Workshop OBI Vancouver 2008 Jan 28 notes

Presentations and Discussions from branches
1. '''Jennifer Fostel. '''

Presentation by PPT.
 * Need for adding roles to other than independent continuant e.g. data transformation processes
 * There are study variables that are factors, and cases where variables are not factors. Added concept to factor, and then need a link to a variable which is in OBI
 * Outcome - StatisticalFactor is a subclass, Outcome=result of plan process etc
 * StatisticalFactor - factor that is a conclusion that has been found to be stat. signigficant
 * Big issues with analyte

Analyte Conclusion by AR [[New page|https://wiki.cbil.upenn.edu/obiwiki/index.php/Analyte]

'''AI: AR/BP/BS - will refine the analyte proposal. There are placement issues to resolve'''

AI:determine if in Bjoern's use case ELISPOT plate is a reagent or an instrument - protocol application and instrument branches

Discussion

RS: fraction is an ouput of assay, I view as a biomaterial transformation. I agree with everything else on slide 5. My use case, outout from assays are dependent continuants e.g. glucose @ 10 pg per dcl. That phrase is an indep cont that inheres on the specimen that went in. Analyte 2 roles here, plays a role as a dependent continuant to represent something about the sample

JF:Trying to divide between quality/not. Conclusion describes a quality of the input material.

RS:Agree, quality is a quality of specimen not the analyte.

BS:Agree with RS, there's a simpler version. When you list objective to measure, you are not measurung glucose, you measure a quality of the specimen - glucose in this specimen. In all cases what you are measuring is a quality and then don't need a role. Is there a case where you don't measure a quality of a specimen.

AR:You want to say which quality, that's why we need analyte

BS:of course

AR:but which quality is what analyte, not just any quality.

BS:more than this. When define an assay, glucose is a not a role quality of a spec, so we use analyte.

AR: we need a name, or placeholder. So this is why we need analyte

JF: we need to know the objective of the assay - measure glucose - how you do it doesn't matter. we are using analyte as shortterm. I want a term that's broader than analyte

AR:option 1. Analyte role inheres in glucose of specimen. What if there is no glucose nothing to inhere in.

RS:plays a role in plan, even if null when measured. Can inhere in the plan.

AR:can only inhere in an independent continuant, plan is not.

BS:can use named parts of plan e.g. goals. Analyte would be name for the part of a plan. When we use analyte, we are talking about plans, and realize

RB:sometime you find something you weren't looking for and wanted to use analyte for that

BS: then you have revised your plan.

BP:if you don't detect glucose - that's a fail, but there is also assay fail

BS:we only do canonical experiment where they are completed. Cases where there is no 'thing' we don't deal with that case, except in plan. To get all investigation we describe all successful things. Failure is a meta thing

AR:question is not about whether we find glucose.

BS:if plan is to measure x, and there is no x, then plan should be that 'we are looking for x' is it there

RS:Ryan's def worked

AR:didn't place in ontology

RB: one proposal

analyte-role, the thing you found has role analyte analyte goal - in plan

Confusing to have analyte as a prefix for both

BP:only problem is that we need to inhere role in something. Inheres in the class glucose for me. As long as we don't specify that there is glucose in the sample.

JF:where do roles that are e.g. on processes go

AR:analyte in in the plan, and this is in your brain

BS:one part of plan is a part labelled for 'analyte'

AR:would need to be 'has_analyte_part'

BS:suggested some horrible concatenation. Issues with refering expressions which have no reference. There is a distinction between plans, measure x, and decide if x is there. Need to handle sep., no good solution in philosophy unless you do this. To say that even when no glucose, there is a glucose quality is bad. Everything would then have a glucose quality even when it is not there, and there are no negative qualities.

JF:what if assay is to measure a behaviour in a rat

BP:this is more complex. We do have protocols that measure non substances. Lets not go there yet.

PRS:how do you measure behaviour. E.g. exploratory behav. in rats, swimming behaviour

BP:we will be able to do that, I want to complete analyte.

AR:there are two analytes - role/goal. Is that OK

BP:No

RS:there is a goal to measure glucose.

JF:objective of plan is to measure glucose and that is a quality of specimen. Not heard anything different.

BS:I think when you say glucose, you mean amt of glucose in specimen where that specimen is blood

JF:if we believe amt glucose is a quality of assay, the specimen has glucose then you determine that. I have a glucometer, and it's function is to measure glucose. How do we define that use of glucose

BS:suggest when you measure glucose you measure a quality, expressed as e.g. a ratio. when measure a behavior is an independent cont aggregate, and determine a quality of that ratio of time spends smoking vs something else

BP:finish glucose first.

