 |
Views
Personal tools
Navigation
Toolbox
|
Main PageFrom Vertebrate Bridging Ontology Project
Vertebrate Bridging Ontology Workshop 1, 28-29 June 2010, Hinxton, UKGeneral InformationThe Vertebrate Bridging Ontology (VBO) Project will host a two-day workshop focused on use cases, mapping strategy, areas of focus, and the identification of expert advisors. The workshop will take place on 28-29 June in Hinxton, UK. Preliminary discussions and logistical information will be posted on the wiki. ParticipationIf you have any suggestions or questions about the workshop, please don't hesitate to contact Ravensara Travillian (regarding content) or Kathryn Hardwick (regarding logistics). Aim of the workshopThis workshop will provide information for the Vertebrate Bridging Ontology project, including collection of use-case details from participating data resources and wider community, mapping strategy, areas of focus, and expert identification. These workshop agenda items will support the larger goals and mission of the VBO project: to produce a robust and user-friendly ontological resource integrating anatomical data from multiple species. Although there are many species-specific anatomical ontologies primarily used to annotate tissue-associated data in organism-specific databases, cross-species integration of such data is not possible due to the lack of a public domain cross-species anatomy ontology. A tool is required to bridge the divide between data from different animals, and leverage the information emerging from high-throughput (HTP) data generation, evolutionary approaches to gene function and regulation and functional genomics data produced using model organisms. VBO's formal goals include:
Agenda, Day 1===AM=== (times are tentative, pending confirmation)
(break at 11:30)
1:00 - 2:00 (5:00 - 6:00 PST) Break for lunch PM
(break at 4) Identification of:
8:00 PM: Project dinner at DeLuca Day 2Morning session 9:00 - 11:30 AM; lunch break 11:30 - 12:30; afternoon session 12:30 - 3:30) (breaks at 10 and 2)
SummaryPlan with Action Item and Milestones 1. Plan for a joint workshop with Phenoscape - VBO ends Feb 2011. remit: date: responsible: AI: Monte will come back with timeline AI: EBI will examine funding situation and see how much money we've got left AI: Paul to come back with info on CASIMIR funding that we could invoice for part of this meeting AI: EBI to ensure get expenses processed ASAP--decide on appropriate timeline and talk to Kathryn 2. Data mining component for skeletal datasets Aim is to deploy the worked up skeletal homology VBO vs. datasets and deploy new models in the context of the Gene Expression Atlas AI: Paul will contact John Sundberg for relevant dataset--EctoDermal mutant database AI: EBI will identify other relevant gene expression datasets in Atlas/Archive - Helen/Raven 3. Two sets of terms skeletal from lexical matching ~90 terms and ~200 remaining matched but non skeletal will be focused on. Homology statements will be for skeletal first, and remainder second. RT 4. Homology statements will be submitted to OBD and therefore can be consumed by other projects e.g. Uberon - RT This will require 4.1 Decision on ontologies in scope. FMA/MA/ZA/VA 4.2 Decision on whether any of these represent a suitable framework, or we need an alternate framework. --decide how to gather use cases and match them to framework 4.3 Decision on how to model these - issue with OBD vs OWL ASAP DOS/JM/RT - - David will commit OWL files to wiki - Discussion with Chris Mungall and Phenoscape needed - Communicate this back to project partners 4.4 Gene expression data use cases require both homology groups, and species-specific terms - need to be included in the resulting ontology 4.5 Modelling language issues - EBI expertise is in modelling in OWL, deployment locally is currently in OWL, Phenoscape is modelling in OBO, investigate any format specific conversion issues - Phenoscape/JM/RT 5. Generate a report on this meeting for BBSRC end of grant report - Raven/Helen 6. Request that Uberon convert commented classes into 'curated' and make these searchable. 7. Refunding needs to demonstrate: novel knowledge - data mining will address this usage - minimum proof of principle, best a useable resource publication agreement in the community 8. A funding proposal could be made to EC - Coordination Action - funding for meeting etc - PS Report
1. joint workshop with phenoscape 2. decide on a design pattern for homology iso_homologous to etc 3. modelling language, issues converting between OWL/OBO? and process 4. Monte requests that we generate a list of requests for phenoscape if we can use and extend that framework 5. decision on scope for VBO remaining time - e.g. ~300 matched terms in the use cases - starting with skeletal system 6. validated mappings for these 300 to source ontologies? Also validate these in Uberon 7. Request that Uberon indicates the curated classes--validate and indicate at class level that validation/human curation has happened 8. Use Terry's list of validated human.mouse mappings a. in working with the 300 terms, and b. for Uberon 9. For short term use is it worth putting some resources into uberon? - will serve some short term use cases 10. Getting VBO refunded - other users, different use cases, e.g. evo-devo not functional genomics. good strong delivered items will help