Workshop OBI Vancouver 2008 Jan 29 notes

Tuesday Jan 29 Notes

Action Items are here: https://wiki.cbil.upenn.edu/obiwiki/index.php/WorkshopNotesVancouver2008

Agenda

Each branch will be allocated one hour for discussion. If some time remains at the end of these 2 days we can come back on some of them.

format: 10 min presentation, 50 min discussion

morning: 9-12

9-10: CENTRA CALL: 9-10 PST

Skype available seems to work for free-to-listen - message susanna

Moderation - self-moderate your slot. Will cutoff discussion at 59 minutes, leaving time to recognize contribution of others.

OBI Completion Milestones

1) OBI "Working" Release - Have OBI at a stage we are ready to have people use requires:

   * Curation_status curation_complete OK (TBD) on all terms - March 1
   * Individual review of OBI
         o Term;problem;solution submitted March 15
         o Problems without suggestions reviewed by group, then advisors 
   * Use cases worked out through OBI to demonstrate "it works"
   * Versioning policy (and other OBO Foundry requirements http://obofoundry.org/crit.shtml)
   * Documentation on how to use 
   * Each branch list of requirements to be completed to complete by March 1.
   * Branch review of other branches

2) Manuscript requires (1) and:

   * Authorship plan
   * Some text 

3) OBO Foundry release

RB:Issues with branch development, easy to find things that are wrong, better to generate solutions. Proposal in AI below.

AI: Term problem solution sheet to be generated by RB to be sent to advisors if needed

RB:problem that we don't have any tools in the short term for naive users, but docs will help us

RS:We need a user documentation plan, and assign people tasks

HP:what's the deadline for the release?

RB:deadline for wrapping up issues at this workshop is March 1, by end of the week we'll decide on a release date

BB:Test plan,

AI: Develop a documentation plan for OBI and assign tasks JM/MC/BB

AI: Develop a test plan for the OBI release, document and execute BB

SS:If I need to add terms in the short term, how do I deal with that?

BP:We need the terms that work for cross branch terms

RB:I would say I need this, and offer a solution

RS:useful if you add to existing h'archy for the other branches

SS:do we use the tracker

RS:Volunteer to do a use case clinical study

RB:Flow cytometry use case

JM/HP:AE

PRS/BII use case

BB:BIRN use case

JG:Immunity use case

HP:What will be the format?

RB:Protege not the first tool that people will need to use

BP:Free text query and translation into OBI terms

AR:Could put them into my triple store. What is the goal?

RB:We want people to start using OBI, and we need to show what OBI is useful for

JM:We are identifying a different user group here

HP:I liked the mapping of the spreadsheet from RS and the use case translation from Bjoern, these are good representations

SS:I think we are mixing two things, showing what OBI could be used for, and we show ourselves for the next workshop, then we can use for other people. I want to work on what I need.

RS:I was thinking making sure that the high level content was there

HP:I think we need to do this for us as internal review and then put it out there, we need to do this.

BS:Grant proposal, all the use case descriptions will be useful. We didn't have a long term plan for timelines for 5 years. We need also to think in longer terms.

RB:If you want to access some funds then talk to BS

AR:That's a larger question for longer term questions

AI: Developing a long term (5 year) management plan for the OBI grant - AR/BP/BS

RB:What's the deadline on the use cases?

BP::Lets do this at the end of today and we'll know then where we are.

RB: I think we will need more time than that

RS:If we are talking about July then I would like to do in 2 phases

RB:Manuscript, all of above must be done. Authorship. Came up as part of the discussion at dinner. Need to think about that in advance as people will commit time for this. Two ways that the authorship is done, authors by name. Co-authors and last get the credit, needed for careers. OR Authored by the OBI consortium, could be authored as a list of credit. Personally think if we go for OBI consortium junior people need the authorship, the effort needs to go in. Is it possible to recognise people with more effort than others. I know this is contentious

BP:I also will author a personal publication IEDB using OBI, put out many OBI related publications

AR:Want to work on small micro parts of OBI, e.g. Information Entity

BS:Swamping is good for raising funding. We need that for the grant. We discussed in some detail at previous meetings, Chris S. There is a management proposal and there are branches that people can get the credit whenever a branch is used. If Chris has worked this out then we need to see if the previous model works.

RS: I do think additional manuscripts allows for additional credit, even if in middle of the pack.