BS:think solution works for both.

BS:Objective is part of the plan.

RB:what's the parent for analyte-role - is it plan

BS:no it's a part of plan. All analyte role is part of some plan.

PRS; goal is a role played by a statement

BS:made before had info object, and needs to be changed

JF: would this term be in plan or DENRI.

BP:lets define the terms

Analyte Page [] - for definitions and conclusions about analyte

MC:objective is in role, then we would need to move that

PRS:should the objective of measuring be a process:

AR:no the process is measuring

PRS:all protocol have objective, and tie then via protocol application

AR:only issue is that may be measuring something not there

PRS:no intent was to measure, and solves that problem

BP:is an objective then of a protocol

HP:absence means undetectable in this case. It's shorthand.

AR:we need to make a logical definition to tie the concept of glucose

BS:we need an 'aboutness' thing and that's been added to analyte page

'''2. Richard Scheuerman ppt. Towards an OBI based data model'''

Similar issues to Jennifer on coverage of OBI, this is a good thing. Major focus can we use OBI as a data model for
 * BioHealthBase - pathogens, and primary expt data, virulence qualities, genome sequence.
 * IMMPORT - primary experiment data, clinical, proteomics, etc
 * ClinicalTranslationalScienceAward - clinical and translational investigations.

Want a common framework for all these databases using OBI

Use case. Mapped a publication into OBI, processes etc then mapped to a spreadsheet to determine all the protocol applications and assays. Composite things that correspond to multiple protocols. Want to describe these composite things somehow, and can't do in OBI.

Developing an OBI based data model, also want to use it for clinical trials e.g case report forms, inclusion criteria, assessment of subjects using assessment assays, composite score, medical history, physical exams.

Issues:
 * time and temperature representation
 * composite protocol application and outcomes, inclusion criteria = age.
 * Need for new dt terms. How age is an inclusion criteria - single conclusion span - start and end determined. Physical, or time dimensions.
 * Need for roles in processes
 * what is time
 * how do we represent context in obi (should we?)

BP:some of this belongs in the ontology some in the database implementation

AR:we have time, and we can say it and we can't reason over it

BS:can these protocols be in public catalogues? If so those protocols can be externally represented

AR:complex protocols are just protocols with parts this is not a big issue

AR:PATO has relational and monadic qualities.

RS:Chris Mungall never commented. There are relative ones to am accepted std e.g. units of measure and qualities of one independent quality to another. (we think this is what is meant in PATO).

BB:depends on what we want to capture in OBI. We don't need to capture what a cm is.

RS:we had a comparator concept, thing and thing we measure vs.

BS:most things in PATO need this, and hasn't been dealt with as an issue.

AR:context - one will make more progress by removing context and talk about components e.g. environment sites and locations (and their qualities) and time.

AR:some belong in EnvO, for site and the qualities of site

AI: Send Env related terms to BS for him to present in Manchester http://darwin.nerc-oxford.ac.uk/gc_wiki/index.php/EnvO_Project

BB:Like the practicality for application of OBI. Some projects have similar objectives, two are NCBO, OBD - OPen Biomedical Data repository - phenotypic descriptions using PATO. Already being discussed, re-use for clinical data. Suggest that RS shares experience with Suzi Lewis's group. Mark Musen's group, combine OWL, relational backend, swirl rules. Also FUGE OM implementations in Newcastle Allyson Lister's work. We grew out of FUGO, connection with the data model. How will the threads come back together? Also Healthcare life science group, looking at neuropathology going through the same process with mining a paper and represent on OBI. Suggest also look at that.

RS:we're already working with Mark Musen's gp. We tried to work with FUGE and they were not receptive. BUT we are developing FUGE for flow cytometry between me and Ryan. Preference not for FUGE but would rather use OBI. Not got a good interaction with FUGE so far, as it's too generic, and you need your own business rules.

BB:Agree with sentiment, but a layer might be needed to implement.

AI:Richard to communicate with Suzi Lewis and Chris Mungall on OBD future, overlap syngeries with OBI

AR:easier in OWL, separate the logic layer and embed in a database system. Think about doing logic first. We decide not to have this discussion until the end of the meeting.