with refunding. query: if same structure has same/similar gene expression patterns, structure, gene, across different organisms over evolutionary time? don't know homology, can we interrogate these ontologies to find if they have this info? do similar organisms show that homology? we have to get those options right so that can address the diversity of queries among users. means starting to get user groups together. need something to give people to look at. get use cases, questions, etc. comparably cost-effective to do that; would have to figure out way to leverage what we already have. usability experts--what do you want to do? look at these UIs and tell us why you can't do it with this. Get copy of Uberon, check each subsumption in tree whether it would qualify as homology. are these valid? if not, throw out. Use template to buiid homology-based hierarchy. "homology cut" of Uberon. FMA extensively developed on basis of use cases, e.g. RSNA radiology lexicon. SNOMED adopting FMA. propose to leverage FMA and clinical annotations. HPO using, but granularity prevents mapping because other species don't have so much. Granularity has not been shown for other organisms. Identify gaps and how to fill them. Caution against reinventing the wheel, re-duplication of efforts. Can leverage what's already been done in FMA? Absolutely vital for EQ definitions. Alignment agasint FMA.use cases, but need workable OWL version, scaled down version. Problem getting OWL-DL version, collaboration happening, but don't know when will be delivered. Needs Owl 2.0/fully tested? Timeframe? Will know within next year or two. Import terms and relationships that we assign values to for that species. Not a particularly difficult problem as long as we don't worry about back-integration, which is not a short-term focus for us. Report any problems found to Onard. Having homology mappings. 11. What established ontologies are we going to use/start mapping? Uberon existing analogy mappings need to be modified; we need homology/evolutionary mappings. 12, Where would we like to be with this in 5 years' time? Clear vision. Model organisms are having to increasingly defending funding; where are communities going? Emphasis on alleviating human disease. Learning about natural world goes along, but human health needs to be driver for impact with funders. Interface with medicine is one of most important things we need to work on. Go seamlessly from one model organism to another to human in a meaningful way, from genotypes to gene expression to phenotypes--what's required? talk same language, clear understanding of homology. Who's annotating properly in large-scale way? Database with useful annotations, can be used and made sense of by users--absolutely compatible with species-specific requirements for model organisms--speaking same language is most important point. Standardization and annotation--human community decade behind model organism community (value of FMA in this regard). Mouse Phenotyping Consortium--generating lots of systematic phenotyping data over next decade. Funders: no point in getting unless relatable to other mod orgs, humans. Amount of insight into gene function, pathway function from standardized annotations on that scale tremendous. Good human disease ontology lacking. Established, accepted, canonical way of doing things--best practices, get it accepted. high-quality anatomical ontologies for all major mod orgs, human. GO--dificulties relating to other classes, but getting there. How to move between two different phenotyping ontologies, but if ontologies not true, resultant reasoning is garbage. Pulling different ontologies in and deploying against data has resulted in improvements; most successful ones driven by users requesting terms, but a lot missing. Validating ontologies over next five years--combo of manual + increased use of reasoning. Need key ontologies in really good shape. Many more of these small expert meetings--very expensive, but very valuable. What are you spending to use ontologies versus what are you spending to make sure they're good? Droso initial iterations good, went downhill after ad hoc additions. Users will ask you to improve, refactor, but we don't know if they're any good till deployed against data. Allows you to improve data collection, synonym handling, even without reasoning. One of the things we do is improve the is-a hierarchy within anatomy. Absolutely need diversity of anatomical approaches across all major organisms, operating within same principles. Improve CARO? not within scope of grant. but we can write up our experience of using CARO. Use cases related to functional genomics, evo-devo, CARO, how do we generate this info? What are milestones for end of grant, what are plans for going forward? Clear statement of what it will provide for the community. Need a clear mission statement. Core of 200-300 terms of use cases is plenty of work in establishing homology mappings. How to represent homology? simple structure or what Phenoscape is already doing. Mechanics: at end of 1-year project is there UI that some groups are using? What for, and what would queries be? Users who want to analyze their existing resources. Also if we had something to back up UI I showed yesterday would help users who want to find available gene expression data for species covered in the