Minutes summerf2f 2009

=Minute from Face to Face meeting summer 2009 @ EBI=

Tuesday AM 9th June
Caveat: Minutes may occasionally be inaccurate during capture

Attending: JM, BP, NM, MC, FG, DD, PL, LS

Collaboration with Envo

JM: We have some sample information in AE which relates to Envo scope

NM: Env are holes or sites with qualities that enable the area or site to support life is current thinking. Env sampling would be taking some quantity from this area or site. I assume you would want to represent env sample from the point of a sample going through an assay.

MC: If you have bacteria in an organism then the organism is an environment, so is it a role?

BP: So we don't have environment at moment, we will have sample collected from site in OBI. There is a class site in BFO so we should define things in OBI under this. We don't worry about whether it is an environment this becomes clear when it is involved in a sample collection process.

LS: It would be good from our point of view to have a higher level class for environment

FG: This shouldn't matter if we use BP proposed approach. So any env should be under site and then in future we can adjust this when envo has definition.

MC: How do we deal with having things under material that we want to say are environments?

BP: If we say mucous membrane then we talk about this class defining a site.

FG: So they are completely different

BP: So inside of my body is a site and this is

FG: material has_part site

BP: Barry has said this is completely clear under BFO so we should follow this

MC: What do we (OBI) need to do for Envo

FG: We submit anything we consider site to the Envo tracker (see OBO foundry sourceforge site)

JM: I can generate list of sites from AE and pass that on to NM if useful

NM: Yes that would be useful. We are doing rocks and sludge, land cover such as agriculture and urban areas from satellite views of world, anatomical regions, holes in which you would find bacteria growing

NM gives demon of wonton ontology browser (URL is not shareable for this as currently private)

NM: Env and habitats and how you instantiate relations in OBI. Envo is trying to remove all reference ot habitats because we want to only reference site.

FG: Defining environments are outside scope of OBI (agreement from room)

Assessing use cases in OBI paper

BP: Discussion about replacing one of use cases from OBI Journal paper with the evidence code use case, discussed during the OBO Foundry workshop

FG: I would suggest adding a further use case.

LS: Should we just add this to existing use cases rather than create a new one?

BP: So the EC ontology exists and the idea is to replace all of the EC IDs with an OBI identifier. This is a use case for OBI.

FG: Competency question is can we replace EC IDs with OBI IDs, and second can we serve same queries. I would like to go through existing use cases first and then do other use cases.

MC: I would like to address EC one

FG: Only if we address other use cases first, this is priority (the journal use cases).

JM: How are we going to do this, I woudl like a process for this session

FG: Can we take each use case and represent it in actual OWL.

GenePattern use case

BP: Visualization is a process so it should be under planned process, presently it is obsoleted

JM: It should be put back in if useful

FG: We should add OBI IDs next to each name in the manuscript

BP: I think instead we should not do this as it is annoying to read, instead we add this in suppl material

JM: I agree, I think we need to make it easy to read first

FG: clustering process needs to be renamed in manuscript to class discovery data transformation (OBI_0200175), normalization process needs to be renamed in manuscript to normalization data transformation (OBI_0200169)

AI: Rename the above in the figure in the manuscript

Discussion about how we link the data item that is output from DT to the report graph that visualizes it.

JM: The part that is missing is the link between the data and report graph

Summary of discussion: We will make the connection by saying a report graph is_about some data item and then add the specific details for each class. So for clustering data the link will be, e.g. dendrogram is_about some tree-like data set (of which clustered data set will be a subclass) or .... etc. This will make the connection between our data sets, the DTs and the report graphs that render the data. This will enable us to represent the GenePattern use case, e.g. tell me all report graphs that can visualise clustered data sets (i.e. output of clustering data transformations).

AI: JM + MC: to add the is_about relations between data sets and report graphs, priority is to get clustering example from GenePattern completed. MC will help from IAO side as most of these axioms will be on IAO classes.

Immune Epitope DB Use Case

BP: In this use case there is an issue which is that I created an intermediate ontology for our purposes because we needed a lot of assays in quickly and OBI is not stable and also takes time to add classes to. So in many ways this does is not the best demonstration of OBI because it advocates a bespoke ontology.

FG: Can't you import this bespoke ontology back into OBI

BP: we have done some work on this

FG: Is epitope within scope of OBI, i would be inclined to say yes

BP: It is not clear to me whether it is or not. What we have done with the ontology is build it around OBI but given them ONTIE IDs and keep them apart for now. (BP shows ontology) There are def some things that belong in OBI ELISA assy for example, but we have very specific assays which are granular because we use them in work, not sure if OBI needs these

JM: Questions are, is in scope, who would be keeper for this if it wasn't and would it become sep ontology or sep file imported in?