'''3. Liju Fan. Relations Branch PPT https://wiki.cbil.upenn.edu/obiwiki/index.php/Image:Fan_RelationsBranchStatusReport_012808.ppt''' Chromium use case assay at the instance level relation. Previously added type level relations. AR:assume that all instance level, and class relations are built on these. LF:new role types proposed:
 * inheres_in (inverse has_function) -
 * has_role
 * has_quality
 * is_realization_of
 * utilizes (a question about child of these being present in the OBI file, this is not general and needs to be dicussed later)
 * hidden relations in text in parts of OBI
 * relations are database field style, represent in OBI are problematic. Want session to do that in this workshop.
 * minor classification errors need to discuss later
 * relations missing from BFO - added time

BP:reference on protocol applications focussed on text defs in protocol application, and these are not yet in OWL, so not really hidden, just not formalized.

JM:has_base should have been defined OK, we need to follow that up.

MC:is utilizes being added ?

AR:no, probably is participants, and there are two ways to say the same thing. this needs to be fixed.

BS:issues with instance vs class level relationships and that we need to be clear what is being used. Not only OWL artifacts, instances that go into relations part need to be comprehensible.

AR:better to have instance and quantification for understanding. only some of the inverse relns will work

BS:all sum assertions will be in the ontology and also one outside ontology. how make these work where they don't satisfy the all sum relations.

BB:feel that relations branch is conscience, they are the ontology police. has_function, has_role etc are these sub types of the RO - bearer of. If so why do we need these instead of RO bears?

AR:special cases. all say that when something has_role - the object has the the role. e.g. if has_function with a role then will generate an error - using bears from RO will not do that.

BB:ok, but this is a problem as OBO relations are being used elsewhere

AR:these have been sent to OBO relations ontology

BS:these likely will be added to RO and are useful

BB:OBO-BFO bridge file? Anything in RO2 will be addressed by Christian right?

LF:yes.

PRS: should we now stop adding more relations now and work with what we have

AR:we are stopping things that don't make sense, very few things are being added

BS:need a process to connect obi needs to RO to make sure that instance level relations are disseminated in OBI to stop people making new relations.

'''AI:Role Branch will develop a way to make it clear for relations using an annotation property (or similar) to make it clear which relations can be used at class and which can be used at instance level. This constraint is needed '''

'''AI:AR,BD will go to the relations ontology meeting in Denver to present our needs. '''

4. Bjoern Peters: Protocol Applications Branch Use Cases and IEDB

Peptide seq of epitope does not equal influenza, needs a translation
 * Common questions: how many influenza T cell epitpoes are there - these are bio, not about the analytes. Also inexact of what they mean, this is a general question frame in OBI: count set of objects derived from influenza for which one or more prot applications measured a positive T cell response to the peptide. Issues that to count in OWL need disjointness, and doesn't scale.
 * In processes have created planned processes and other things. There is always an execution that can refer to. BS thinks planned process is not a special kind of process, suggests 'process that is a realization of x plan'. Some protocols induce natural processes e.g. methylation and need a way to connect natural processes.
 * lots of things ended up with protocolapplications e.g. purchase of an antibody. Doesn't belong to ProtocolApplication. May need to go up one level represent other processes in a different way. Planned processes in OBI scope for e.g.
 * broke protocols into sub processes, also a role recipient, something that has things happen to. These should be enumerated. Need to make sure that inputs and outputs are connected, can't do in OWL. (use case for SWIRL)
 * inputs and outputs are things necessary to realize the protocol. hard to say when the output exists e.g. when is a cake produced in a baking protocol.
 * could agree that for something need to be an output then something needs to have changed (but am I a new instance if my blood pressure changes)
 * have text definitions for many things.
 * would lie to decide where to stop e.g. administer substance in vivo - not all possible children. define precisely by using inputs and outputs, not in instantiating all possible children.
 * discussion on whether a sample needs to be representative of the whole, and cases where it's just partitioning where criteria are used to make a separation. RS:suggest sample is an output of a process and specimen could be an input to an assay. Could have a generic representation for all specimens derived from organisms.
 * Need a way to connect inputs and outputs - maybe with SWIRL rules
 * Need controls, time intervals, multiple repeats
 * To answer use case above - get all assays that have as output data 'on the T-cell response to an object' the final part of this is not clear yet.

BS:planned process is a realization of a plan, should be able to infer planned process as an is-a to avoid mult. inheritance. Planned process then should be an inferred class.

