Workshop notes 2007 july 13

Discussion on Milestones

CC. Milestone comments, while working on the branches, think and separate terms that might belong to a separate ontology for clinical trials. Don't steal them all. Forward them to us. Establishment of a new branch, or layer.

BP.There is no difference for terms for clinical trials vs e.g animals. JF:we agreed that there would not be two things CC:there will be an OCI.owl yes AR:Although OCI is part of OBi directed at a certain community, Question for that view how to manage the development process, and 2. how to develop that view. Meta data property records view perhaps

SS:True for all communities, all want their own view.

CC:Not my desire to give myself extra work

BP:We had talked about this mechanism

AR:No this is different. Not just the name but the content. OCI may not need the bulk of the instrument branch. Want to know which terms are for a community - could do with a script.

RS:OCI working group will deal with that

AR:two phases, some idea from community what would be in a view, and what from OBI would not be in OCI. Implementation second.

SS:can't do that till we have terms BP:We had an alternate term tag, use that for communities -> script. CC:My quibble is with the development process, as an OCI group and mailing list, we will be gathering terms, some branch might have the same terms RS:we went through this already, we will submit terms to branches CC:I get that, at some point talk of importing OBI, everyone seemed to agree RS:I believe that OCI is a community and not a separate ontology BP:A clinical trial role I can imagine GF:Point out that at this point he can't see a clear sep in terms of views. In NCICB caBIG we have an intermed layer between ontologies, and each domain generates their own view and data elements and specific terminologies and populate their view. CC:the milestones look fine there are time commitment issues, is anyone getting paid. PRS:I would like to know what the competency questions from OCI, I would like to know what these are and we could review the time. BP:I like the idea of looking at the file, not OCI though.

AA:Review branches for 'community views'

19. Alan's NSF funding opportunity

Insert link to proposal. [[1]]

Excerpt:

Program Title: Community-based Data Interoperability Networks (INTEROP)

Synopsis of Program: Digital data are increasingly both the products of research and the starting point for new research and education activities. The ability to re-purpose data – to use it in innovative ways and combinations not envisioned by those who created the data – requires that it be possible to find and understand data of many types and from many sources. Interoperability (the ability of two or more systems or components to exchange information and to use the information that has been exchanged) is fundamental to meeting this requirement. This NSF crosscutting program supports community efforts to provide for broad interoperability through the development of mechanisms such as robust data and metadata conventions, ontologies, and taxonomies. Support is provided for Data Interoperability Networks that will be responsible for consensus-building activities and for providing the expertise necessary to turn the consensus into technical standards with associated implementation tools and resources. Examples of the former are community workshops, web resources such as community interaction sites, and task groups. Examples of the latter are information sciences, software development, and ontology and taxonomy design and implementation.

Science commons were invited to submit. Might be an opportunity to fund OBI. Want university(ies) to be involved as PI 1. Hire 1 person FT OBI developer @ Science Commons 2. Series of short grants to members OBI community to work as visiting researchers 3. Remaining money, meeting support etc 4.

BP:More full time OBI person is better. RS:May have enough for 2FTE, and 3rd FTE as a rotating person. SS:Can you provide workshops and travel RS:Expect network projects with PI and co PI. Not at one site. HP:Looking at the regulations seems UK salary costs would not be allowable. Need a clarification costs on this. RS: I could participate without any funds. I can justify that as it would save money and effort on my existing grants. Some others might like to get funds. AR:Deadline in Aug 31. RS:OBi is a test case for semantic web for other disciplines HP:reach out to the model organism communities, Sue Rhee TAIR, Suzi Lewis for Fly, Dawn Field, Env. ontology DD: the BIRN Ontology TF could certainly provide a letter of support

Next steps: 1. Need letters of support from as many OBI people as possible. 2. Review a draft of the proposal 3. Boiler plate that can be re-used 4. Send a template for a letter of support 5. AR to send a title to be included in a letter

20. Data transformation workshop, 5th-9th November 2007

Agenda item carried over from Thursday If interested email helen parkinson @ebi

Speedos Banned

21. GF wants to sort out the SVN so he can load all the files

Revisit in PM

AR:Protege 3 and 4 are messing up the merged files, next tryinh swoop. Frame deleted issue has also been seen by NCICB's development fork. Protege added an import to relations.owl which is causing downstream problems.

