Difference between revisions of "Workshop on Clinical Trial Ontology"

From NCBO Wiki
Jump to navigation Jump to search
Line 3: Line 3:
 
The National Center for Biomedical Ontology will host a two-day workshop focused on the Ontology of Clinical Trials. The workshop will take place in March 2007; venue to be determined.
 
The National Center for Biomedical Ontology will host a two-day workshop focused on the Ontology of Clinical Trials. The workshop will take place in March 2007; venue to be determined.
  
−
== Aim of the workshop  ==  
+
== Aim of the workshop  ==
 +
 
 +
Ida,
 +
 
 +
I think that the issue of models vs. ontologies is indeed important. Before we can address this issue sensibly, however, I think we need to have some idea of what a clinical trial ontology should be. My ideas (which are still inchoate) are as follows:
 +
 
 +
1. it should fulfil the three requirements you listed already in
 +
http://smi-web.stanford.edu/pubs/SMI_Reports/SMI-95-0584.pdf, which is to say:
 +
 
 +
(1) fully and faithfully capture trials of any experimental design
 +
(2) represent all the concepts [better: contain all the terms] needed for the task of meta-analysis
 +
(3) be in a form to be used for trial bank interoperation, as much as is possible
 +
 
 +
2. it should share an upper level with the FuGO (Functional Genomics Investigation Ontology), which is attempting to create an ontology whose terms can be used in the annotation of experiments of other, though related, sorts (some members of the FuGO Consortium, e.g. Susanna Sansone have an interest also in the clinical trial ontology). See http://fugo.sourceforge.net/
 +
 
 +
3. it should interoperate with a drug (trial) ontology (or perhaps include the latter as a proper part)
 +
 
 +
4. it should interoperate with an epidemiology study ontology
 +
 
 +
5. it should interoperate with a disease ontology and with the other biomedical ontologies currently being developed within the OBO Foundry
 +
 
 +
6. it should provide a formal representation of, and definitions for, terms like:
 +
 
 +
cohort, randomization, placebo, response, efficacy, control, protocol, etc., including also major relevant statistical terms.
 +
 
 +
My proposal would be that we focus on 6., since this is already a very considerable challenge. We might then create a small advance team to collect terms and create a preliminary test version. I hope that you will feel free to make alternative proposals, to maximize the usefulness of the end-result from your point of view.
 +
 
 +
With greetings
 +
Barry
 +
 
 +
PS Nota bene: I am no fan of HL7 (see http://ontology.buffalo.edu/HL7/doublestandards.pdf). Indeed I think HL7 V3 is in large part responsible for the $30 bn. cost overruns incurred by the still non-functioning UK national health information infrastructure announced just this week. 
 +
 
 +
At 01:11 AM 6/8/2006, you wrote:
 +
Barry,
 +
 
 +
What are the objectives you envision for this workshop? You may know that
 +
BRIDG is "comprehensive domain analysis model representing
 +
biomedical/clinical research", taking after HL7. We could probably spend the
 +
whole workshop on philosophical discussions of models vs. ontologies (I had
 +
wanted to ask you your views of that at Stanford but didn't get around to
 +
it). Defining the objectives and anticipated outcomes would be important for
 +
deciding who should/must attend.
 +
 
 +
Initial names:
 +
BRIDG: Doug Fridsma, Joyce Niland (maybe)
 +
CDISC: Becky Kush
 +
HL7: Lakshmi Grama
 +
caBIG: not sure exactly who would be right
 +
NIH Roadmap: Stuart Speedie (U. Minn), maybe others -- ask Carol Bean
 +
AMIA CT WG: Michael Kahn, Charlie Barr,
 +
AMIA KR WG: not sure who
 +
Columbia has a group on Clinical Trials Informatics that might be relevant
 +
 
