microarray-ontol-digest Monday, September 16 2002 Volume 01 : Number 029 ---------------------------------------------------------------------- Date: 05 Sep 2002 13:31:02 -0600 From: jason@openinformatics.com (Jason E. Stewart) Subject: [microarray-ontol] Feedback needed for MAGE vote on OntologyEntry Hey All, There is a debate going on in the mage list: mged-mage@lists.sf.net discussing how to modify the current MAGE model's support for Ontologies. The current model was built to only support Ontologies that are *static* (to use Michael Miller's terminology). This is exemplified by GO where every entry in the Ontology has a unique identifier. The MGED sample Ontology does *not* fit this category, and so there is a proposal to modify MAGE's view of an OntologyEntry to include a self-referential association from OntologyEntry to itself. Although this change, seems minor - one more association - it is actually a fairly dramatic increase in the expressivity of the model, and what can be encoded in an OntologyEntry. However, I believe that there are more fundamental conflicts between the MGED sample ontology and MAGE that were discussed on this list a number of months ago, and never resolved (at least not to my satisfaction), and so I'm recommending that we do not approve the recommended change to the model. To be clear, I have no vote in this matter, just an opinion and a recommendation as to how Ugis should vote. I suspect folks on this list feel differently than I do, and so I encourage you to voice your ideas on the MAGE list. Also, please send us examples of how the MGED sample ontology would be used if this modification were to be approved. Cheers, jas. ------------------------------ Date: Fri, 6 Sep 2002 10:44:53 +0100 (BST) From: Alvis Brazma Subject: Re: [microarray-ontol] Feedback needed for MAGE vote on OntologyEntry Jason, I can't comment on the MAGE details, though I trust Ugis, who says that most of the detailed questions have been discussed for quite some time. >From ArrayExpress point of view it is absolutely necessary that we adopt MAGE as standard ASAP, even if it is not perfect. We are beginning to accumulate legacy problems already. For instance, the MAGE-ML files generated by the Sanger relate to older DTD than, for instance, Paul's files. How many times will we migrate the data from older versions? How many times the labs will have to change their LIMS? MAGE is not an academic project, the community needs the standard right now, otherwise we are in a typical catch 22. I'm sure Rosetta, Affymetrix and other companies have exactly the same problem. So, I strongly urge not to postpone the vote. When we have enough use-cases, we can start working on MAGE2 instead of perfecting this one. Cheers, - - Alvis On 5 Sep 2002, Jason E. Stewart wrote: > Hey All, > > There is a debate going on in the mage list: > > mged-mage@lists.sf.net > > discussing how to modify the current MAGE model's support for > Ontologies. > > The current model was built to only support Ontologies that are > *static* (to use Michael Miller's terminology). This is exemplified by > GO where every entry in the Ontology has a unique identifier. > > The MGED sample Ontology does *not* fit this category, and so there is > a proposal to modify MAGE's view of an OntologyEntry to include a > self-referential association from OntologyEntry to itself. > > Although this change, seems minor - one more association - it is > actually a fairly dramatic increase in the expressivity of the model, > and what can be encoded in an OntologyEntry. > > However, I believe that there are more fundamental conflicts between > the MGED sample ontology and MAGE that were discussed on this list a > number of months ago, and never resolved (at least not to my > satisfaction), and so I'm recommending that we do not approve the > recommended change to the model. > > To be clear, I have no vote in this matter, just an opinion and a > recommendation as to how Ugis should vote. > > I suspect folks on this list feel differently than I do, and so I > encourage you to voice your ideas on the MAGE list. Also, please send > us examples of how the MGED sample ontology would be used if this > modification were to be approved. > > Cheers, > jas. > ------------------------------ Date: 06 Sep 2002 11:19:52 -0600 From: jason@openinformatics.com (Jason E. Stewart) Subject: Re: [microarray-ontol] Feedback needed for MAGE vote on OntologyEntry "Alvis Brazma" writes: > From ArrayExpress point of view it is absolutely necessary that we > adopt MAGE as standard ASAP, even if it is not perfect. We are > beginning to accumulate legacy problems already. For instance, the > MAGE-ML files generated by the Sanger relate to older DTD than, for > instance, Paul's files. How many times will we migrate the data from > older versions? How many times the labs will have to change their > LIMS? Hey Alvis, You mentioned some things which I found strange - Why are you retaining XML files? Why haven't you imported them into ArrayExpress? Why would you need migrate from one DTD to another? > MAGE is not an academic project This is exactly *my point*. I'm alarmed at the subjectivity of some of our decisions. There are standard IT procedures to handle this process, and we've used none of them. It was my suggestion that we begin to use them sooner rather than later. That the OMG didn't require their use was surprising to me, given the scope of this specification. > the community needs the standard right now, otherwise we are in a > typical catch 22. I do comprehend the real-world situation you are in. I was my desire to help you avoid potential problems in your data handling process by bringing up some of the serious issues that I believed had not been sufficiently well thought out. > I'm sure Rosetta, Affymetrix and other companies have exactly the > same problem. So, I strongly urge not to postpone the vote. When we > have enough use-cases, we can start working on MAGE2 instead of > perfecting this one. Well, my appologies for cynicism, but don't hold your breath on those use cases - I don't see the volunteers lining up. jas. ------------------------------ Date: Fri, 6 Sep 2002 10:36:46 -0700 From: "Miller, Michael (Rosetta)" Subject: RE: [Mged-mage] Re: [microarray-ontol] Feedback needed for MAGE v ote on OntologyEntry--Versioning Jason, > Hey Alvis, > > You mentioned some things which I found strange - Why are you > retaining XML files? Why haven't you imported them into ArrayExpress? > Why would you need migrate from one DTD to another? > We have found that old files never go away. Since we haven't varied the MAGE-ML file name, if someone tries to load these old files they will not validate. It is absolutely essential that if a file ever validated against a production DTD, it will always validate. Because the MAGE-ML DTD was a known to be a work in progress, implementers have been good about the changes that have been occurring (thanks all). After v.1 is complete it will be essential that future versions have a different filename. As I've suggested, that may work well for our version control and automatic generation of the DTD. In retrospect, it might have been good to have started with a filename of MAGE-MLv01.dtd then moved that to MAGE-MLv10.dtd for the available one but I think we're fine for now with MAGE-ML.dtd for the v.1. thanks, Michael > -----Original Message----- > From: jason@openinformatics.com [mailto:jason@openinformatics.com] > Sent: Friday, September 06, 2002 10:20 AM > To: mged-mage@lists.sourceforge.net; microarray-ontol@ebi.ac.uk > Subject: [Mged-mage] Re: [microarray-ontol] Feedback needed for MAGE > vote on OntologyEntry > > > "Alvis Brazma" writes: > ------------------------------ Date: Sat, 7 