BP: I would be happy to give these OBI IDs but equally fine with them being in sep file.

FG: This is a good use of how to build an application focused ontology under OBI

JM: I agree, similar to way we are trying to build for AE

AI: BP decide whether these should be added to OBI and come up with list of requirements before this can happen (i.e. adding things to OBI)

AI: MC + AR to check if this can be accommodated once above is done

FG: The slant needs to be changed in journal so that it shows that OBI can be extended for application ontology successfully

JM: Agreed this will become AI for BP (see below)

AI: BP The slant needs to be changed in journal so that it shows that OBI can be extended for application ontology successfully

FG: Presently, as OBI stands, we can't do all of this use case

BP: No, it still requires a few things, but all doable before journal is out

BP: I can add more specific info as well, like these are classews you get back from this query too, but i'm not sure it is interesting and we already have the GenePattern querying use case

FG: I think this is still useful as it's from a diff perspective, using application specific external ontology

MC: Do we explain why this useful, what advantages of using this approach are

BP: Yes, we explain in journal paper where demonstrate how this could tie in different resources using a single query

MC: But you are going to query using SPARQL oui? Something about ducks and genes querying GO and pubmed at same time, so why is this good to have in OBI instead of 3 different resources?

BP: This is fine as long it is all in based on OBI structure

MC: Paté

BP: Second part for IEDB was to use ontology to validate using consistency checks

FG: Add this to discussion perhaps to save space, it is useful to say though

JM: Agreed with above

BP: OK will move it

MC: Shall we put in examples, for example we can check for x, y or seagull?

BP: Yes so we do this

Use case 3: JF representing a protocol

BP: So the us case discusses a protocol and how we can represent this. We don't have in OBI, ethics committee, enroll balanced cohorts, placebo is still problematic, none of this can be done presently in OBI.

FG: Maybe we should stop at first part or not include it at all given we can't do it

JM: So pros and cons as I se it, pro is it does talk about a study and the parts therein which is different and is critical to OBI's scope. Con, I don't see this is an actual use case, what is USAGE here, is it annotation, template query, etc, also does this overlap with early methods section describing how we represent bits of studies?

FG: I think we don't cover most of this in ontology and it doesn't show use as in use case so we should consider dropping it

JM: Can we make modifications rather than drop?

MC: In what way is it not use case?

JM: Use cases have an actor that show how the ontology will be used in a given scenario, this is largely a coverage example at present, but this can be modified to become a use case.

FG: Issues still remains we don't cover it

PL: This is a use case which is describing utility

FG: yes but it needs updating as it's not correct at present. I'm not saying delete it but we should reformulate it.

JM: I think we need a third use case and if we even thought about replacing it we would still be faced with issue of requireing some work for a new use case so doesn't remove the problem. i would prefer to reformulate this and fix OBI

FG: so we need to go back to JF with the following four questions what is it the experiment you are trying to do? how do you model this? how are you using OBI? what are queries?

AI: BP: to ask JF the following four questions what is it the experiment you are trying to do? how do you model this? how are you using OBI? what are queries? This will enable use case to be completed

Tuesday PM 9th June
Looking at first section of manuscript

BP: We need a class that is high level, such as our anatomical entity, that we can import from external ontology. We need to be able to import a species neutral

AI: Biomaterial to contact Chris about species neutral anatomy (UBERON) as we need to import this into OBI

FG: chemical entities in solution why is this solution and not liquid?

MC: shall this be a defined class?

BP: No. more generally we are trying to model mixtures but we have limited to this for now.

BP: In an ideal world we would have a term 'mixture'. This is where we need to demonstrate the has_grain relation

MC: If we add this into the journal now we will take up a lot of space

FG: so chemical mixture is defined as mixture of one or more homogenous chemicals. So I think what chemical entities in solution describe is chemicals in a solution which is not a mixture. So are you saying you want mixture class?

BP: Yes I feel we should have this, since we have mixed materials generally not accounted for in OBI. At same time we have some issues saying, for example, why a cell is not one.

ET: yes

BP: Process material we have in, specimen and blood sample etc needs to be removed from manuscript and realigned with what we have in OBI. Postpone for 30mins for Phil to get back as we would like an outside review. We had discussion in previous workshop to skip specimen and sample but we can't avoid this any more.

JM: Agree we need to address these terms, they belong in OBI if anywhere.