BP:I see a Nature Biotech. Article, we can use the fact that there will other papers to help determine that paper.

AR:Other papers, suppose I want to write a paper re use of Roles, what's the protocol?

BS:Writing is hard, everyone will not do that. People should just do it, contact the other relevant people, e.g. role branch, and if you want to do this send round a larger group, and most of the work will always be done by the first author.

BP:This is important, we need to focus on OBI Deb

BS:We need to think about different journals. Journals - Journal Royal Soc of Chem, Colin Batchelor, already involved in OBO foundry efforts and we can use the journal to work on that. If someone has a chemistry use case contact him. We need to target Nature etc as well in the long run

RS:I don't want to make decisions about authorship at this meeting.

RB:I want a policy.

SS:We put something in a doc

AI:Find the original doc for authorship policy and take to a coordinators conference call SS

AR:do small papers preceed OBI

RS: I don't think we should put out a paper before the main OBI paper

AR: The publication could take a year

RS: Then we need some impetus

RB: We need to get this done, and so authorship is a key thing

AR: What about workshop publications

RS: We are all doing that should be the public face of the presentations

JM: There were some questions during a recent presentation I did. If we release papers now then we need answers to these questions. I would be concerned about releasing it like it was finished

CS: If there is an OBO foundry paper then the decision was that it is by the OBO consortium

RB:I brought this up, as there are some junior people that are doing some of the work, we need a policy for people that we can recognise.

JW:One way that this could work for my community would be a real practical example with real data that people have some interest. Will there be instance level data. We'd like all the data in that format? Want to see that level of integration of that data. Would be a nice way of dealing with data in that way.

AR:that will not happen without active work to do that. That is worth considering as a monthly meeting.

RB:Process requirements for OBO release are now in discussion. We have two more weeks to do that.

BS:OBI will need a couple of people to be associate OBI editors, and their role will be to nominate coord editors every couple of years

AI:OBI coordinators should nominate 1 or 2 people to be OBI associate editors - agenda item for a coordinators call

1. Plan Branch - Philippe Rocca Serra (no ppt - just the ontology)

• InformationEntity is what's realized, this is basically what is plan - should this be renamed as plan?
• Should plan, protocol and algorithm be merged has been discussion on that. Alg could be a specialization of a plan?

BS:There is a shareable plan which is documented, when someone realizes it then they have a private plan.

BP:fine, can we call the shared plan, plan. Rename then information entity plan?

BS:No. Why do you want digitalEntity to be realizable

BP:It's not

BS:You have non realinfoent, and infoent

AR:InfoEnt are all plans

BS:solution call InformationEntity, if you are happy that an algorithm is a plan then rename plan is OK

AI:Edit:InformationEntity will be renamed plan, synonym RealizableInformationEntity DONE

Alan fixed a minor edit error that caused the file opening to fail and note that we need to work on a branch basis to avoid Protege errors.

NonRealizableInfoEntity has one child. If only needs one child then maybe merge.

AI:Edit:DigitalEntity->kid of NonRealizableInfoEntity, Plan will be a sibling of NonRealizeableInfoEntity, both children of InfoEntity

BP:Do we also need to deal with DENRI and Plan?

PRS:want to deal with Plan first. Should these be disjoint

AR:probably, issues with the protege adding disjoints. Want to remove all disjoints now and add a script later to deal with that. All should be pairwise disjoint apart from couple of cases where BFO is not clear. AR

AI:Disjointness, currently incomplete. AR will strip all out and script adding back in. Al will be pairwise DJ apart from a couple of places in BFO where it's fuzzy AR

BS:Gene seqs are not info, they can be represented as imformation

BP:looks like it does now, as this is a BFO issue if these definitions are now considered to be wrong

AI:Definition of InformationEntity to be done by BS/AR/BB</b>

PRS:issue with algoritm being define in terms of function and the logical overhead

BP:this is replicated in planned processes. Def for me of alg is something realized as a data transformation process

AR:This excludes cases where a robot is doing something

RS:still data as input, output is data that drives the motor

BP:If you want a set of robot instructions where movement is an output then we can add it later now we need data trans

AR:what about alg as something that deals with sw interpreters

JM:call it SW algorithm, as it's not alg in general

PRS:go back to plan, need objective or goal etc where plan. Should all plans have a goal and need a role.