Sep 2002 12:55:25 +0100 (BST) From: Alvis Brazma Subject: Re: [microarray-ontol] Feedback needed for MAGE vote on OntologyEntry > Hey Alvis, > > You mentioned some things which I found strange - Why are you > retaining XML files? Why haven't you imported them into ArrayExpress? > Why would you need migrate from one DTD to another? > because ArrayExpress loader is working not on the most recent DTD, and because AE implements not the very latest MAGE-OM. We are about to update that, but we can't afford to do that too many times. Therefore we need the final standard. - - Alvis ------------------------------ Date: Mon, 09 Sep 2002 08:12:38 -0400 From: Bruce Aronow Subject: Re: [microarray-ontol] Feedback needed for MAGE vote on OntologyEntry Hi guys, Couple of simple questions.. What if an ontology that is used for an annotation is defective, contradictory to other field values in MAGE, or unavailable or changed or just incomplete with respect to the actual experiment using an external DB lookup? What sort of checking mechanisms could we have? regards, Bruce On 9/6/02 5:44 AM, "Alvis Brazma" wrote: > Jason, > > I can't comment on the MAGE details, though I trust Ugis, who says that > most of the detailed questions have been discussed for quite some time. > > From ArrayExpress point of view it is absolutely necessary that we adopt > MAGE as standard ASAP, even if it is not perfect. We are beginning to > accumulate legacy problems already. For instance, the MAGE-ML files > generated by the Sanger relate to older DTD than, for instance, Paul's > files. How many times will we migrate the data from older versions? How > many times the labs will have to change their LIMS? MAGE is not an > academic project, the community needs the standard right now, otherwise we > are in a typical catch 22. I'm sure Rosetta, Affymetrix and other > companies have exactly the same problem. So, I strongly urge not to > postpone the vote. When we have enough use-cases, we can start working on > MAGE2 instead of perfecting this one. > > Cheers, > - Alvis > > > > On 5 Sep 2002, Jason E. Stewart wrote: > >> Hey All, >> >> There is a debate going on in the mage list: >> >> mged-mage@lists.sf.net >> >> discussing how to modify the current MAGE model's support for >> Ontologies. >> >> The current model was built to only support Ontologies that are >> *static* (to use Michael Miller's terminology). This is exemplified by >> GO where every entry in the Ontology has a unique identifier. >> >> The MGED sample Ontology does *not* fit this category, and so there is >> a proposal to modify MAGE's view of an OntologyEntry to include a >> self-referential association from OntologyEntry to itself. >> >> Although this change, seems minor - one more association - it is >> actually a fairly dramatic increase in the expressivity of the model, >> and what can be encoded in an OntologyEntry. >> >> However, I believe that there are more fundamental conflicts between >> the MGED sample ontology and MAGE that were discussed on this list a >> number of months ago, and never resolved (at least not to my >> satisfaction), and so I'm recommending that we do not approve the >> recommended change to the model. >> >> To be clear, I have no vote in this matter, just an opinion and a >> recommendation as to how Ugis should vote. >> >> I suspect folks on this list feel differently than I do, and so I >> encourage you to voice your ideas on the MAGE list. Also, please send >> us examples of how the MGED sample ontology would be used if this >> modification were to be approved. >> >> Cheers, >> jas. >> > > ------------------------------ Date: Mon, 9 Sep 2002 09:28:38 -0400 (EDT) From: Angel Pizarro Subject: Re: [microarray-ontol] Feedback needed for MAGE vote on OntologyEntry Bruce, this is really an application dependent question, but I will try and answer some of it w/ repsect to various scenerios that are general enough to apply to most applications. For checking data integrity, the first scenerio we had discussed at the Seattle Jamboree use the BioMOBY project (www.biomoby.org) for cheking these various types of information. What this entails of course is having URI's that conform to (or able to translate to) bioMOBY's expectations. Read more about it on their website. The other method is of course in-house data warehouse of ontologies you are interested in. The work involved in maintaining the warehouse determines which ontologies you are in need of and are worth the effort. This still leaves the question of what to do with erroneous entries. If you don't want to make your DB 'dirty' then you can't put the errors into the regular slot. They must go someplace else, either by storing the MAGE-ML instance itself or using some set of "scratch" tables to hold the data until it can be verified (i.e. our warehouse is out of date), fixed or plain ol' deleted because it wasn't important in the first place. You could also just import the data as-is and mark it dirty, but until an annotator gets a chance to look at it, people will have access to faulty data. Hope this helps. Angel On Mon, 9 Sep 2002, Bruce Aronow wrote: > Hi guys, > Couple of simple questions.. > What if an ontology that is used for an annotation is defective, > contradictory to other field values in MAGE, or unavailable or changed or > just incomplete with respect to the actual experiment using an external DB > lookup? What sort of checking mechanisms could we have? > regards, > Bruce > > > On 9/6/02 5:44 AM, "Alvis Brazma" wrote: > > > Jason, > > > > I can't comment on the MAGE details, though I trust Ugis, who says that > > most of the detailed questions have been discussed for quite some time. > > > > From ArrayExpress point of view it is absolutely necessary that we adopt > > MAGE as standard ASAP, even if it is not perfect. We are beginning to > > accumulate legacy problems already. For instance, the MAGE-ML files > > generated by the Sanger relate to older DTD than, for instance, Paul's > > files. How many times will we migrate the data from older versions? How > > many times the labs will have to change their LIMS? MAGE is not an > > academic project, the community needs the standard right now, otherwise we > > are in a typical catch 22. I'm sure Rosetta, Affymetrix and other > > companies have exactly the same problem. So, I strongly urge not to > > postpone the vote. When we have enough use-cases, we can start working on > > MAGE2 instead of perfecting this one. > > > > Cheers, > > - Alvis > > > > > > > > On 5 Sep 2002, Jason E. Stewart wrote: > > > >> Hey All, > >> > >> There is a debate going on in the mage list: > >> > >> mged-mage@lists.sf.net > >> > >> discussing how to modify the current MAGE model's support for > >> Ontologies. > >> > >> The current model was built to only support Ontologies that are > >> *static* (to use Michael Miller's terminology). This is exemplified by > >> GO where every entry in the Ontology has a unique identifier. > >> > >> The MGED sample Ontology does *not* fit this category, and so there is > >> a proposal to modify MAGE's view of an OntologyEntry to include a > >> self-referential association from OntologyEntry to itself. > >> > >> Although this change, seems minor - one more association - it is > >> actually a fairly dramatic increase in the expressivity of the model, > >> and what can be encoded in an OntologyEntry. > >> > >> However, I believe that there are more fundamental conflicts between > >> the MGED sample ontology and MAGE that were discussed on this list a > >> number of months ago, and never resolved (at least not to my > >> satisfaction), and so I'm recommending that we do not approve the > >> recommended change to the model. > >> > >> To be clear, I have no vote in this