BS: all investigations have some result, data, report etc. Output also biomaterials some times, sometimes both. Suggest don't merge the two outputs.

BS:Assays are determined as output data, mouse is not an output from this assay

BP:No it's an output of a protocol application, data comes from assay

BS:want to be clear in the use case assay, output data.

AR:distinctness is a problem, as when need to count need at different levels, and traverse across multiple levels for objects at a consistent type. Users might not know levels, and this is a problem.

BP:we do define distinct locally, if something is different in database fields=distinct' can be different for different types. Get questions that cannot answer, need to offer some sort of answer and that's an issue

AI:determine if in Bjoern's use case of an ELSIPOT assay an ELISPOT plate is a reagent or an instrument.

5. Philippe Rocca-Serra Plan Branch


 * Determine purpose of a study, treatments, populations, temporal component, replications, nature of measurements, applications/incompatibilities between protocol applications, reagents etc
 * Plan, protocol, algorithm, still defining, limit on what is maths and comp sci. deciding where to draw the line and try to decide what class constraints. Do plans need goals? What process is being executed.
 * If have to declare protocol as a plan, by realizing a DNA extraction process, how much do I need to represent in plan, how much in protocol application? Branch divisions? Some things could be merged. Need a way to automate this process of development.

BS:P/A should be a single tree (no multiple inheritance) but need assays, inputs, outputs, plans.

BS:should just do protocol, and infer protocol application, good way to validate inferred classes.

BP:we had this discussion before, protocol, protocol application are interdependent and term submission showed that. Wanted to put it all together. Put in protocol application, easier to handle in terms of what was done rather than what was planned.

RS: Should be in plan as can describe without having to realize it.

BS:we don't know what protocols are until realized as protocols are abstract. The best source is therefore protocol applications.

AR:since we connect, hook up, and things that are hypotheticals are split, and merge by making a connection not infer

HP:bad, need a single way to do this, and remove redundancy so either one or the other and not both

BP:Possible that won't be symmetrical anyway.

RS:can we define the objective from the protocol application?

PRS: all protocols have an objective

BB:if have multiple protocol applications that have the same objective then can get into problems.

BP:different levels of preciseness, so can be as granular as you need. Class of protocolapplications all of which have a measure of glucose as an output are the inverse of the protocols that have goal - determine glucose. Easier to determine at the process level therefore as can be precise and granular.

PRS:same issue in data transformation and algorithm, see no reason why we can't deal with it in this direction

BS:some things are avoid doing - don't communicate protocol application.

HP:these are negation statements they can be transformed into positive statements.

AR:if you have both ways to represent the protocol/protocol applications, need to look in both cases and deal with on the implementation side, by making a mapping table

HP:that's solving on the implementation side a bad design decision in the ontology

BP:We want to remove redundancy. No one wanted to have duplicates.

HP:this needs to be done with testing from both directions, we are discussing what the tests should be not where to start from, need to start from both places.

BS:lots of clinical trials things have things in plan that can't be in protocol application

PRS:should we have input from the clinical trial field

 AI:BP and AR will try and develop ProtocolApplication from Plan

AI:Protocol and ProtocolApplication branches will be merged by evaluating some use cases

AI:Produce an example a ProtocolApplication from ArrayExpress Protocol- HP

AI:RS will provide a clinical trial example of protocol

AI:How do we embed OBI in different systems - AR wants as focus of a session this week



6. Susanna Sansone - Biomaterial Use Case https://wiki.cbil.upenn.edu/obiwiki/index.php/Image:Sansone_OBIBiomaterialBranchUCs.ppt


 * Multi-omics technologies, sample details etc
 * Went through the use cases from the wiki, summarize most want to describe the subject of the study, innate, also during and after the treatment, historical characteristics e.g. medical history
 * Definition of biomaterial branch, objects that have a role, sample or reagent roles - has been redefined recently.
 * Issues about how to deal with defined classes
 * Dealing with mixtures sum of parts, how describe if add something RS: a new mixture
 * Track biological and anatomical origin, derived from mesoderm, link to EnvO
 * result_of centrifugation - AR: can deal with as from ProtocolApplication. Need to remove the duplication.

BS:the way that BP presented his queries was useful for marketing purposes, and this documentation is key, liked the way that BP presented this

JM:These could be considered as competency questions

AR:competency questions are part of use cases no?