BS:can't make a generic goal for a protocol that is it's applied in a protocol application, it's a trick sounds like evading the question

BP:it has context though

BS:ProtocolApplications have goals, and protocol has same goal. The goal of a protocol is not to be applied.

AR:can define in terms of inputs and outputs, goal is to get from inputs to outputs

BS:no. You are using output as goal then.

RB:State my goals for an expt is within a population, need a testable hypothesis

HP:in some cases there are non hypothesis studies, but there can still be a goal

AR:default goal could be in terms of the protocol application, better to do more clearly

BP:Do we need a class goal - another info entity, protocol has a part goal/objective

PRS:objective is there, it's a role played by a statement

JF:Objective should be moved from role. It's in DENRI not role

RB:I agree with Barry, that there are things that I want to test, and there will be a protocol application

BP:We are separating goal which not realizable from plan which is realizable, that's a good split. If you goal is to cure cancer, you don't realize that.

AR:realize doesn't mean that it can happen, but that action or process will follow from it

BP:the thing that translate to action into plan, not the goal

AI:EDIT:Objective now child of InfoEntity

AR:not happy with goal not being realizable

BS:objective in plan and objective as realized is different from the thing that is cardinal part of plan. Think here we mean plan in first sense and it can be realizable

BP:Is objective a plan

BS:no it's part of a plan

AI:Define Objective add a definition, synonym is GOAL DONE

BS:Objective should be lower in the h'archy than plan.

BP:can you realize an objective without a plan, must have a plan to realize it. Therefore not in itself realizable

BS:not clear if it's an info entity from working through the definition

PRS:we need to decide this now

BS:This is an ontology for protocol driven research, everything is connected to protocol application

AR:I want to make a KO mouse - is that an objective. This is a process to make a continuant

BP:objective is use a KO in summer 2008

AR:this is realizable, future point

BP:I need a example

BS:I could buy it in 2008

BP:I will have a plan that I realize as well as the objective, this is std project management. The objective is separate from the plan. It's important to keep it sep.

JF:This is now starting to sound like a role.

JW:this couples implementation with output

BP:not all objectives have a plan

JW:the plan comes after objective is implemented

BS:BP wants to create a KO mouse, and to fulfill this objective is the real world part of objective. Fulfillment is not realization. Realization is always a process. Objective is not a process, as its not the process that is the objective, it's the boundary of the process. protocol application is a process, you processes like climbing a mountain, being on top of a muntain is not an achievement, it is getting to the top by climbing, not being there.

BS:OBJECTIVE belongs under NONREALIZABLEINFOENTITY

AR:If you can use achieve - you need to have an entity achievement

BS:there are already good ontologies for achievements

PRS:Want to see an application of the ontology. How can I do that in OBI. There is no constrain on it. We need to say what are the means that we use to execute the plan. If we can't do that then we are failing, or I have misunderstood.

RS:All plans need instructions and objectives

PRS:and the means

BS:will be the end point of the process, special kind of process

BS:plan is to watch the fish - what's the objective

RS:to make associations

BS:this is then a plan that doesn't have an end point objective

RS:there are objectives that don't have hypothesis e.g. collection of data

RB:you'd be observing for a reason, achieve some kind of understanding, which is an end point

BP:now hypothesis, what's the difference between investigation and protocol application. I propose that protocol application never tests a hypothesis, but an investigation sometimes does.

AI:Resolve the difference between Investigation and ProtocolApplication - BP suggests that PA never tests a hypothesis and that Investigations can have hypotheses - THIS WEEK

PRS:what about a dt where there is a hypothesis test

RS:as soon as you do that it's an investigation

MC:would like to see Plan, Hypothesis, Inv and PA interbranch issues need to be resolved

BB:Agree. Also function is in there

BP:Don't see why function is in there. We care now about hypothesis

BB:May relate to objective, may use an assay to use a thing that helps measure function

BP:not need to discuss function to differentiate between Investigation and PA

AR:Investigation doesn't need to have a plan, you don't know all that before start.

BP:A funding agency can have objective to cure cancer. When I list aims those are objectives. exp design is then plan when I apply for money

BS:all technical terms in an NIH grant might need to be in OBI, good validation step. Plans can be general, and sometimes need to be modified.