matter, just an opinion and a > >> recommendation as to how Ugis should vote. > >> > >> I suspect folks on this list feel differently than I do, and so I > >> encourage you to voice your ideas on the MAGE list. Also, please send > >> us examples of how the MGED sample ontology would be used if this > >> modification were to be approved. > >> > >> Cheers, > >> jas. > >> > > > > > - -- Angel Pizarro Programmer Analyst Center for Bioinformatics angel@pcbi.upenn.edu ------------------------------ Date: Mon, 09 Sep 2002 18:57:43 +0100 From: Helen Parkinson Subject: Re: [microarray-ontol] Feedback needed for MAGE vote on OntologyEntry Hi all, Angel gave a nice answer. Just to elaborate, at ArrayExpress we are not really in a position to tell anyone that their chosen annotation ontology is defective, though we may point them in the direction of an alternative one if we felt it to be more appropriate. We know that building a good ontology is tricky and time consuming and often ontologies are being used for purposes for which they were not designed. Defective is also a difficult term, almost no ontologies are perfect, and most are works in progress. If we really felt that there was a problem we would suggest talking to the ontology builders and working with them. In reality a lack of appropriate ontologies, use of synomyms may be more of a problem and getting users to use anything other than free text have been our major challenges. One of the things we have discussed is having an ArrayExpress list of approved terms/sources (for our use obviously, though we would make this public) and we might flag data as annotated in way that may cause problems with our query capability. We would still load such data and make it available. Contradiction with other areas in MAGE is something that our data curation team will take care of, and is a parsing issue from external applications. We are working with MAGE-ML exporters to try and get this right at present. These aren't hard and fast rules, just how we have talked about dealing with things till now. Bruce, is there a particular area you feel is problematic, or were you just throwing it out there for discussion? regards Helen Angel Pizarro wrote: > Bruce, this is really an application dependent question, but I will try > and answer some of it w/ repsect to various scenerios that are general > enough to apply to most applications. > > For checking data integrity, the first scenerio we had discussed at the > Seattle Jamboree use the BioMOBY project (www.biomoby.org) for cheking > these various types of information. What this entails of course is having > URI's that conform to (or able to translate to) bioMOBY's expectations. > Read more about it on their website. > > The other method is of course in-house data warehouse of ontologies you > are interested in. The work involved in maintaining the warehouse > determines which ontologies you are in need of and are worth the effort. > > This still leaves the question of what to do with erroneous entries. If > you don't want to make your DB 'dirty' then you can't put the errors into > the regular slot. They must go someplace else, either by storing the > MAGE-ML instance itself or using some set of "scratch" tables to hold the > data until it can be verified (i.e. our warehouse is out of date), fixed > or plain ol' deleted because it wasn't important in the first place. > > You could also just import the data as-is and mark it dirty, but until an > annotator gets a chance to look at it, people will have access to faulty > data. > > Hope this helps. > > Angel > > On Mon, 9 Sep 2002, Bruce Aronow wrote: > > >>Hi guys, >>Couple of simple questions.. >>What if an ontology that is used for an annotation is defective, >>contradictory to other field values in MAGE, or unavailable or changed or >>just incomplete with respect to the actual experiment using an external DB >>lookup? What sort of checking mechanisms could we have? >>regards, >>Bruce >> >> >>On 9/6/02 5:44 AM, "Alvis Brazma" wrote: >> >> >>>Jason, >>> >>>I can't comment on the MAGE details, though I trust Ugis, who says that >>>most of the detailed questions have been discussed for quite some time. >>> >>>From ArrayExpress point of view it is absolutely necessary that we adopt >>>MAGE as standard ASAP, even if it is not perfect. We are beginning to >>>accumulate legacy problems already. For instance, the MAGE-ML files >>>generated by the Sanger relate to older DTD than, for instance, Paul's >>>files. How many times will we migrate the data from older versions? How >>>many times the labs will have to change their LIMS? MAGE is not an >>>academic project, the community needs the standard right now, otherwise we >>>are in a typical catch 22. I'm sure Rosetta, Affymetrix and other >>>companies have exactly the same problem. So, I strongly urge not to >>>postpone the vote. When we have enough use-cases, we can start working on >>>MAGE2 instead of perfecting this one. >>> >>>Cheers, >>>- Alvis >>> >>> >>> >>>On 5 Sep 2002, Jason E. Stewart wrote: >>> >>> >>>>Hey All, >>>> >>>>There is a debate going on in the mage list: >>>> >>>> mged-mage@lists.sf.net >>>> >>>>discussing how to modify the current MAGE model's support for >>>>Ontologies. >>>> >>>>The current model was built to only support Ontologies that are >>>>*static* (to use Michael Miller's terminology). This is exemplified by >>>>GO where every entry in the Ontology has a unique identifier. >>>> >>>>The MGED sample Ontology does *not* fit this category, and so there is >>>>a proposal to modify MAGE's view of an OntologyEntry to include a >>>>self-referential association from OntologyEntry to itself. >>>> >>>>Although this change, seems minor - one more association - it is >>>>actually a fairly dramatic increase in the expressivity of the model, >>>>and what can be encoded in an OntologyEntry. >>>> >>>>However, I believe that there are more fundamental conflicts between >>>>the MGED sample ontology and MAGE that were discussed on this list a >>>>number of months ago, and never resolved (at least not to my >>>>satisfaction), and so I'm recommending that we do not approve the >>>>recommended change to the model. >>>> >>>>To be clear, I have no vote in this matter, just an opinion and a >>>>recommendation as to how Ugis should vote. >>>> >>>>I suspect folks on this list feel differently than I do, and so I >>>>encourage you to voice your ideas on the MAGE list. Also, please send >>>>us examples of how the MGED sample ontology would be used if this >>>>modification were to be approved. >>>> >>>>Cheers, >>>>jas. >>>> >>>> >>> > ------------------------------ Date: 11 Sep 2002 13:55:12 -0600 From: "Jason E. Stewart" Subject: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML "Chris Stoeckert" writes: > What I would suggest (and what we are doing) is using the MGED > ontology class for the 'category,' So in this case the category would > be "Individual" and the value would be the patient ID. The database > entry can point to the MGED ontology. Hey Chris, The problem is that 'Individual' is a not as informative as I would like it to be. I really want to know that it's coming from the MGED sample ontology classification of Individual and not someone elses. Also, the category should include 'Patient ID' if that's what it is, I feel that naming it Individual leaves it unclear - maybe it's their SSN or their Military ID # ... Can I propose something (to borrow ideas from MyGrid, I3C, and MOBY) along the lines of: mged.org:Ontologies/Sample:Individual/PatientID Yes, it's verbose, but it's also very clear. * mged.org - the Authority responsible for the Ontology * Ontologies/Sample - since MGED will likely have more than just the sample ontology * Individual/PatientID - if PatientID exists in the Ontology already, then that should be enough without 'Individual' Another way that I'm less happy about would be: mged.org:Ontologies/Sample:Individual/ID_Number and then using the description to indicate that it was a Patient ID number. Although it's more general and can be used for SSN and other ID number schemes, it relies on a free text description to disambiguate it from other types of ID numbers. jas. ------------------------------ Date: Wed, 11 Sep 2002 16:36:02 -0400 From: Chris Stoeckert Subject: Re: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML Hi Jason, You would know that it's coming from the MGED ontology (and the version of the ontology) the same way you would know where any OntolgyEntry came from i.e, the DatabaseEntry association. The URI:http://www.cbil.upenn.edu/Ontology/MGEDontology.html already tells you mged.org:Ontologies/Sample: FYI, There is only one MGED ontology which includes sample descriptions but also other parts of microarray experiments (e.g., study design). The category 'Individual' is an ontology term which is defined. Description can be used to add information such as 'Patient ID.' PatientID is not a registered term in the MGED ontology so you can't just make it up and say it came from there! If the ontology group felt that Individual required further specification this would change things. For example, we could make PatientID a subclass of Individual or we could give Individual an attribute (property) has_ID_type (patient ID, SSN, ..). The only use case I can think of where specifying the type of ID would be important as part of the ontology is if multiple types of identifiers are used in a study and desired queries are made based on the type of identifier. Unless I hear that this is an important use case, I'm inclined to stick with Individual and use Description to describe it further as necessary. Comments from others? Cheers, Chris On Wednesday, September 11, 2002, at 03:55 PM, Jason E. Stewart wrote: > "Chris Stoeckert" writes: > >> What I would suggest (and what we are doing) is using the MGED >> ontology class for the 'category,' So in this case the category would >> be "Individual" and the value would be the patient ID. The database >> entry can point to the MGED ontology. > > Hey Chris, > > The problem is that 'Individual' is a not as informative as I would > like it to be. I really want to know that it's coming from the MGED > sample ontology classification of Individual and not someone elses. > > Also, the category should include 'Patient ID' if that's what it is, I > feel that naming it Individual leaves it unclear - maybe it's their > SSN or their Military ID # ... > > Can I propose something (to borrow ideas from MyGrid, I3C, and MOBY) > along the lines of: > > mged.org:Ontologies/Sample:Individual/PatientID > > Yes, it's verbose, but it's also very clear. > > * mged.org - the Authority responsible for the Ontology > > * Ontologies/Sample - since MGED will likely have more than just the > sample ontology > > * Individual/PatientID - if PatientID exists in the Ontology already, > then that should be enough without 'Individual' > > Another way that I'm less happy about would be: > > mged.org:Ontologies/Sample:Individual/ID_Number > > and then using the description to indicate that it was a Patient ID > number. Although it's more general and can be used for SSN and other > ID number schemes, it relies on a free text description to > disambiguate it from other types of ID numbers. > > jas. > > > > ------------------------------ Date: Wed, 11 Sep 2002 13:44:00 -0700 (PDT) From: Cathy Ball Subject: Re: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML Hi all, I think that this is an important use case NOT to use. There will be absolutely no reason to schlep around Patient IDs or SSNs -- it's a huge privacy issue. There might be some use for projects to share such IDs internally, but it would be an error to assume we'll ever be reporting such IDs between projects. Therefore, I side with Chris and think we can safely stick witn Individual for the time being. Cathy On Wed, 11 Sep 2002, Chris Stoeckert wrote: > Hi Jason, > You would know that it's coming from the MGED ontology (and the version > of the ontology) the same way you would know where any OntolgyEntry came > from i.e, the DatabaseEntry association. The > URI:http://www.cbil.upenn.edu/Ontology/MGEDontology.html already tells > you mged.org:Ontologies/Sample: FYI, There is only one MGED ontology > which includes sample descriptions but also other parts of microarray > experiments (e.g., study design). > > The category 'Individual' is an ontology term which is defined. > Description can be used to add information such as 'Patient ID.' > PatientID is not a registered term in the MGED ontology so you can't > just make it up and say it came from there! If the ontology group felt > that Individual required further specification this would change > things. For example, we could make PatientID a subclass of Individual > or we could give Individual an attribute (property) has_ID_type (patient > ID, SSN, ..). The only use case I can think of where specifying the type > of ID would be important as part of the ontology is if multiple types of > identifiers are used in a study and desired queries are made based on > the type of identifier. Unless I hear that this is an important use > case, I'm inclined to stick with Individual and use Description to > describe it further as necessary. > > Comments from others? > > Cheers, > Chris > > On Wednesday, September 11, 2002, at 03:55 PM, Jason E. Stewart wrote: > > > "Chris Stoeckert" writes: > > > >> What I would suggest (and what we are doing) is using the MGED > >> ontology class for the 'category,' So in this case the category would > >> be "Individual" and the value would be the patient ID. The database > >> entry can point to the MGED ontology. > > > > Hey Chris, > > > > The problem is that 'Individual' is a not as informative as I would > > like it to be. I really want to know that it's coming from the MGED > > sample ontology classification of Individual and not someone elses. > > > > Also, the category should include 'Patient ID' if that's what it is, I > > feel that naming it Individual leaves it unclear - maybe it's their > > SSN or their Military ID # ... > > > > Can I propose something (to borrow ideas from MyGrid, I3C, and MOBY) > > along the lines of: > > > > mged.org:Ontologies/Sample:Individual/PatientID > > > > Yes, it's verbose, but it's also very clear. > > > > * mged.org - the Authority responsible for the Ontology > > > > * Ontologies/Sample - since MGED will likely have more than just the > > sample ontology > > > > * Individual/PatientID - if PatientID exists in the Ontology already, > > then that should be enough without 'Individual' > > > > Another way that I'm less happy about would be: > > > > mged.org:Ontologies/Sample:Individual/ID_Number > > > > and then using the description to indicate that it was a Patient ID > > number. Although it's more general and can be used for SSN and other > > ID number schemes, it relies on a free text description to > > disambiguate it from other types of ID numbers. > > > > jas. > > > > > > > > > ------------------------------ Date: Wed, 11 Sep 2002 13:56:37 -0700 From: "Miller, Michael (Rosetta)" Subject: RE: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML Hi All, > > You would know that it's coming from the MGED ontology (and the version > > of the ontology) the same way you would know where any OntolgyEntry came > > from i.e, the DatabaseEntry association. The > > URI:http://www.cbil.upenn.edu/Ontology/MGEDontology.html already tells > > you mged.org:Ontologies/Sample: This was