents (for a given version of the ontology) • Given an ontology relating to any domain, we can use COHSE to dynamically link suitable resources • e.g. MGED • Separation of the maintenance of link data from the underlying content enables us to manage the task of updating databases • e.g. UniProt/SWISS-PROT links directly to the Gene Ontology, but its predecessor SWISS-PROT did not.

18. Discussion(2/2) • We believe Semantic Web technology will have its initial successes – within well-defined communities • The identification of potential link sources is done in a rather naive fashion • e.g. formatting information is included in the source • We are investigating the use of the GATE framework • Navigation of the ontology is driven by the COHSE Agent rather than the user • we are investigating support for more effective personalization.

19. Conclusion • Bioinformatics relies on access to knowledge held in its databases • The dynamic linking process helps to alleviate some of the problems with traditional Web linking • GOHSE is relevance to biological data where cross-linking is known to be fragile • GOHSE does not provide us with any new knowledge • However, it allows us to link together diverse biology resources