BP:Investigations need more than one PA. PA can also have multiple PA. What else is it that differentiates. Could be conclusions that are not simply an output, this is an interpretive step.

JW:the outcome of a protocol is known, the outcome of an investigation is not know at the outset

RB:RS used the work experiment. Are experiments different than PA. or did you mean Investigations

RS:There are gps of specimens that are an intermediate layer

HP:This makes things more complex, suggest that use in that case a sub investigation

PRS:Conclusion is under role - may need to move.

AR:Is the example of the nurses study an an investigation

We think it is

BP:Barry, PA gives instance level data, e.g. has these outcomes. Conclusions are general statements.

BS:Often true not always true, you might have an experiment to find an instance e.g. a planet found

RB:Assay is a type of PA. PA is about instance level data, has multiple info.

BP:Study we tried to take care of in San Diego. we need these terms, we couldn't decide what the difference is between PA and Study - we had a defined class of multiple PA that are in a defined class based on some thing e.g. done at the same time.

We decide to focus on difference between Investigations and Conclusions

BP:where you find an instance like a star I make a universal statement 'new star' conclusion

BS:BP is right, according to Popper says all science is designed to falsify hypotheses - testable hypotheses - we can't prove hypothesis is true, we can prove not true. Science not interested in the instances, and there are hypotheses that have been tested that = science Goal is to get to testable hypothesis but do so processes to refute hypotheses.

BP:PA give you instance level data which can refute a hypothesis. Investigations should make a new hypothesis again testable.

AR:this is a rathole. We can talk about conclusions as part of documents. Now we are doing something hard, and we may not make process.

BP:If we do that then docs need digital entities what kind of entities are these

BS:PA provides evidence to support the conclusion - we can define conclusion from there

AR:PA and Investigation - Inv get funded for, has a narrative object that contains conclusions etc,

JW:in a PA doing a well defined steps, in an Investigation not the case

RS:Assay has a finding, conclusion is an outcome of a investigation

AR:I don't want to change PA.

BP:finding is data from an assay, is instance level data. it works. Alan says that we shouldn'y define hypothesis is hard. I state that any statement beyond instance level data are conclusions and these are investigation level data. If there is a problem we can move on

BS:Conclusion is too general, conclusion goes in the paper.

HP:don't think terms of paper, there are too many things that are conclusions that are not published.

BS:defining conclusions: the conclusion of an investigation is an info content entity resulting e.g. E.g. Samson Tu's description of autism.

AR:measuring pollution example - doesn't satisfy universal,

RS:measure vs decide - decide=investigation measure =PA

AR:need to excise the proxy for in invesigation. Reserve investigation to mean those cases where there is a class level finding. Some things might not them meet investigation in OBI

HP:what about cases where there is no conclusion and no paper just cell lines

BP:Then it's a study, not an experiment

MC:we have two levels id perhaps, processes that we do and deal with the 2 levels and see if we need the other levels. We agreed on PA being at the instance level.

MC:BP had a good definition of investigation and PA

AR:proxy for messed that up

BP:some conclusions are uninteresting. Conclusions for investigations are interesting across investigations.

HP:If I add a conclusion then I had a bunch of PA and now I have an investigation

BS:That's a non cannonical case.

HP:How does that help me

BS:This is a problem is we define investigation as a protocol appl and a conclusion. These are not a problem as they are non in a meta analysis or they are non cannonical

JF:why is production of data not a conclusion. In role we had an outcome that is data production. outcome is a subclass of conclusion

BS:too vague, we want to know why something is an investigation.

BS:If you collect data for drawing a conclusion then you are performing an investigation.

RS:As soon as you want to add a conclusion that is a new submission

BP:no the intent was that an investigation was done

HP:OK and I can agree with that

AR:biobanks might submit data?

RS:then they start an investigation

AR:no

RS:you are making incorrect assumptions. Needs to be an investigation using biobank samples

Consensus has been reached. We have agreed that an investigation has a conclusion and this differentiates it from a PA

We are now adding to protege. Placement of Investigation is OK as is.

BS:Wants to add survey - an investigation without a conclusion.

HP:why can't this be done by grouping into assays?

Definition of Investigation:

Definition of ProtocolApplicaton:

BP:MaterialTransformation, DataTransformation, etc easier to define start with these.

AIChildren of ProtocolApplication are redefined based on this discussion

BP:Now into PA review. Want to say that there are 2 processes, get data and separate cells in flow sorting. The assay then is only data out.