indeed how I thought we had modeled this. > There might be some use for projects to share such IDs > internally, but it would be an error to assume we'll ever be reporting > such IDs between projects. Therefore, I side with Chris and > think we can > safely stick witn Individual for the time being. But, without giving away who the actual individual was, there might be a need to make sure that multiple documents that reference this individual have a single, consistent reference to the individual. This can be accomplished using the same encryption algorithm on something like "person name:ssn:data of birth" each time the identifier is needed. Michael > -----Original Message----- > From: Cathy Ball [mailto:ball@genome.stanford.edu] > Sent: Wednesday, September 11, 2002 1:44 PM > To: Chris Stoeckert > Cc: Jason E. Stewart; Guy Karlebach; mged-mage@lists.sourceforge.net; > microarray-ontol@ebi.ac.uk > Subject: Re: [microarray-ontol] Re: [Mged-mage] A question > about MAGE-ML > > > Hi all, > ------------------------------ Date: Wed, 11 Sep 2002 14:15:33 -0700 From: paul spellman Subject: Re: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML I can't say that I particularly agree with Cathy (we're not the anonymity police). But I don't think there is a use case for patient_id explicitly, primarily because I can't see how anyone would search on it. If you wanted to search for a patient id and it was stored as the value for individual, you would know what it was and where it was so you just parse through the individuals til you find the right one. In some ways though my bigger concern is the redundancy of what can be encoded in MAGE and what can be encoded by a reference to an ontology. I think it is very interesting that Chris actually mostly agrees with me. The exact type of id (patient id) is pretty much unimportant. The key is that it relates to a particular individual. I would argue that this is better represented in MAGE than through an ontology ref called individual. The reason is that the object (represented by patient id XXX) can have explicit associations to other objects (represented by their tumor, or tumor biopsy, or heart, or lung or whatever), while representing each of these separately with an ontology entry to individual with the same identifier requires more software intelligence... Paul On Wednesday, September 11, 2002, at 01:56 PM, Miller, Michael (Rosetta) wrote: > Hi All, > >>> You would know that it's coming from the MGED ontology (and the >>> version >>> of the ontology) the same way you would know where any OntolgyEntry >>> came > >>> from i.e, the DatabaseEntry association. The >>> URI:http://www.cbil.upenn.edu/Ontology/MGEDontology.html already >>> tells >>> you mged.org:Ontologies/Sample: > This was indeed how I thought we had modeled this. > >> There might be some use for projects to share such IDs >> internally, but it would be an error to assume we'll ever be reporting >> such IDs between projects. Therefore, I side with Chris and >> think we can >> safely stick witn Individual for the time being. > But, without giving away who the actual individual was, there might be > a > need to make sure that multiple documents that reference this > individual > have a single, consistent reference to the individual. This can be > accomplished using the same encryption algorithm on something like > "person > name:ssn:data of birth" each time the identifier is needed. > > Michael > > >> -----Original Message----- >> From: Cathy Ball [mailto:ball@genome.stanford.edu] >> Sent: Wednesday, September 11, 2002 1:44 PM >> To: Chris Stoeckert >> Cc: Jason E. Stewart; Guy Karlebach; mged-mage@lists.sourceforge.net; >> microarray-ontol@ebi.ac.uk >> Subject: Re: [microarray-ontol] Re: [Mged-mage] A question >> about MAGE-ML >> >> >> Hi all, >> > > > > > ------------------------------------------------------- > In remembrance > www.osdn.com/911/ > _______________________________________________ > Mged-mage mailing list > Mged-mage@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/mged-mage ------------------------------ Date: Wed, 11 Sep 2002 17:39:05 -0400 From: Chris Stoeckert Subject: [microarray-ontol] Fwd: Non-member submission from ["White, Joseph (Joe) A." ] > From: "White, Joseph (Joe) A." > To: "'Jason E. Stewart'" , > Guy Karlebach > , mged-mage@lists.sourceforge.net > Cc: microarray-ontol@ebi.ac.uk > Subject: RE: [Mged-mage] A question about MAGE-ML > Date: Wed, 11 Sep 2002 16:07:45 -0400 > MIME-Version: 1.0 > X-Mailer: Internet Mail Service (5.5.2653.19) > Content-Type: text/plain; > charset="iso-8859-1" > > Hi Jason, Helen, Guy, and Michael, > > I don't mean to tell anyone what data to use or submit to AE, but would > recommend that we steer clear of social security numbers and any > identifier > that could conceivable be misused. The possibility that these IDs > could be > misused is real, and the consequences of misuse are severe. > > So I'm suggesting that we think very hard about what identifiers to > supply > and who will have access to them. > > Joe White > > > -----Original Message----- > From: Jason E. Stewart [mailto:jason@openinformatics.com] > Sent: Wednesday, September 11, 2002 3:55 PM > To: Guy Karlebach; mged-mage@lists.sourceforge.net > Cc: microarray-ontol@ebi.ac.uk > Subject: Re: [Mged-mage] A question about MAGE-ML > > > "Chris Stoeckert" writes: > >> What I would suggest (and what we are doing) is using the MGED >> ontology class for the 'category,' So in this case the category would >> be "Individual" and the value would be the patient ID. The database >> entry can point to the MGED ontology. > > Hey Chris, > > The problem is that 'Individual' is a not as informative as I would > like it to be. I really want to know that it's coming from the MGED > sample ontology classification of Individual and not someone elses. > > Also, the category should include 'Patient ID' if that's what it is, I > feel that naming it Individual leaves it unclear - maybe it's their > SSN or their Military ID # ... > > Can I propose something (to borrow ideas from MyGrid, I3C, and MOBY) > along the lines of: > > mged.org:Ontologies/Sample:Individual/PatientID > > Yes, it's verbose, but it's also very clear. > > * mged.org - the Authority responsible for the Ontology > > * Ontologies/Sample - since MGED will likely have more than just the > sample ontology > > * Individual/PatientID - if PatientID exists in the Ontology already, > then that should be enough without 'Individual' > > Another way that I'm less happy about would be: > > mged.org:Ontologies/Sample:Individual/ID_Number > > and then using the description to indicate that it was a Patient ID > number. Although it's more general and can be used for SSN and other > ID number schemes, it relies on a free text description to > disambiguate it from other types of ID numbers. > > jas. > > > > > > > ------------------------------------------------------- > In remembrance > www.osdn.com/911/ > _______________________________________________ > Mged-mage mailing list > Mged-mage@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/mged-mage > ------------------------------ Date: 11 Sep 2002 15:48:20 -0600 From: "Jason E. Stewart" Subject: Re: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML Hey Chris, Thanks for filling me in. The tone of your letter seems a bit aggrevated, I'm sorry if I'm being ignorant, I'm doing my best to understand the issues. "Chris Stoeckert" writes: > You would know that it's coming from the MGED ontology (and the > version of the ontology) the same way you would know where any > OntolgyEntry came from i.e, the DatabaseEntry association. The > URI:http://www.cbil.upenn.edu/Ontology/MGEDontology.html already tells > you mged.org:Ontologies/Sample: There's a couple of important points. 