db. Even if only protoype for skeletal terms as proof of concept would be great. Would be usable, because would be gains for community over status quo. Rather than use lexical mapping as bridge, map back out to all single-species ontologies. homologous to mapping: "give me equivalent to in other species". Also, "give me evidence for...". Also want to run queries that are not species-specific, because that's what many of our users need. Assuming h-to is transitive, derive from mappings we have. We have that infrastructure, figure how usability, interface. If we can do this, can use query expansion against sample database, species-specific queries. Come in with FMA ID, "I want all of the single-species terms that correspond to this ID". Collect homology statements. Action item: make sure formalization is really solid. David has OWL version of Chris' design. Attach knowledge to terms. Work out formalisation as soon as possible. OBD can do things you can't do in OWL? HP mappings, zebrafish in OBD. Load our statements into OBD; should be doable. Musculoskeletal section of HPO most complete/stable. 300+ statements for which we have mappings. Exquisite example of new knowledge gained from this. Is there a new pub from this? Stats from UI usage. Is there agreement in the user community that this is the right strategy. Use for public interface, anatomograms (how to query?), interface into integrated bit--other kinds of data get used for that; one sportal for users. All of these things are happening, we need to get the content and coverage. ID of key data sets for data mining from queries based on working ontology as well; skeletel can provide high-quality example. Integrate Human phenotype data on top of it. Need to find and obtain representative data sets. Not a problem; big problem is finding homology information. Generate new info--hey, look, we didn't know this before, but it really works! Complementary to Uberon, b/c computed/not curated. Is VBO absolutely essential? Yes, and we need to communicate this strongly. All other dbs realise their shortcomings; want it to complement their work. What are we doing that cannot be done as effectively any other way, generate new info? Counterintuitive or unexpected from existing inf? Incorporate data mining system on skeletal system. ParticipantsMichael Ashburner, University of Cambridge Jonathan Bard, University of Edinburgh John Hancock, Bioinformatics Group, Harwell Terry Hayamizu, The Jackson Laboratory Peter Holland, University of Oxford Onard Mejino, University of Washington David Osumi-Sutherland, University of Cambridge Helen Parkinson, European Bioinformatics Insistute Paul Schofield, University of Cambridge Claudio Stern, University College London Ravensara Travillian, European Bioinformatics Institute Monte Westerfield, University of Oregon Links to resourcesAnatomy and ontology resourcesOntologies in Biomedicine and Life Sciences (OBML) Presentation, Mannheim, Germany, 10 Sept 2010: http://sig.biostr.washington.edu/~raven/travillian-OBML-2010-presentation.pdf OBML paper: http://www.onto-med.de/obml/ws2010/obml2010report.pdf
Logistics resourcesTaxis to and from EBI: Mid-Anglia Private Hire, http://midangliaprivatehire.co.uk, Huntsman Site, Ickleton Road, Duxford, Cambridgeshire CB22 4QA, Tel: 01223 836 000 (Please note that we will make arrangements for international participants to be picked up at Heathrow upon arrival, and delivered to Heathrow for return flights--you do not need to make these taxi arrangements. If you need any other taxi service during your stay, Mid-Anglia provides discounted rates for trips to and from the EBI campus.) VenueThe event will be held on the Wellcome Trust Genome Campus in Hinxton, UK. Vertebrate Bridging Ontology Workshop 2, 28-29 June 2010, Hinxton, UKGeneral InformationThe Vertebrate Bridging Ontology (VBO) Project will host a two-day workshop focused on use cases, mapping strategy, areas of focus, and the identification of expert advisors. The workshop will take place on 7-8 February 2011 in Cambridge, UK.
Agenda, Day 1AMPM7:30: Project dinner at The Oak Bistro
PapersOntologies in Biomedicine and Life Sciences, Mannheimhttp://sig.biostr.washington.edu/~raven/travillian-OBML-2010-presentation.pdf : presentation at OBML Journal of Biomedical Semantics, in pressICBOsubmitted ISMB Bio-Ontologies SIGin revision journal paperin preparation OntologyTest set for ICBO paper(consists of classes of vertebrate and invertebrate acetabulum, femur, tibia, fibula, humerus, radius, ulna, and homologous-to and not-homologous-to relationships) http://sig.biostr.washington.edu/~raven/icbo-test-set.xlsx (set of pairwise mappings in Excel format) Test set for automation of pairwise mapping generationtibia set: contains homologous relations among tibia in Gallus gallus, Homo sapiens, Xenopus laevis, Rana catesbeiana, Bufo marinus, Rattus norvegicus, Sus scrofa, Mus musculus, Canis lupus familiaris, Oryctolagus cuniculus, Ovis aries, Cavia porcellus, Australopithecus afarensis. not-homologous-relations between vertebrates above and Aphidius ervi, Comicus calcaris, Drosophila melanogaster, Gryllus assimilis, and Anax imperator. http://sig.biostr.washington.edu/~raven/gold-standard-set.xlsx (set of pairwise mappings in Excel format) This page was last modified 14:12, 23 May 2011. This page has been accessed 722 times. |
|