 +
As for dates, towards the middle of May is probably best.
 +
 
 +
Ida
 +
 
 +
 
 +
On 6/4/06 6:43 AM, "Smith, Barry" <phismith@buffalo.edu> wrote:
 +
 
 +
> Dear Ida,
 +
> Thanks for this -- it is clear that there is basic agreement as to
 +
> the facts of the matter concerning 'instances': the problem is rather
 +
> terminological. But more on this later.
 +
>
 +
> Concerning clinical trials, do you have specific persons at BRIDG,
 +
> HL7, CDISC, and caBIG, who you think should be involved in this
 +
> meeting? And do you know anyone at the FDA and at the AMIA WG on
 +
> Clinical Trials (or other similar bodies) who might want to be involved?
 +
>
 +
> As to date, I am thinking of the second half of April or the first
 +
> half of May, and on the East coast. Do you have preferences within that
 +
> window?
 +
>
 +
> With greetings
 +
> Barry
 +
>
 +
> At 01:34 AM 6/3/2006, you wrote:
 +
>> Barry,
 +
>>
 +
>> A clinical trial ontology workshop would be very interesting. Other
 +
>> groups working in this area include BRIDG, HL7, CDISC, caBIG, etc as
 +
>> you probably know. It would be great to use such a meeting to work
 +
>> towards a first reference ontology for clinical trials. Do keep me
 +
>> posted on the meeting as it develops.
 +
>>
 +
>> I agree that 'instance' is used in all the ways you suggest. I think
 +
>> the main source of type-instance confusion in cBIO is in the way the
 +
>> bio and medical informatics communities use defintions 2 and 4. I
 +
>> think that in medical informatics, we tend to conflate #2 and #4
 +
>> because we're always thinking of data for a particular patient (or
 +
>> clinical trial) as the most specific information that needs to be
 +
>> captured. So, needing to capture eg Jane's particular oophorectomy,
 +
>> the type may be SURGICAL-INTERVENTION, the sub-type may be
 +
>> OOPHORECTOMY, and the instance JANE'S-OOPHORECTOMY. From the point
 +
>> of view of an ontologist of surgery, however, who doesn't deal with
 +
>> actual cases of oophorectomies, OOPHORECTOMY can well be an
 +
>> instance, with no need to consider or represent particular
 +
>> oophorectomies. In contrast, because many medical informaticians
 +
>> take care of patients, we tend to care about particular
 +
>> oophorectomies and consider those to be the 'instances'. This then
 +
>> causes us to draw the type-instance distinction at a different level
 +
>> of specificity than a pure ontologist of surgery would.
 +
>>
 +
>> In sum, I don't think the naming of meaning #4, or any of the
 +
>> meanings, is the problem, but rather a different conceptualization
 +
>> of how specific/particular a representation has to be. I hope that
 +
>> clarified rather than confused things?  And hard to do this by email!
 +
>>
 +
>> Do you have a tentative month/days for the workshop?
 +
>>
 +
>> Best,
 +
>> Ida
 +
>>
 +
>>
 +
>> On 5/27/06 2:54 AM, "Smith, Barry" <phismith@buffalo.edu> wrote:
 +
>>
 +
>> Dear Ida,
 +
>> I plan to organize a clinical trials ontology workshop in 2007, as
 +
>> one of the Center's dissemination workshops, and I am hoping very
 +
>> much that you will be involved. Here, to set us going, is my
 +
>> thinking on types vs. instances:
 +
>> The term 'instance' is used in different ways by different communities.
 +
>> 1. In the database world it means, officially, a complete database
 +
>> environment, including software, table structure, etc.; used when
 +
>> administrators need to describe multiple instances of the same database type.
 +
>> 2. However, the term 'instance' is also used loosely in the
 +
>> information modelling world (e.g. by HL7), to mean the data in a
 +
>> database about a specific case or particular (the data about  Jane,
 +
>> the data about Jane's oophorectomy).
 +
>>
 +
>> 3. Sometimes ontology people use 'instance' to mean 'leaf node' in an is
 +
>> 4. In the philosophical ontology world, however, and increasingly in
 +
>> the GO/OBO world, 'instance' is used to mean: a particular
 +
>> (individual, spatiotemporal, normally observable) entity which
 +
>> instantiates a given type. Thus a particular organism, cell, tumor,
 +
>> pill, surgical intervention, ....
 +
>> Did I miss anything out?
 +
>> I think we need a term to cover meaning 4., though I would be happy
 +
>> to substitute another term for this if one is suggested.
 +
>> With greetings
 +
>> Barry
 +
>> At 10:58 PM 5/26/2006, Ida Sim wrote:
 +
>> Chris,
 +
>> Thanks for this. It's great to see the trial bank schema and instance data
 +
>> imported into OBD as a first step.
 +
>> We are indeed interested in a reference ontology for clinical trials but I
 +
>> don't think we've resolved the issue of types vs. instances. This requires
 +
>> much more discussion.
 +
>> Simona Carini has just joined our project and will work with UVic to push
 +
>> the visualization project further along. That will start to provide concrete
 +
>> uses for OBD and we can then work towards some more implementation
 +
>> questions.
 +
>> Best,
 +
>> Ida
 +
>> On 5/24/06 5:14 PM, "chris mungall" <cjm@fruitfly.org> wrote:
 +
>>>
 +
>>> Sorry for the lack of subject line in the previous email! I hate it
 +
>>> when people do that. Here it is again, with subject line.
 +
>>>
 +
>>> On May 24, 2006, at 5:06 PM, chris mungall wrote:
 +
>>>
 +
>>>> I've added a new wiki page:
 +
>>>>
 +
>> <http://smi.stanford.edu/projects/cbio/mwiki-internal/index.php/>http://smi.s
 +
>> tanford.edu/projects/cbio/mwiki-internal/index.php/
 +
>>
 +
>>>> Clinical_Trials_in_OBD
 +
>>>>
 +
>>>> This is reachable from the OBD wiki page
 +
>>>>
 +
>>>> The page contains a link to the demo and a discussion of where to go
 +
>>>> next for representing clinical trials in OBD. In particular OBD, will
 +
>>>> require that terms are represented as types and not as instances,
 +
>>>> which will require a shift away from the current trialbank
 +
>>>> representation. From Ida's slides at the last all hands, it seems
 +
>>>> that trialbank may be moving in this way anyway, towards a reference
 +
>>>> ontology for clinical trials?
 +
>>>>
 +
>>>> Cheers
 +
>>>> Chris
 +
>>>> --++**==--++**==--++**==--++**==--++**==--++**==--++**==
 +
>>>> cbio-developers mailing list
 +
>>>> cbio-developers@lists.stanford.edu
 +
>>>> https://mailman.stanford.edu/mailman/listinfo/cbio-developers
 +
>>>>
 +
>>>
 +
>
 +
>
 +
Ref: ontology for clinical trial
 +
 