1) This isn't completely correct, it might have just been a simplification on your part, but I wanted to point it out. http://www.cbil.upenn.edu/Ontology/MGEDontology.html is the URI to the Ontology itself, it should be listed as the URI in the Database object associated with the DatabaseEntry. The URI in the DatabaseEntry is intended to retrieve the instance of the OntologyEntry from that database. This is what doesn't quite work for 'dynamic' ontologies like the MGED sample ontology, there is no way to pull out the exact instance of: category: Individual value: 123-45-6789 from the MGED Sample ontology - because that's a specific instance, and the instances don't live in the Ontology only the categories. (please correct me if there is a better differentiation than 'static' and 'dynamic'). Since I designed this mechanism with static Ontologies, e.g. GO, in mind that was the reason we re-used the DatabaseEntry/Database associations. Because GO:123456789 is an accession number that pulls exactly that information out of the DB. With dynamic ontologies it is possible to do other things, but we'd at least have to modify the documentation. 2) The DatabaseEntry is *optional*. The idea was to include enough information in the OntologyEntry so that it wasn't necessary to use the DatabaseEntry/Database associations. 3) It is a good idea to begin thinking about MOBY and MyGrid and WWW services. The nomenclature I suggested will make the MGED Sample Ontology much more accessible because it is in the form of a Life Science Identifier (LSID) - which is modeled after the unique identifier format the grid computing people have developed. > FYI, There is only one MGED ontology which includes sample > descriptions but also other parts of microarray experiments (e.g., > study design). In MGED IV I mentioned that MAGE has ~30 simple ontologies such as the dataType and scale associations for QuantitationType. Since these are simple controlled vocabularies, I thought that it would make the most sense to keep them separate from the more complicated Sample Ontology structure. > The category 'Individual' is an ontology term which is > defined. Description can be used to add information such as 'Patient > ID.' PatientID is not a registered term in the MGED ontology so you > can't just make it up and say it came from there! I'm sorry if I offended you, Chris, I know very little about the Sample Ontology, and I did the best I could to come up with an example. > If the ontology group felt that Individual required further > specification this would change things. Last I knew, I had joined the mailing list because I intended to participate in the Ontology group. > For example, we could make PatientID a subclass of Individual or we > could give Individual an attribute (property) has_ID_type (patient > ID, SSN, ..). This is what I was wondering. > The only use case I can think of where specifying the type of ID > would be important as part of the ontology is if multiple types of > identifiers are used in a study and desired queries are made based > on the type of identifier. Unless I hear that this is an important > use case, I'm inclined to stick with Individual and use Description > to describe it further as necessary. I agree, that I don't think multiple ID's for a single individual is likely. Perhaps 'description' is the best place to put that information - once again that wasn't what it's intended use was for - it was meant to be a static description of the 'meaning' of the term, i.e. a dictionary definition. I'm happy to be having this discussion, and as I've said, there's much that I haven't yet learned about the structure of the Sample Ontology, so please forgive my ignorance. I think the Sample Ontology will become very important to us all, and so I believe that it's very important that we have a recommended 'best practice' on how to use it with MAGE. I hope that I can be of help in working this out. Cheers, jas. ------------------------------ Date: 11 Sep 2002 15:53:28 -0600 From: "Jason E. Stewart" Subject: Re: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML "Cathy Ball" writes: > I think that this is an important use case NOT to use. There will > be absolutely no reason to schlep around Patient IDs or SSNs -- it's > a huge privacy issue. There might be some use for projects to share > such IDs internally, but it would be an error to assume we'll ever > be reporting such IDs between projects. I agree that the privacy issue is a big one, but just because we don't thinks it's a good idea for people to be making the data public doesn't mean that MAGE or the MGED Sample Ontology should ignore it. Because it *will* be used internally in data pipelines within an organization, this is *exactly* the type of information that must be conveyed in some way. Perhaps this is the wrong way to do it, but we shouldn't ignore it. jas. ------------------------------ Date: Thu, 12 Sep 2002 08:07:26 -0700 From: "Miller, Michael (Rosetta)" Subject: RE: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML All, Just this morning I read a bit about the IGC (http://www.genomeweb.com/articles/view.asp?Article=20029111668) and their efforts to create a public repository of tissues. Projects such as this will need to identifier individuals, but the point is that for public purposes, the identification used does not lead back to the person. It's only neccessary that the identification is unique in the scope of the project or repository. Michael > -----Original Message----- > From: Jason E. Stewart [mailto:jason@openinformatics.com] > Sent: Wednesday, September 11, 2002 2:53 PM > To: mged-mage@lists.sourceforge.net; microarray-ontol@ebi.ac.uk > Subject: Re: [microarray-ontol] Re: [Mged-mage] A question > about MAGE-ML > > > "Cathy Ball" writes: > > > I think that this is an important use case NOT to use. There will > > be absolutely no reason to schlep around Patient IDs or SSNs -- it's > > a huge privacy issue. There might be some use for projects to share > > such IDs internally, but it would be an error to assume we'll ever > > be reporting such IDs between projects. > > I agree that the privacy issue is a big one, but just because we don't > thinks it's a good idea for people to be making the data public > doesn't mean that MAGE or the MGED Sample Ontology should ignore it. > > Because it *will* be used internally in data pipelines within an > organization, this is *exactly* the type of information that must be > conveyed in some way. > > Perhaps this is the wrong way to do it, but we shouldn't ignore it. > > jas. > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Mged-mage mailing list > Mged-mage@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/mged-mage > ------------------------------ Date: Fri, 13 Sep 2002 10:57:56 +0100 From: "Causton, Helen" Subject: [microarray-ontol] RE: A question re. MAGE-ML and ontologies Ontologies should define only subject specific terms. The OE has to be very comprehensive in describing entities relevant to microarray experiments. Terms that need not be included are those that are widely and precisely understood by those in the field and terms that are not directly part of the subject. We believe that characteristics such as IDs, which fall into the latter category, should not be included in ontologies and such values need not have ontology references. If biosource id has an ontology entry then there is no reason why bioassay (ie hybridisation, scanning etc) id's can't have ontology entries. This can be extended to potentially all attributes in MAGE-OM. One solution to this problem is to leave IDs as part of MAGE-OM without any links to ontology entries. IDs could be stored, as