HP:but there is still material out

AR:then it's not called an output

BS:Assay outputs data is about qualities (or other features) of the input material (dependent continuant) or evaluant

AI:to prepare a better definition for assay based on the discussion and convince Alan that it is always a dependent continuant

HP:you need to define generally in terms of materials that could also occur in chemistry not specimen

RB:we want to add evaluant? Is evaluant in OBI? Needs to be added for this definition

AR:if material is being used can't distinguish reagent

AI:EDIT:evaluant is added as a role, RS will make a definition

RB:protocol application has instance level data -

BP:that's assay we defined there, we should now define Material and Data transformation next

AI:Edit:re-define material transformation - added to OWL file. .. is a protocol application where the objective is to produce output material from input material'

BS:are mice inputs to an observation?

All agree yes. And this is then an assay not an MaterialTransformation. Agreement

AR:overlapping issue?

BP:cell sorting, sort cells, while you sort you count, count is an assay, then you have 2x process of materialseparation and an assay simultan

BP:These happen in parallel

BS:that's Ok if works for the community

AI:Edit:Check the definition of DataTransformation and reworded for consistency - DONE

PRS:another class - 'Waiting' doing nothing to a material for a duration of time. No-one wants waiting, it comes under the definition of MaterialTransformation

BP:we have the general problem that we want to connect bits of a protocol. We don't have any temporal concepts. As we don't have the parent sub parts here so we don't gain anything from adding waiting

BS:waiting goes in the human behaviour ontology - it's too general for OBI

PRS:Material and Data Trans are processes that have the 'objective to' included in the definition. Should this only be on plan and at least should be an objective on all of the children of PA

JF:we had time tiggers to deal with the use case Alan described wait 2 mins after washing.

RS:I prefer incubation to waiting

BS:all these have to deal with time, and this needs to be resolved

AI:Edit:waiting was moved next to acclimatization and is no longer a child of PA DONE. Needs to be resolved when time representation is resolved (not done)

MC:we need to make sure that all the PA definitions are consistent. All are checked for consistency and minor modifications are made.

AI:TIME - needs to be discussed and a proposal produced for representing temporal events, waiting etc in OBI

BS:PA differs from a planned process in that a protocol is executed not plan.

Lack of conviction about this, we look at the definition of PA and planned process.

BP:we agreed that investigation and protocol are different. There are 2 types planned processes - protocolapplication and investigation. This makes the difference

We check def of Protocol for consistency. We have circular definitions.

BP:the 3 subtypes if protocols, we could define the ProtocolApplication in terms of the union of its children. Would solve the problem.

BS:are you sure that dealt all cases where there are 2 PA going on at the same time. issues with disjointness then. Some investigation where assay/material transformation - if use disjunction then can't represent them in investigation. There is another PA - aggregate PA.

RS:this was in the use case - composites

RB:is it a child or sibling.

BS:Aggregate protocol application is a protocol appl according to Alan

BS:BFO issues, processes and aggregates of processes are siblings - problem in BFO. Define 3 PA as PA units and mix and match to make larger and more complex PA

BS:before using BFO need to fix granularities and we don't have this; defining PA in terms of kids then is a bad idea

BP:I made this proposal, lets go back to the question, how is PA different from PlannedProcess - I withdraw suggestion to define in terms of children, but we still need to know how PA/PlannedProcess are different

BP:do we want to continue this discussion.

BS:Has some notes of things that need to be changed, need to work on these. RB volunteers Melanie, James.

This was unresolved, AI have been added so we don't lose track of this.

Bill Bug, Function Branch Report

• Function tricky - no-one using BFO function much, no community experience
• Issue to separate function and role
• Function branch alone is useless, and may be thin, most of value is in relation in linking bits of OBI back together.
• OWL implementation will need new relations
• Function in BFO is a RealizableEntity - definition is horrible, inheres in object of a kind defined to perform x function
• BFO list not helpful in guidance on using function
• developed OBI examples, are quite complex
• Function in OBI should have limited scope in Instrument, Plan, PA,
• Objective, role, function not orthogonal
• what are the use cases e.g. separate molecule by size
• Instrument and reagent use function, and the dfunctions are realized as PA, got mirroring between function and realization of function same as in other places in OBI. Define by specifying PA.

