Synopsis of the Bio-Ontologies Workshop at the EBI for MGED, Dec. 5, 2001Synopsis provided by Helen Parkinson who organized the workshop with Susanna Sansone
Alvis Brazma, EBIAlvis presented the general problem of sample annotation both in submission of data to ArrayExpress and also in querying data. (slides)
Robert Stevens, University of ManchesterRobert presented an overview of building and using ontologies, an defined both ontology types and their meanings, in terms of data, uninterpreted signals, information, data, information with meaning and knowledge, all information with menaing and an understanding to infer things and carry out new tasks. (slides)
He introduced three types of ontologies, natural language, such as GO, Object Based and Logic based. He highlighted the lack of methodology for biulding an ontology and introduced the Oiled editor for logic based ontology building and the stages of ontology design that wil help give better results. He concluded with the idea that ontologies need to be both human and machine readable.
Chris Stoeckert, U. Penn.Chris gave a detailed description of the MGED ontology which is a logic based ontology built with the Oiled editor.
Some of the post talk discussion indicated that the definitions in the MGEd ontology, example, sex were inadequate for some species and these are being revisited with a view to coverage of other species. (slides)
Duncan Davidson, MRC HGU, EdinburghDuncan presented the mouse developmental ontology and illustrated some of the problems related to describing temporal and spatial descriptions of patterns, with some beautiful animations of expression patterns.
Chemical ontology: IUPAC projects - Peter Murray-Rust, University of NottinghamPeter described an XML based system for chemical nomenclature and also highlighted the problems of working within the chemistry field where many ontologies are proprietary and overlapping. Access to these onlogies must be bought and as this has been the case for many years there is no grass roots movement in the chemistry analagous to that in biology.
Go developments - Michael Ashburner, Go ConsortiumA bit of history: three model organism database groups (FlyBase, Mouse Genome Informatics, Saccharomyces Genome Database) identified a common need for controlled vocabulary for function annotation. Their collaboration evolved into the GO Consortium.
Three ontologies: GO covers molecular function, biological process, and cellular component (slide with brief descriptions; more detail in documentation on web).
Annotation: databases export "association file"; briefly described the content & format. We encourage (but can't require) users to make annotations public via GO
Cross-products: narrative versus combinatorial approach, with simple example of combining terms from two orthogonal vocabularies (one with sucrose misplaced); also mentioned David Hill's heart development experiment; GO will use DAG cross-products for developmental and physiological process terms.
GoBo, or EGO extended go, other ontologies to be included in the go model with the following criteria: open source DAG-Edit compatible non-overlapping defined terms common ID space
The coverage of medical ontology - Jeremy Rogers, Medical Informatics Group, Univeristy of ManchesterJeremy's talk covered the wide range of medical ontologies, all different yet all addressing the same questions. He also showed that by adding too much detail that ontologies become unusable (though entertaining)for humans and result in less clarity not more. Attempts are being made to create usable tools for humans to enter data into ontologies in intuitive ways, this has moved away from form filling and towards a graphic interface where users point an image of e.g. the affected area instead of attempting to describe it as text.
RACE-D: Roche Affymetrix Chip Experiment - Detlef Wolf Bioinformatics, F.Hoffmann-La Roche Ltd, Basel.Detlef presented Roche's in house LIMS and array database which has been developed to store and analyse data across different studies and sites.
ArrayExpress: data submission and annotation tool - Helen Parkinson - Microarray Informatics Team, EBIMIAMExpress the EBI's MIAME compliant annotation tool was presented, a demo can be be seen at www.ebi.ac.uk/arrayexpress The problem of ontology creation for ArrayExpress was highlighted, and the lack of user knowledge and user defined ontologies for microarray data were discussed. (slides)
Open Discussion:Alvis Brazma introduced the open discussion with the following points:
Outcomes of the meeting: Jason Stewart, Open Informatics will circulate a list of all the ~70 places that the MAGE-OM model references an ontology. These will be circulated to the ontology and MAGE lists and as some of these are very small controlled vocabularies can be collected in the first instance and then ontologies can be constructed. An example is surface chemistry.
Preliminary plans for a large ~200+ people bio-ontologies were suggested. The meeting would be held in Hinxton end of 2002. Mike Ashburner has been suggested for the head of the programme committee.
There was a discussion about whether this should be in the form of workshops and tutorials lead by the great and the good or a more formally structured conference. There are also issues of whether users of microarrays should be invited as well as informaticians. The benefits of this would be some training for these guys in the construction of ontologies and also some user testing of the MGED ontology, and some information gathering of instances for ontologies. Funding bodies should also be invited as there is no dedicated funding for the MGED ontologies or the gobo effort that the GO consortium are suggesting.
There was a curation discussion about how terms submitted to array express should be curated and fitted into the MGED ontology, John Quackenbush's brilliant suggestion was an EBI approved list vs a list of user defined terms. These would be visible from the query interface and submission interfaces and flag datasets for MIAME compliance and approved term use. It was suggested that the model organism databases would be a good place to start looking for EBI approved terms. These databases should also be represented at the conference.
The issue of synonyms was raised and is is clear that these will cause problems and will most likely be resolved by curation and that their use should be discouraged.
The MAGE developers raised a problem QVS concept which has been included in the MIAMExpress submission tool. they suggested that it be exported from MIAMExpress as an ontology reference not as a QVS.