attributes, in their respective classes. Helen Causton and Mahendra Navarange Head, CSC/IC Microarray Centre Imperial College Hammersmith Campus Du Cane Road London W12 ONN UK http://microarray.csc.mrc.ac.uk - -----Original Message----- From: Jason E. Stewart [mailto:jason@openinformatics.com] Sent: 11 September 2002 20:55 To: Guy Karlebach; mged-mage@lists.sourceforge.net Cc: microarray-ontol@ebi.ac.uk Subject: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML "Chris Stoeckert" writes: > What I would suggest (and what we are doing) is using the MGED > ontology class for the 'category,' So in this case the category would > be "Individual" and the value would be the patient ID. The database > entry can point to the MGED ontology. Hey Chris, The problem is that 'Individual' is a not as informative as I would like it to be. I really want to know that it's coming from the MGED sample ontology classification of Individual and not someone elses. Also, the category should include 'Patient ID' if that's what it is, I feel that naming it Individual leaves it unclear - maybe it's their SSN or their Military ID # ... Can I propose something (to borrow ideas from MyGrid, I3C, and MOBY) along the lines of: mged.org:Ontologies/Sample:Individual/PatientID Yes, it's verbose, but it's also very clear. * mged.org - the Authority responsible for the Ontology * Ontologies/Sample - since MGED will likely have more than just the sample ontology * Individual/PatientID - if PatientID exists in the Ontology already, then that should be enough without 'Individual' Another way that I'm less happy about would be: mged.org:Ontologies/Sample:Individual/ID_Number and then using the description to indicate that it was a Patient ID number. Although it's more general and can be used for SSN and other ID number schemes, it relies on a free text description to disambiguate it from other types of ID numbers. jas. ------------------------------ Date: 13 Sep 2002 09:30:11 -0600 From: "Jason E. Stewart" Subject: Re: [microarray-ontol] RE: A question re. MAGE-ML and ontologies "Causton, Helen" writes: > Ontologies should define only subject specific terms. The OE has to be very > comprehensive in describing entities relevant to microarray experiments. > Terms that need not be included are those that are widely and precisely > understood by those in the field and terms that are not directly part of the > subject. > > We believe that characteristics such as IDs, which fall into the latter > category, should not be included in ontologies and such values need not have > ontology references. If biosource id has an ontology entry then there is no > reason why bioassay (ie hybridisation, scanning etc) id's can't have > ontology entries. This can be extended to potentially all attributes in > MAGE-OM. > > One solution to this problem is to leave IDs as part of MAGE-OM without any > links to ontology entries. IDs could be stored, as attributes, in their > respective classes. Awesome! Thank you very much Helen for helping me think about this issue. I believe that is really a good solution. Yes, BioMaterial already has a MAGE-OM identifier attribute, and using the LSID nomenclature we can uniquely specify exactly what type of identifier we are using, something like: LSID:army.mil:MilityaryID:123456789 Examples such as this - which help deliniate the boundary between MAGE-OM and the Ontologies - are really what I need to understand. Thanks again, jas. ------------------------------ Date: Fri, 13 Sep 2002 17:24:16 +0100 From: "Navarange, Mahendra" Subject: [microarray-ontol] Using Ontologies with MAGE-OM Hi, If ontologies were to be used only to define terms that are ambiguous in our subject area then it might be useful not to insist on defining everything. Another drawback of defining a rigid framework of terminology is that people will assume that its a directive and will stop at providing only that information which has ontology reference. (Everything is defined in ontologyref including even IDs, so just give'em that and we're done!!!) Obviously this not what we want. We want to use MAGE-OM as our data model for storage of data with Ontology references being used only to avoid ambiguity. we also want the facility to enter more data that may not have been defined in ontology yet. Besides the above, I would also suggest not using the GO style number based ontology but instead use generally understood and accepted dot convention to represent hieratchical relationship between ontology terms. For example, lets assume that I want to describe "bedding" which is an ambiguous term (but fortunately I found an ontology entry for it), I would desribe my OntologyReferenceID as "MGEDOntology.BiomaterialDescription.BiomaterialManipulation.EnvironmentalHi story.bedding". Its mouthful but easy to understand than a plain simple number. It could be argued that a number translator could be written to switch back and forth from numbers and dot separated names. But from the point of view of researchers numbers are hard to understand and also numbers by themselves may not represent the hierarchical relationship between terms. In above example its readily apparent that the "bedding" is being used in BiomaterialManipulation context. Regards, Mahendra Navarange - -----Original Message----- From: Jason E. Stewart [mailto:jason@openinformatics.com] Sent: 13 September 2002 16:30 To: Causton, Helen Cc: mged-mage@lists.sourceforge.net; microarray-ontol@ebi.ac.uk Subject: Re: [microarray-ontol] RE: A question re. MAGE-ML and ontologies "Causton, Helen" writes: > Ontologies should define only subject specific terms. The OE has to be very > comprehensive in describing entities relevant to microarray experiments. > Terms that need not be included are those that are widely and precisely > understood by those in the field and terms that are not directly part of the > subject. > > We believe that characteristics such as IDs, which fall into the latter > category, should not be included in ontologies and such values need not have > ontology references. If biosource id has an ontology entry then there is no > reason why bioassay (ie hybridisation, scanning etc) id's can't have > ontology entries. This can be extended to potentially all attributes in > MAGE-OM. > > One solution to this problem is to leave IDs as part of MAGE-OM without any > links to ontology entries. IDs could be stored, as attributes, in their > respective classes. Awesome! Thank you very much Helen for helping me think about this issue. I believe that is really a good solution. Yes, BioMaterial already has a MAGE-OM identifier attribute, and using the LSID nomenclature we can uniquely specify exactly what type of identifier we are using, something like: LSID:army.mil:MilityaryID:123456789 Examples such as this - which help deliniate the boundary between MAGE-OM and the Ontologies - are really what I need to understand. Thanks again, jas. ------------------------------ Date: 13 Sep 2002 11:11:35 -0600 From: "Jason E. Stewart" Subject: Re: [microarray-ontol] Using Ontologies with MAGE-OM Hi Mahendra, "Navarange, Mahendra" writes: > Besides the above, I would also suggest not using the GO style number based > ontology but instead use generally understood and accepted dot convention to > represent hieratchical relationship between ontology terms. I believe it is important to have: 1) versioning - so we have a record of how the ontology has changed over time, getting to past states is important. I think systems like CVS and subversion address this issue adequately. 2) unique identifiers - so that we can unambiguously locate a term within the ontology. 