 +
 
 +
Ricardo Pietrobon co-chairs the MSI Ontology Working Group that represents the metabolomics domain in FuGO.
 +
 
 +
 
 +
Ricardo is Assistant Professor at Duke University and his current work is primarily in clinical and translational research. He uses an ontology to encode randomized trials.
 +
 
 +
 
 +
In particular, his group seeks to expand existing ontologies, by including 'quantitative terms' (e.g. such as risk ratio,
 +
confidence intervals, variables) , so that these would work with statistical packages such as R/bioconductor.
 +
 
 +
 
 +
Ricardo has also been in contact with Ida Sim.
 +
 
 +
 
 +
Therefore, also as MSI Ontology Working Group, we will be very interested to know more about the plans of this ontology for clinical trial and contribute where possible.
  
 
== Agenda ==
 
== Agenda ==

Revision as of 06:10, 19 June 2006

General Information

The National Center for Biomedical Ontology will host a two-day workshop focused on the Ontology of Clinical Trials. The workshop will take place in March 2007; venue to be determined.

Aim of the workshop

Ida,

I think that the issue of models vs. ontologies is indeed important. Before we can address this issue sensibly, however, I think we need to have some idea of what a clinical trial ontology should be. My ideas (which are still inchoate) are as follows:

1. it should fulfil the three requirements you listed already in http://smi-web.stanford.edu/pubs/SMI_Reports/SMI-95-0584.pdf, which is to say:

(1) fully and faithfully capture trials of any experimental design (2) represent all the concepts [better: contain all the terms] needed for the task of meta-analysis (3) be in a form to be used for trial bank interoperation, as much as is possible

2. it should share an upper level with the FuGO (Functional Genomics Investigation Ontology), which is attempting to create an ontology whose terms can be used in the annotation of experiments of other, though related, sorts (some members of the FuGO Consortium, e.g. Susanna Sansone have an interest also in the clinical trial ontology). See http://fugo.sourceforge.net/

3. it should interoperate with a drug (trial) ontology (or perhaps include the latter as a proper part)

4. it should interoperate with an epidemiology study ontology

5. it should interoperate with a disease ontology and with the other biomedical ontologies currently being developed within the OBO Foundry

6. it should provide a formal representation of, and definitions for, terms like:

cohort, randomization, placebo, response, efficacy, control, protocol, etc., including also major relevant statistical terms.