22. Relations agenda item added by GF

LF brings up last slide. Barry suggested to minimize the relations, and to not define minimal properties. GF: I don't entirely agree with that. There will be similar relns but will have difft domains and ranges. I suggest that if we need to create relns with difft domain and ranges, we make difft relns. This is a design isse

GF: Next step is to continue to review and define relations contributed by branches and add to OBI's Relations.owl. Also, continue to discuss RO II. In the long term, will assist branches in identifying new realtions hidden in class definitions, model classes and classify and make inferences, and promote OBI relations to RO.

RS:why differ domain and ranges, would be a subset of the RO GF:in one domain the range is restricted, difft in another, if you are not specific then neither of the domains and ranges are specified and you can commit a syntactic error that will not be picked up as a logical error. AR:A lot of these cases handle by GCI - global concept inclusion - looks like a class expression as a sub class of another subclass expression. Have an example for part of - works for processes and continuants, p and c cannot be part of ech other. In OWL can say anything part of a process and detects errors. Not in BFO yet, will be at some point. Protege 3 doesn't allow entry of these. P4 does.

GF:so this is an implicit range GF:a localized range, domain and range are not constraints, consistency check is more than that. GF:how declare: AR:can't do in Protege, sub class reln where subclasses are class expressions. Can show some worked examples. is BFO improved in principle. GF:I would defer this discussion to offline then GF: I would say keep the relns as you make them, id in relns id if they can be re-used and need branch communication for that. Any new relns you feel are needed that can't be derived from RO we can suggest to RO. If they don't include it we can still define in OBI until is included in the RO. LF: I will keep these as is for mapping purposes GF:last concern, if you cannot find a more generic relation, say is instrument_relation_x be specific.

23. Walking through OBI Investigation use case (PRS suggestion)

PA: An intervention investigation has a specific design, e.g. factorial design (e.g. 2-factor design). AR: Another way to define is it is a design realizedAs a protocol application that hasInput something-with-a-treatment-role. That way it will automatically get classified as an intervention if you forget to call it that. BP: Where do we currently have study design? Under information_entity, and is for the whole investigation. Overall plan on a high level. We can say study_design interventionStudyDesign realizedAs interventionStudyDesign hasPart interventionProtocolApplication. AR: InterventionPA should be a defined class, not an asserted class. Otherwise we get multiple inheritance with the class we already have (something like addingSubstanceToOrganism). CC: We cannot use an intervention role as you cannot apply roles to processes. AR: If we rephrase as "treatedRole" it is fixed.

RealizableEntity
  InterventionStudyDesign is_realized_as some (InterventionStudyDesignProcess and (has_part some (has_input some (hasRole some treated_role))) 
                                                                              and (has_part some (has_input some (hasRole some control_role))) 

BP: We need InterventionStudyDesignProcess as there could be more things than just PA that are related to it. AR: The parenthetical section beginning with "hasPart..." would be the defined class InterventionProtocolApplication. CC: More description on why processes cannot have roles. The continuant has the role, and undergoes a process.

PA: How do we cover the stuff that _isn't_ treated? How does the above example relate to the factor variables we will need to define? JF: It isn't just TreatedRole - you can be treated with all sorts of things. We need to clarify what TreatedRole can encompass, and if that includes "nothing", "placebo", etc. BP: We would define an appropriate Treatment Process (we already have a few in the OWL File) such as AdministrationOfLowDoseOfAspirinToMouse. TreatedRole can be used for both treatment and placebos. "Exercise of mouse" is a treatment of a mouse JF: Placebo should be assigned to a drug, and is also a TreatedRole. AR: Treated in the sense of control vs. treated; we also need treated in terms of a substance administration. The control role is the placebo - you just have a process that only has placebo, it wouldn't be an intervention study. GF: Could exercise regiment be a treatment? JF: We need a class of study where you can separate observational study from perturbational study. Then this whole thing becomes moot. AR: We need those terms yes, but they may not be asserted classes. BP: I'm not worried at this stage if they will eventually be defined or not - we're dealing with it at the level of plan. AR: Yes, it's at the process level when we will come into problems. TreatedRole is in the sense of treated vs. control. BB: that PATO Phenotype statement example I did SPECIFICALLY addressed the issue of treatement groups. It didn't use roles - in its current "raw" form, but it provides a detailed example of how PATO and other OBO Foundry ontologies can be used to explicitly define treatment groups using PATO and the Phenotype syntax. Go to http://esw.w3.org/topic/HCLS/OntologyTaskForce/OboPhenotypeSyntaxExperiment and scroll down to Experiment Assertions. age mapped controls in particular

PA: Can we use the current OBI term control_role to explicitly state that some elements are controls, using the structure outlined by Alan?