3) immutible identifiers - once a term is given it's identifier it doesn't change. This helps prevent problems when viewing legacy data. In my mind, any scheme that satisfies these can work. Using a number as an identifier makes it simple, but it is not a very informative piece of information. The problem I can foresee with using dot-style identifiers is that they can potentially break 3) if a decision is made to re-order the ontology. > If ontologies were to be used only to define terms that are ambiguous in our > subject area then it might be useful not to insist on defining everything. > Another drawback of defining a rigid framework of terminology is that people > will assume that its a directive and will stop at providing only that > information which has ontology reference. (Everything is defined in > ontologyref including even IDs, so just give'em that and we're done!!!) > > Obviously this not what we want. We want to use MAGE-OM as our data model > for storage of data with Ontology references being used only to avoid > ambiguity. we also want the facility to enter more data that may not have > been defined in ontology yet. Yes, in my mind the key role of Ontologies is to provide rigorous meaning to information - to avoid ambiguity. > For example, lets assume that I want to describe "bedding" which is an > ambiguous term (but fortunately I found an ontology entry for it), I would > desribe my OntologyReferenceID as > "MGEDOntology.BiomaterialDescription.BiomaterialManipulation.EnvironmentalHi > story.bedding". Its mouthful but easy to understand than a plain simple > number. It could be argued that a number translator could be written to > switch back and forth from numbers and dot separated names. But from the > point of view of researchers numbers are hard to understand and also numbers > by themselves may not represent the hierarchical relationship between terms. > In above example its readily apparent that the "bedding" is being used in > BiomaterialManipulation context. I believe you've correctly identified the key issue: It could be argued that a number translator could be written to switch back and forth from numbers and dot separated names I believe that this is a User Interface issue. In that community they talk about the Model-View-Controller architecture and one of the important pieces is separating how the information is stored (the Model) from how the information is presented (the View). Numbers are generally not all that useful to people, but to computers they are extremely useful. If you are using a tool (a piece of computer software) to assign an ontology entry to describe your sample, what matters is how the tool presents the information to you and the way it is presented is separate from how the information is actually stored and retrieved by the software. Most people would prefer a nice hierarchical browser that enabled them to locate the most appropriate term, or a search field to give them a few potential suggestions - few people I know would be satisfied if all they were given is numerical ID's. Also, if you were viewing a piece of data that had already been annotated, it would be important for that tool to translate numerical ID's into the more informative dot-style nomenclature. So I believe it is possible (and potentially desirable) to use a numerical identifier system, while ensuring that researchers never have to actually see and learn what the identifiers are - that is something that should stay in the realm of the software developers. Just my opinion, jas. ------------------------------ Date: Mon, 16 Sep 2002 13:22:25 +0100 From: Helen Parkinson Subject: [microarray-ontol] Re: [Mged-mage] RE: A question re. MAGE-ML and ontologies Hi all, I am coming into this discussion a bit late as I've been away, few comments below. Having reviewed what has been posted I am still of the view that patient id is an ontology entry, and is adequately described in MGED. There is nothing in the definition that says this should actually identify a real patient, it can be anonymised, but I think the use of such an ID is critical when different consortia members are using the same samples. Michael made an excellent point about when the sample may be from the same pateint but a difft tissue, these then both ontology entries. I do not see the distinction. We also have particular needs in AE for unique id for parts of MAGE, so I don not recommend use of such id's for these cases. Causton, Helen wrote: > Ontologies should define only subject specific terms. The OE has to be very > comprehensive in describing entities relevant to microarray experiments. > Terms that need not be included are those that are widely and precisely > understood by those in the field and terms that are not directly part of the > subject. > > We believe that characteristics such as IDs, which fall into the latter > category, should not be included in ontologies and such values need not have > ontology references. If biosource id has an ontology entry then there is no > reason why bioassay (ie hybridisation, scanning etc) id's can't have > ontology entries. This can be extended to potentially all attributes in > MAGE-OM. I agree, but this is why such a thing as a patient id, which is in my view a BioSource Characteristic and hence an Ontology Entry should not be used as BioSource Id, these should be just simple identifiers. > > One solution to this problem is to leave IDs as part of MAGE-OM without any > links to ontology entries. IDs could be stored, as attributes, in their > respective classes. Exactly cheers Helen > > Helen Causton and Mahendra Navarange > Head, CSC/IC Microarray Centre > Imperial College > Hammersmith Campus > Du Cane Road > London W12 ONN > UK > > http://microarray.csc.mrc.ac.uk > > > -----Original Message----- > From: Jason E. Stewart [mailto:jason@openinformatics.com] > Sent: 11 September 2002 20:55 > To: Guy Karlebach; mged-mage@lists.sourceforge.net > Cc: microarray-ontol@ebi.ac.uk > Subject: [microarray-ontol] Re: [Mged-mage] A question about MAGE-ML > > > "Chris Stoeckert" writes: > > >>What I would suggest (and what we are doing) is using the MGED >>ontology class for the 'category,' So in this case the category would >>be "Individual" and the value would be the patient ID. The database >>entry can point to the MGED ontology. >> > > Hey Chris, > > The problem is that 'Individual' is a not as informative as I would > like it to be. I really want to know that it's coming from the MGED > sample ontology classification of Individual and not someone elses. > > Also, the category should include 'Patient ID' if that's what it is, I > feel that naming it Individual leaves it unclear - maybe it's their > SSN or their Military ID # ... > > Can I propose something (to borrow ideas from MyGrid, I3C, and MOBY) > along the lines of: > > mged.org:Ontologies/Sample:Individual/PatientID > > Yes, it's verbose, but it's also very clear. > > * mged.org - the Authority responsible for the Ontology > > * Ontologies/Sample - since MGED will likely have more than just the > sample ontology > > * Individual/PatientID - if PatientID exists in the Ontology already, > then that should be enough without 'Individual' > > Another way that I'm less happy about would be: > > mged.org:Ontologies/Sample:Individual/ID_Number > > and then using the description to indicate that it was a Patient ID > number. Although it's more general and can be used for SSN and other > ID number schemes, it relies on a free text description to > disambiguate it from other types of ID numbers. > > jas. > > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > Mged-mage mailing list > Mged-mage@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/mged-mage > > ------------------------------ End of microarray-ontol-digest V1 #29 *************************************