HP:did any of the use cases submitted speak to the function branch?

BB:not checked in detail

BB:would like to know from branches how they thing we should proceed with function.

BS:seems that BB is clear on BFO, and also proved need between function and role

RS:don't need to say that heart pumps blood in OBI. My use cases need function not role. Equip manufacturers need function. I need role.

BB:Simpler better as looked at clinical assessments, you often don't get anything more than glucose was measured, don't know how. How would you represent that.

RS:these are assay types

We think function inheres in reagent and PA, and no other use cases have been found.

MC:distinction between function and role - who decides which? In role in BFO - role of artifical heart is to pump blood - that's a role

BS: role:borne by an independent continuant e.g. person, not need to play that role e.g student, roles are optional, many socially assigned, some from bio reality

functions: entities esp independent continuants have functions as they were made to have them. all hearts were made to pump blood, even if some do badly. artifical heart was made to pump blood, - then it's a function, error in BFP

dispositions:entities have these and are not a design feature - going grey for e.g.

BB:not worried about this distinction. We have now thin specified cases. We want to see how would use in OBI.

RS:functions are different from PA, but we should work on one branch and then have a way to deal with parallelism in other branches

BB:function seems to be covered in other bits of OBI

BP:leave function for now, and see what it should be later after we work through PA/

MC:Clear answer - artifical heart function pumping blood - centrifuge - function spin, never have the role centrifuge - will it have both? Will it never have a role.

BP:separator role never held by a centrifuge -

BB:it would in some cases if some device was specified to play role spin.

RB:was built to have that function

so function or role? still not clear for the centrifuge.

JW:can any object have multiple functions? Yes, function and role then validating. Can then be validating when represented in data acquisition

HP:so summarizing then if something is being used for its std function then role=function, if there is a case when it's being used for something not designed for then you can use role.

BP:we could define cannonical function for instruments then.

JW:is the scope of functions and roles the same set - not nec, but mostly

BP:we define the things that live in the process, and we only need to define the process

RS:I think in my use case I need the role an instrument plays when performing a function. not same as the role in the PA

JF:RS saying that he uses a centrifuge in a PA, you describe as a role, and poss a function DT branch has role of normalization, may also be the function of an algorithm

AR:question is there a function and role that are the same, answer if there is a function it's a realization of a function, not also a role. RS needs the role.

MC:For an instrument it can be obvious, for a protein is much harder. Is reagent a function or a role.

BS:have a protocol, 3 centrifuges, 1 and 2 backups. have a backup role, and first run is a role . Roles are optional and how you decide to use them in a PA. The GO tells what the function of protein is, the role is your role, as a label, as a suppressor.

MC:reagent is always a role not a function. said opposite before

BB:yes, but seems role is more important for reagent

BS:role is often dependent on function

RS:there are some molecules that have a role in an assay, not to do with function, some reagents are produced where it realizes function in an assay. two types of reagent.

BP:we referred to the doorstop problem doorstop=function doorstop, centrifuge has role doorstop. so instruments can perform a role so AR is wrong.

AR:rare that have instrument that is not designed fr a function that can play the role where it is playing a role that another instrument was designed to do. In that case need both role and function.

JW:more generic the instrument the more roles it will have

AI:MC will write up the discussion on Role, Function and what we understand for distribution and we will then develop a consensus on going forward on this. No consensus reached yet.

AI:BS will resolve BFO issues reported to him, all examples will be resolved by May 4 2008

Ryan calls time again.

Role Branch Jennifer Fostel ppt

• Hypothetical other branch - need for roles on processes
• Tina's use case
• hypothesis is included at present, happy if included elsewhere
• Need for organization and aggregates of people to be in OBI

AI:BS/AR to explain why roles cannot be born by processes

AR:need to fill in bearer for each role, and the process in which it's realized

BS:internal review board was built to do something it may have a function

AR:if it's a social inst then it is treated as role -

HP:you said that you could put in the process in which a role is realized, you don't want to define all processes.

AR:we might not complete the logical definitions

BP:use of space role suffix - BS agrees there may be a reason to allow that to help distinguish could get in a mess if we don't do it.

JF:OK

AI:update naming conventions page to allow this case of suffix role being allowed in this special case

AR:drug role, could be therapeutic instead.