MC: grml

FG: Protein is not a type of peptide

BP: This is not an OBI issue, it is defined as such in external ontologies, chebi and pro. We don't have recombinant expression system in OBI can we add it

FG: We need to say proteins can be synthesised or whatever

BP: This is not the same, i think natural language works here.

FG: We don't have agent in OBI

BP: No but this is in preparation for objective specification-planned process part, but we perhaps don't have space to add all the details in manuscript

FG: Agreed

MC: I agree with Frank, as always

FG: I disagree

BP: has_specified_output and has_specified_output_data - the issue with these two relations is that they are correct if we wish to assign 'output' role to material entity and data because we don't know if we can assign the role of input and output to data currently. If we can have input and output role for data then we don't need these two relations.

FG: We should just have input and output relations nothing more specific than that.

MC: I think everyone agree we need roles for data so we should push to barry/bfo, the problem is with the specification of the relation they need to have role output or input

LS: That needs to be fixed

JM: I don't see issue, can someone give a concrete example

MC: Issue is when we have role input/output it means that your role is realized within a process so it is a 'special' status

BP: When you have an experiment and have instrument, cell etc, then they are all participants in the process. We should remove the roles part of the definitions for the input and output relations and we can remove the information related input and output relations: has_specified_output_information, has_specified_input_information and instead has_specified_output and has_specified_input should be used.

AI: LS and relations branch: remove the roles part of the definitions for the input and output relations and we can remove the information related input and output relations: has_specified_output_information, has_specified_input_information and instead has_specified_output and has_specified_input should be used.

Phil reenters room so we go back to sample#

PL: Agreed with the changes that output range should become a GDC or material entity. (some rant about GDCs and why boundaries are Indep Cont)

Sample/Specimen

BP: One decision we made in biomaterial call is to try and model something simple first rather than try and do all possible uses of the words. This example was blood sample. This became hard because we ended up referencing other terms that are also contentious

frank takes the floor to demonstrate his thinking

PL: There are two uses we are debating here, first is i took some blood from someone. Second, this a randomly selected subset of a population which is representative of larger population

BP: I can have a population of 1, if i look at a whole body for example.

FG: Yes, my definition (Referring to my google doc) is that a sample is a role that is borne by a material entity that is, the part (subset) of a population (whole or class) which is assumed to be representative of the population.

PL: Data sets would not work here since it involves both roles and also Information which is not a material entity

JM: So we need two sample, one for material and one for statistics

BP: Yes

Agreed definition: a material sample is a role that is borne by a specimen that is, the part (subset) of a population (whole or class) which is assumed to be representative of the population.

FG: We can also add the defined class for material sample which is material bearing role of sample

BP: This is fine but it is not main controvesial part, if I do sample extraction is it material separation or sample.

PL: If you think of material sample as a role, I might want to collect butterflies but that does not mean they are samples they are just a collection or specimen.

BP: Question is when does it become a sample and not just a specimen?

A specimen role is a role borne by a material entity that is collected for potential use during an investigation

Furthermore; material sample is_a specimen (but not other way around)

specimen examples: Biobanking, e.g. blood taken and stored in a freezer for potential further investigations

material samples: blood drawn from a patient to monitor glucose level,

JM: There is also issue that role is only borne during a process so once process ends it's not a sample

FG: Well problem is in BFO these may not work since the role is dropped at end of the process

PL: Stored is key here, if we say it is stored until it thrown away it still has role during this storage process

specimen will lose role when it is no longer stored sample loses it representativeness of whole

BP: Role is a social construct so as long as someone has the intent to derive something from it then it still has the role

PL: it will fall down for measurements of height since heights are quality

specimen gathering process is_a planned process with the objective to obtain and store a material entity for potential use as an input during an investigation

material sampling process is_a planned process with the objective to obtain and store a specimen that is representative of an input material entity (we will have to be able to answer for a given material sample what material it is representative of; that may require adding a role for the 'sampled material'. Right now we will use the input relation)

JM: We need to say that when it is no longer stored it is not a specimen not just when it is destroyed since this is our definition

FG: I don't really care about this all that much, even though you make a very good point James and are correct.

JM: Why thank you Frank.

BP: bronchial alveolar lavage is this part of lung or is it extracellular

PL: it is extracellular it is secreted fluid, the lavage step suggests that something extra has been added to the process though

PL: I do not understand the derives_from relation, this not a definition by BFO standards. derivation by this is only frmo biological materials as mentions biologically significant.

FG: Can we use has_grain?

BP: I don't see how we would use has_grain, how would we?

FG: Beer

PL: derives_from suggests that original does not exist from the new part so it would not work for lavage. What you are trying to say is has_part some stuff that derives from some anatomical entity

BP: we need to be able to say that there was a link, that it_was_part_of but original still exists, similar to a sample process, but essentialy it is what did you take the specimen from.