JM:in use case is a scenario and an example, comp questions is what is returned when I ask the ontology a question

RS:the use cases are 3 kinds of derived from, 1. where no net increase/decrease 2. partioning. 3. merging. Do we want to describe all three independently rather than using derives_from

PRS:we can described char of a sample e.g. lake sample, not affect the bugs in it, originates from gives more power.

BP:these things that are in biomaterial branch have all gone through a process e.g. env sample collection, this is an object that is an output from a process. This should give us the connection to env etc.

7. Alan Ruttenberg, DENRI ppt http://ashby.csail.mit.edu/presentations/DenrieReport.pdf 


 * worked from the google docs, now in OWL
 * made recent fixes, added arist. definitions and find missing relations 'encodes' and 'about' - about is not in wordnet in the way that we use it
 * think objective goes here, movd shared plan into realizable entity
 * algorithms - what's the difference between this and a plan, maybe alg is just a shared plan?
 * file formats are realizable as file exists
 * programming language - patterns of words, syntactic plan, pl are ways of expressing algorithms, need to express that and can't
 * digital entity - bits - relationship to other information entities is that digital entity encodes them. Not realizable - could be a sibling of non realizable information entity
 * source code module - see discussion below
 * datum - qualities and datums are converging, some datums are proxy for qualities. Should we repeat, if not where does it live?

'''AI:BS will provide a better name for MeasurementUnit as Units pre-exist in the world outside data and so the name is bad. '''

BS:there are information content entity outside OBI

AR:should be in BFO, so should narrative object and symbolic

BS:even if we did that, I am trying to keep BFO small

AR:you are missing abstract entities, acc to Stanford Philosophical Enc.

BS:BFO is a small ontology for research, subset of SUMO want to keep it small. Unit should not be in OBI

AR:the measurement x cm should be in OBI. We could call unit unitname instead but it does belong here

BS:already have a document ontology

AR:could make this information entity a new obo foundry ontology

PRS:how is a source code module and related bits helping the biologist?

BB:we need to know about NIH source code, and make it searchable and useable

BP:we are getting away from the point

JM:is there a useage for these?

AR:no we were asked for these

BP:I think some came from test cases and are artefacts

AR:I think we need all these

BP:I don't think we need gene list

AR:Chris S wants this somehow.

HP:can we have an accession number list

RS:it's a special case of matrix - and we need matrix

BS:we should think prospectively about things like gene list

AI:Alan will generate a list of problems and solutions with DENRI for discussion on a conference call

8. Melanie Courtot, Instrumentation


 * Labelling flow cytometry protocol extract instrument represent using OBI
 * Do we need pipettes, ice bucket, small lab ware
 * does size matter
 * we have sample tube, should this be defined as holding a sample
 * falcon is needed here, where should we stop with tubes - decision with tube and properties that can be used if needed e.g.
 * question is it enough for instrument manufacturer, model, serial, (or other unique id and serial is a subtype) JW:physical location? AR: that's handled separately, as anything can have location.

AR:when do we give an id for something, sample tube, has_quality volume. You only need to id when

BP:sample tube shouldn't be asserted should be defined, otherwise need to embed all possible uses

RS:is this flat? looks to be, tubes, beakers, etc do these need to be added?

MC:we have labware and instrument and it's not clear

RS:we can determine in terms of function

AR:these can be anon classes - may need an id, even for inferred classes

RB:we don't put it in unless someone asked for it

MC:this would go under labware - but no good definition

HP:the ice bucket cover should not be modelled as this would be represented as text for humans and if it's needed by the community it will be added, no one has asked for this and aiming at completeness will not be productive

BS:needed for the Robot scientist

BS:functions for artifacts - function is what it was built for, roles are when put to other uses.

BP:then you need 2 things, role and function for all artifacts

AI:BS will provide an example of function/role that doesn't involve doubling up with role and functions 

BP:Instrument Manufacturer - where does this live? It's an instance of Organization

AI:Organization needs a place in the ontology, not yet determined

RS:when you use instrument in an assay - there are config parameters.

MC:the instrument is not enough - filter etc. Is it the same instrument. If we want to do that we need to model all the parts of the instrument to do that. Don't expect a user to annotate these. Template for these?

AR:we can define a class expression for the specific type.

HP:is the config a plan? Has anyone asked for it?

MC:yes, so we need it

RB:how do we implement - and OWL file from vendors

BP:then we need to have laser for e.g. so that the vendors can have the vocab, don't want it in the same file

HP:practically though we need these and the people in this room will deal with this, whatever we do with the OWL file.