AI:Susanna to supply subtype of compound role toxin etc

JF:issues with defining metabolite, not always in the body

RS:we need to make sure we are not defining the other metabolites in cell, suggest active_drug_metabolite

MC:do we keep metabolic process?

AR:not an issue here, so we use from GO

MC:PA could be things are investigated, metabolic process could be investigated

RS:phosphorylation natural - not in OBI if you do it in yr expt then it's in OBI. Treatment and result go in OBI

SS:metabolite are in chebi, should it be there

JF:not the compound, derived via a process

PRS:should be in chebi. Here we mean active metabolite, and this is in chebi

JF:not need to be a role

AI:Philippe to make a list of obi roles that should be in chebi

AI:Define hospital when do organization

JF:Biological replicate - the other batch of cells, or the process.

HP:the process is replication, this is a role. BS agree

JF:reference treatment - reference medical device is instrument, sham treatment - bearer is a protocol application - mimic the harmless portion of an intervention

BS:treating the type not the instance, sham is a type of treatment, so particular sham treatment is the reference, you need an instance of a sham treatment. When we have a reference substance - is that an instance or a type?

BS:baseline participant - is that a person, or any person who in the context of that study

BP:in design you speak in general, when you start these are instantiated

BS:these are then playing a role, the treatment is not the same. you create de novo. every instance of sham treatment is created as such, it is not playing a role

JF:is a function

BS:these are processes created fr a purpose they do not play a role

BS:not function

BP:these sound similar as baseline sounds like belogs in study - sham_treated_participant

BS:some reference participants are determined by that fact that they had a sham treatment

RS:sham is a different treatment, not some other treatment playing a sham role. Can't equate with role.

AR:we will get multiple inheritance, e.g. heart surgeries, each case there can be a sham. Lead us to role again

RS:define as surgery when do the dt then have the gp play a role as the comparator. Tie role to a data item or to a person

BP:a treatment is a case in an investigation and only in the case of a single investigation

BS:shaving is it a sham treatment?

JF:could be,

RS:not always so, contextual

BP:so animal output of the protocol was sham treated

RS:take the animals, get results inhere in the animals not in the processes, and need to define the animals not the processes

RB:close branch and this term. had an hour

AI:no resolution on processes being bearers of roles, functions, variables, propositions needs to be resolved

AR:experimental factor - study variable -

RS:already constant and conditional variables

AR:I don't know where variable is

RS:in same place as conclusions and investigations

PRS:variable is a synonym for a factor, time is not part of the treatment in one of these examples it's not a variable.

BP:if you are doing a time course then this is in protocol application. If you conclude something about time then your conclusion is different, how do you design what you do

RS:general case of independent var, conditional population a disease b no disease, or time

JF:we are calling those experimental factors

AR:We have no factor or variable

PRS:can we not start with time, it's already a problem

AR:is variable part of a hypothesis - certain type of hypothesis - BP agrees

AR:about would link variable between drug and it's dosage

BP: in describing the drug you would have all the details.

JW:needs to be part of investigation

PRS:intervention design and declaration of the variables, some variables are not independent of each other. can't have only compound

HP:we need to deal with variables linked with study and also links between variables

BP:suggest that variables are on investigation design, it it only a relation or is it a thing

BS:suspicious - not a role, these are type entities, no instances unless we choose the types

AR:if have an independent variable, where the protocols are different you can do a consistency check

BS:if we put in part of investigation design we would need to say something about the types in reality that can serve as factors

AR:can do as types, define relations

BS:we maybe able to narrow the types

AR:not a class, is a part of has_independent_variable has type

BB:if we do as a relation what is the domain. Study Design.

AR:use a relation and then use to related to material and protocolapplication

PRS:what's the inverse relation, does this break

AR:we still need a link from specification to the type

BS:we have a new term specification of study variable - is a part of a study design which specifies which variables are objects of interest in the study

AR/BS will discuss relation proposal or the new term proposal for variable in study design

JF:roles on propositions, if we move hypothesis this goes away. mostly resolved.

Liju Fan, roles branch report

• has_base - what is domain?

BS: belongs to maths not biomedicine, discussed if things should be included

MC:we don't want a maths ontology, we have a need for these, 2, 10, we have other relations

RS:can we use has parameter

AR:what do you do where a function has multiple inputs, is this different from other functions would parameter use

PRS:no functions these are log_transformations - has input role.