PL: If you have the notion of lavage, the process to lavage, then you already have the notion of what it was

PL: we need to distinguish between process and result of the process, rest seems reasonable

BP: So we need to have a lavagae process we can add this now

MC and FG shake head, not at work but reading Twitter

LS: What language is word lavage from?

MC: français

LS: Do you know OBI in japanese means sash

BP: yes logo is meant to reflect that

BP: Can we call things urine specimen and blood specimen

JM: Yes

Looking at agenda for tomorrow

BP: We could consider, tomorrow, having time where 2 people leave room and edit the owl files while rest continue on with other things to speed things up, e.g. release process is not of interest to all, so that should split off with MC & someone else

BP: Also add reasoning to the agenda for Alan as it has slowed again

JM: Would be useful if Alan could give us some tips tomorrow on helping speed up reasoning as we edit

Tuesday Action Items
AI: JM to generate list of sample envs from AE and submit to Envo

AI: JM to edit the figure for the GenePattern manuscript to reflect the wording in OBI, e.g. normalization process in figure should be normalization data transformation

AI: JM + MC: to add the is_about relations between data sets and report graphs, priority is to get clustering example from GenePattern completed. MC will help from IAO side as most of these axioms will be on IAO classes.

AI: BP decide whether these should be added to OBI and come up with list of requirements before this can happen (i.e. adding things to OBI)

AI: MC + AR to check if this can be accommodated once above is done

AI: BP The slant needs to be changed in journal so that it shows that OBI can be extended for application ontology successfully

AI: BP: to ask JF the following four questions what is it the experiment you are trying to do? how do you model this? how are you using OBI? what are queries? This will enable use case to be completed

AI: Biomterial branch: to contact Chris about species neutral anatomy (UBERON) as we need to import this into OBI

AI: LS and relations branch: remove the roles part of the definitions for the input and output relations and we can remove the information related input and output relations: has_specified_output_information, has_specified_input_information and instead has_specified_output and has_specified_input should be used.

Wednesday AM 10th June
Attending: JM, BP, MC, FG, DD, PL, LS, AR, JZ, SAS, PRS, RS

BP gives overview from Tuesday OBI meeting

AR gives overview from industry meeting

Discussion on sample/specimen

AR: we need to add realises to the definition as this is a role

Bjoern edits to add realises

BS: We don't need storing process in the definition

BP: It is needed as otherwise the role is lost once it is collected and this is not what we desire, it needs to a be a specimen long after it is collected

RS: What is main distinction here between sample and specimen

FG: The distinction is that sample is representative of a whole whereas specimen is not, e.g. blood sample is representative of the whole blood, whereas you cut an arm off and use it then it it a specimen

AI: AR to think about the definition of sample since he had a couple of minor issues

BS: Earlier you said gathering and storing - do you mean gathering is one process and storing is another, or do you mean gathering and storing as one single process? This needs to be

AI: BS to tinker with wording a little of specimen and sample

Continuing manuscript work

BP: Are we saying we have a relation with IAo which is that we are also building IAO within OBI or that IAO is sep effort

AR: I think we should say that IAO was spun off from IAO as a result of the process of OBI

BP: rename information content entity in IAO to Information artifact

AI: MC rename information content entity in IAO to Information artifact

Discussion on parameter

AR: So there is a question whether an indep and dep variable should be in OBI

BS: They should be in OBI they don't exist outside biological investigations

AR: Indep and dep variables are clear, parameters are not clear. This requires discussion. Best way would be for JM and I to

recapitulate our views

BS wants to redfine independent variable (focusing on definition on board)

Lots of continuous discussion by everyone at once so not able to capture

Wednesday PM 10th June
example for parameter

for planned process k-means has_objective (for example) class prediction objection:

and this objective specificatoin has part_of some input parameter and some input dataset and some output

AI: JM to do parameter mark up from above

Data types

BP and JM: Discussion on data types, we would need ranges represetned by things such as 'string', 'floating point' and so on

JM: I already submitted some of these to IAO list

BS: integer and string are not data types, they are numbers or letters. Data types are social artifacts

AI: AR, following completion of JM parameter, AR will take this to IAO

BFO Discussion

BS: The common upper ontology which was to go into ISO is on hold for now. BFO 2.0 will be released soon and will incorporate RO 2.0 into the release. There will be several version of BFO 2.0, e.g. common logic version, owl version.

PL: Which was one is the 'correct version'?

BS: Common Logic version is the canonical version of BFO

AR: We have to live with the differences as a consequence of language differences.