BP:we wanted to keep the h'archy outside the OWL file

RB:so organizations - e.g. Affymetrix, and there would be an org.owl file and this could be opulated by us

RS:hardware is config and then changes that are made in a run. e.g. parameters - which are qualities.

HP:I think we need to only deal with the hardware configuration

AR:do we use classes or instances?

RS:want to group on e.g. common laser

AR:can do that at query time

HP:if you group on parts then there will be mult inheritance so can't do that

RB:would this be in the OBI SVN?

General yes it's OK in OBI SVN.

AI:JW will provide a link for PDBs effort to do this for beamlines

9. James Malone - Flow Cytometry Data Transformation Use Case


 * plots - histogram, dot plot, density plot - visualization, need to be added and probably others too.
 * minor granularity issue with a specific type of test, new terms added
 * see table that shows mapping
 * SW ontology link needed to deal with instances of e.g. sw names

'''AI:Make the connection between the visulation here between table and graph that comes out of this process currently that is in the DENRI branch : DT and DENRI branches. Could be done with defined classes'''

RB:where would sw type live?

AR:information ontology, beside algorithm, they can be realized

BB:need packages vs libraries as well.

BS:process vs roles for dt. confusing instances with types. instance of the dt process is only ever used to normalize and is therefore an instance.

AR:if a process is the same but used for different processes - then you get 2 parents

BS:I agree that I have not made my case, but this is an instance level thing

BP:did you consider the objective of the protocol application - that was the other solution.

HP:this could still happen, and this was considered.

AR:I am more worried about users understanding this.

AI: dt branch will consider using objective, not role - might make things easier

Role is confusing and all the discussions will be distilled by BS and AR for clarity to come up with a less confusing definition

JM:SW ontology is nascent and we have a contingency

BB:has a Birnlex ontology that might also help with SW.

Discussion on how to prioritize for the workshop
RB:What do we need to publish?

RS:What's missing? Investigation etc

AR:Documentation?

JM:Use cases, and annotation examples?

BS: Grant question, packaged for grants?

HP:what is a 1.0 release?

RB:W3C Model for release: 
 * we are at a working draft now
 * next Public working draft, - iterate
 * next candidate recommendation, iterate, 12 months, publish before OBI is accepted ideally
 * next proposed recommendation, needs to be stable.
 * Need a versioning policy and an obsoletion policy

 We all agree that we need a process

BP:how much do we need to ask the community to comment on

SS:before we go to the community, we will have another workshop in the summer. Suggest internal evaluation in July.

HP:Suggest limit attendance to people who have tested it so that we make progress

BB:worked examples, and also queries and building some data sets

AR:Part of OBI foundry, they are also trying to develop a process, which is more limiting on release. They do rolling releases. Maybe we can be thinking about this. Needs to go on the obi-discuss list.

BB:As soon as we put things out we need a versioning policy and a way to channel feedback - report generation etc. This was the need for curation status, not just human readable.

AR:Go don't do that

HP:they do, there is always an alert and a consultation period for GO changes esp. obsoleting.

BP:they keep the 3 top levels fixed in GO, but there will be changes esp. additive.

RB:I'd like to know what the release policy is, we have made 2 releases and I think now publication might be a good example

BS:we would need commitment from large data holders to use ontology and evaluation

RB:not for publication, but for OBO peer review

AR:let's not invent a new process for this, there is a process document, see what's missing

RS:we need to see the branch ppt to see if we are ready.

SS:This will also influence what we can do at the next workshop

AR:Has everyone read OBI, all of it?

BS:I have read it, maybe 6 months ago - all has changed a lot

BP:each branch should be happy all other branches look at it, read and understood

MC/JM we had an interbranch review, we went through all terms that we thought were complete. We all reviewed and we know is incomplete, but what is there is useful and has high level complete.

MC:CS will look at the branch and will comment, is aware of what we are supposed to do

RB:3 things, branch review, inter branch review, obi release, publication - we need to see what the requirements are for all of these and they are not the same.

HP:inter branch review needs to be done as evaluation

AR:I'd like to set a std for our advisors, they need to do things.

RB:did we ask them to specific questionbs

PRS: do you have a large set of data to annotate and can you prove that it's useful. We identified some process to review, plan to define and objective etc. We gave PA and DT and these are similar and these need to be aligned to work in the same way. If we don't align then it will be a waste of effort.

1. Plan/ProtocolApplication needs to be done this week

2. Identify main cross branch relations and work on that.