AR:could we deal with this as an input

RS:input is data, output is data, base is a parameter.

AR:input has role, implies that there is a role, base is a role

BS:has_variable, is a quality not a relation, is a quality of a function not a relation. DT can live with roles as a quality. But we need something to inhere the quality in and we can't

BS:math entities are like types, don't change. seems that types can play a role, we select them, same thing might apply here. they might be bearers of qualities. BFO says that instances of dep continuants can't have qualities. Says nothing about types.

AI:types can bear quality like things mathematical function and features and this is a workaround here for math entities. Outside BFO so needs to be OBI and would mean that we don't need a relation. New class called feature - has_feature relates a type to a feature, all base, degree and number are features of type. DT branch will look into this.

• repels/has_affinity_for - outside obi scope. Kevin Clancy proposed as did protocol application branch

BS:suggest that you create whatever relation you want and make a placeholder until chebi or someone steps up.

BB:affinity is a quality

BS:send to PATO

BP:quality will not help, we want to say something about binding

JW:how deal with hydrophobic interaction

AR:already in biopax land.

BP:need to specify that this label binds this molecule for protocol application

AR:these are functions, there are binding function in go

BP:we need to connect objects

BB:PATO has this.

JW:what level of detail would you go to

BS:depends on what's needed by a branch

AR:events like this are in biopax obo - not in any place now. good to have an example, so we can do something

BP:binding assay - detects a binds b - needs the relational quality of A for B

JW:what to do with cleavage

BP:all branches try to submit elsewhere and keep if not elsewhere, we only submitted these

AI:binds/repels etc needs to go to PATO, and Bill will provide an example for how to do a relational quality and will also talk to Chris Mungall about adding to PATP

• Instrument branch has submitted a bunch of terms to relation

AI:relation branch terms will be sent to RO2

MC:there are some that did not yet any feedback

JM:when you get issues with terms, you need to help us by bouncing things back to the branch

RS:we need a better way of solving this, there are too many things here

MC:today I need to say that instrument has model_x

BB:models are classes, and manufacturers are instances

PRS:we are still struggling with using OBI classes to input data. We know that instrument has model, and we need a way to view that.

AR:model, instrument is an instance of a model, language is database centric. This is hard in realism based thing and we didn't want simply to reject. Proliferation is part of the database terms.

BS:model - if we introduce an aboutness relation then we already can use this to deal with model. There will be an InfoEntity that is model number

AI:BS will review relations s/sheet and document the approach for dealing with has_model that need to be represented in OBI

• runs_software - think the instrument has part that bears a sw interpreter, how related SW to Instrument

RS:is this a special case of a realize a plan? So use realize in this case

BS:the SW was built to run and the machine was built to run SW, is a realization of software

AR:so instance of a machine needs to say that it can run SW

BS:disposition maybe function

HP:we need to solve the problem link HW to SW

BS:SW is a plan, is realized by the machine - the relationship is therefore instrument is able to realize the plan. The machine has disposition of being able to realize the SW.

AR:need a relation now between disp and SW

Disposition is in OBI and SW is not and we need the special disposition to deal with that. BS will tell us what SW is. Generically dependent continuant and specifically dependent continuant are linked by concretize already, but we need to specialize it for SW. Will be sent to RO

• has_image_format - could be has_format, this is a plan, and an instance bits of a pdf file

JW:you made file format specification already, use aboutness

BP:file that is generated by a realization of a file format specification, where tiff, gif etc are file format specifications

AI:issue that we need a quality like thing of gifness. could also be solved by using feature om OBI as above, again a work around BS:

AI:branches to prioritize which relations are needed and tell relation branch

BP:we need to define something that works as a placeholder.

Melanie Courtot - Quality

• few qualities in OWL, CS is the point man for PATO-OBI
• issue with PATO not accepting experimental qualities- emulsion etc from biomaterial. They now agreed
• terms from DENRI - voltage etc need better definitions - the definitions were old for datum. They agreed that these are qualities. AR thinks they are dispositions.

BP:Pato is full of things like this, these are qualities of non biologicals.

AR:Chris said parameters will not be accepted.

BS:Settings on equipment are an OBI thing, if exists only in a biomedical investigation belongs to OBI.

AI:We will need to deal with Parameters, probably in instrument

AI:JW to look at use case docs and suggest some for stuctural biology