AR: Alan does not want object, fiat object part and object aggregate in BFO

BS: There are many people in community use this so it is staying. Generically dependent continuant is being added to BFO 2.0 as previously agreed.

Chris Mungall: What is currently under disposition

BS: Nothing presently

BS: Function and capability will be added under disposition

AR: Are there any objections to the realizable entity changes?

BP: Given the approach we are going to take by putting everything just under realizable entity and then infer the different subtypes, this is fine for OBI I believe.

BS: We wish to remove processual context, this will become deprecated.

PL: What is relation between spatial region and spatiotemporal region

BS: That is a projection relation, so that you can prject a spatial region onto a temporal region. I have a paper on this very topic, unfortunately there are too many relations in that paper and that will potentially introduce problems. Do we have relation of process to temoral region

CM: duration of is in

BS: This is somewhat ambiguous, we need to address this

PL: What is an instance of a zero-dimensional region?

BS: Take a map of the heavens and pick a point in that map that is vacuumous and that is a zero-dimensional region

PRS: Surely there is a relation between two-dimensional region and site for example

BS: Yes and that is something that I need to address

BP: do you envisage any OBO foundry ontology that will use spatial region or temporal region

BS: I have thought of removing that.

AR: are subclasses under material entity needed?

BP: What is the need for spatial / temporal regions? If they are included, they seem to be missing combinations (scattered ...)

BS: length is only a property of spatial regions. There is an is_projected relation, which links objects etc. and the regions,

PL: Example: length of longest entity in the univers + 1 meter

PL: how do we express time duration for a process

BS: duration of time interval into which the process is projected

PL: How do you then describe velocity, accelaration etc?

BS: Will follow up.

Discussion on IAO

There are some apparent issues with slow progress of IAO. Alan has declared that he is the branch leader in effect and that there will now be calls at which term submission will be resolved to enable faster resolution of classes

AI: AR to organise calls for IAO

Change of URL in OBO Foundry

AI: AR + MC: MC will send email out to OBO-discuss to let everyone know of URL change and AR will do change

AI: AR measured data item - alan will organise a call to discuss the conflicts between this class, data and so on also add to tracker to IAO

back to journal paper

Frank closes "Rugby Union Heroes" website and back to Protege

Looking at the figure in the journal paper of the analyte mouse assay and checking for coverage in OBI

AI: MC will MIREOT mouse.

AI: MC or AR Author role is to be moved to IAO

AI: JF and Role branch to coordinate and improve/remove study specimen, study participant and anything else related in light of the new definitions for specimen and sample.

AI: Biomaterial branch to add class 'test tube'

Wednesday Action Items
AI: AR to think about the definition of sample since he had a couple of minor issues

AI: BS to tinker with wording a little of specimen and sample

AI: JM to do parameter mark up from above

AI: AR, following completion of JM parameter, AR will take this to IAO

AI: AR to organise calls for IAO

AI: AR + MC: MC will send email out to OBO-discuss to let everyone know of URL change and AR will do change

AI: AR measured data item - alan will organise a call to discuss the conflicts between this class, data and so on also add to tracker to IAO

AI: MC will MIREOT mouse

AI: MC or AR Author role is to be moved to IAO

AI: JF and Role branch to coordinate and improve/remove study specimen, study participant and anything else related in light of the new definitions for specimen and sample.

AI: Biomaterial branch to add class 'test tube'

Thursday 11th June
Attending: BP, MC, FG, DD, PL, LS, AR, JZ, SAS, PRS, RS

OBI review: Chris Mungall's review point to huge discrepancy between the logical and textual definitions.

We need to address style inconsistency in the definitions. The issue is how to make the OWL definitions that can be easly stated in natural language.

Logical vs textual definition: the goal of the english defintion is to convery to a reader what the term denotes, and to reflect the logical definition as much as sensible. The goal of the logical definition i to enable data integration. The 2 will not necessarly mach but should never contradict.

Top level defintions have been defined, we only managed to do 'chemical solution', 'organization', 'processed material', 'investigation', 'documenting', 'assay', 'material separation'.

MO and OBI integration: relevant/suitable terms will be moved to OBI; then when a view of OBI is created for the MGED community, then this file will also include additional modules (MO specific).

Actions

Alan: will edit the OBI definition in english

Richard (Foundry Coord): discuss new rules for all ontologies, where definitions are preceded with '=def' for any necessary + sufficient definition

Alan: bring chemical solution discussion to ChEBI

Dirk: to reorganise the wiki pages

Jie: will work with MGED reps in OBI to sort MO terms for inlcusion to OBI