My proposal would be that we focus on 6., since this is already a very considerable challenge. We might then create a small advance team to collect terms and create a preliminary test version. I hope that you will feel free to make alternative proposals, to maximize the usefulness of the end-result from your point of view.

With greetings Barry

PS Nota bene: I am no fan of HL7 (see http://ontology.buffalo.edu/HL7/doublestandards.pdf). Indeed I think HL7 V3 is in large part responsible for the $30 bn. cost overruns incurred by the still non-functioning UK national health information infrastructure announced just this week.

At 01:11 AM 6/8/2006, you wrote: Barry,

What are the objectives you envision for this workshop? You may know that BRIDG is "comprehensive domain analysis model representing biomedical/clinical research", taking after HL7. We could probably spend the whole workshop on philosophical discussions of models vs. ontologies (I had wanted to ask you your views of that at Stanford but didn't get around to it). Defining the objectives and anticipated outcomes would be important for deciding who should/must attend.

Initial names: BRIDG: Doug Fridsma, Joyce Niland (maybe) CDISC: Becky Kush HL7: Lakshmi Grama caBIG: not sure exactly who would be right NIH Roadmap: Stuart Speedie (U. Minn), maybe others -- ask Carol Bean AMIA CT WG: Michael Kahn, Charlie Barr, AMIA KR WG: not sure who Columbia has a group on Clinical Trials Informatics that might be relevant

As for dates, towards the middle of May is probably best.

Ida


On 6/4/06 6:43 AM, "Smith, Barry" <phismith@buffalo.edu> wrote:

> Dear Ida, > Thanks for this -- it is clear that there is basic agreement as to > the facts of the matter concerning 'instances': the problem is rather > terminological. But more on this later. > > Concerning clinical trials, do you have specific persons at BRIDG, > HL7, CDISC, and caBIG, who you think should be involved in this > meeting? And do you know anyone at the FDA and at the AMIA WG on > Clinical Trials (or other similar bodies) who might want to be involved? > > As to date, I am thinking of the second half of April or the first > half of May, and on the East coast. Do you have preferences within that > window? > > With greetings > Barry > > At 01:34 AM 6/3/2006, you wrote: >> Barry, >> >> A clinical trial ontology workshop would be very interesting. Other >> groups working in this area include BRIDG, HL7, CDISC, caBIG, etc as >> you probably know. It would be great to use such a meeting to work >> towards a first reference ontology for clinical trials. Do keep me >> posted on the meeting as it develops. >> >> I agree that 'instance' is used in all the ways you suggest. I think >> the main source of type-instance confusion in cBIO is in the way the >> bio and medical informatics communities use defintions 2 and 4. I >> think that in medical informatics, we tend to conflate #2 and #4 >> because we're always thinking of data for a particular patient (or >> clinical trial) as the most specific information that needs to be >> captured. So, needing to capture eg Jane's particular oophorectomy, >> the type may be SURGICAL-INTERVENTION, the sub-type may be >> OOPHORECTOMY, and the instance JANE'S-OOPHORECTOMY. From the point >> of view of an ontologist of surgery, however, who doesn't deal with >> actual cases of oophorectomies, OOPHORECTOMY can well be an >> instance, with no need to consider or represent particular >> oophorectomies. In contrast, because many medical informaticians >> take care of patients, we tend to care about particular >> oophorectomies and consider those to be the 'instances'. This then >> causes us to draw the type-instance distinction at a different level >> of specificity than a pure ontologist of surgery would. >> >> In sum, I don't think the naming of meaning #4, or any of the >> meanings, is the problem, but rather a different conceptualization >> of how specific/particular a representation has to be. I hope that >> clarified rather than confused things? And hard to do this by email! >> >> Do you have a tentative month/days for the workshop? >> >> Best, >> Ida >> >> >> On 5/27/06 2:54 AM, "Smith, Barry" <phismith@buffalo.edu> wrote: >> >> Dear Ida, >> I plan to organize a clinical trials ontology workshop in 2007, as >> one of the Center's dissemination workshops, and I am hoping very >> much that you will be involved. Here, to set us going, is my >> thinking on types vs. instances: >> The term 'instance' is used in different ways by different communities. >> 1. In the database world it means, officially, a complete database >> environment, including software, table structure, etc.; used when >> administrators need to describe multiple instances of the same database type. >> 2. However, the term 'instance' is also used loosely in the >> information modelling world (e.g. by HL7), to mean the data in a >> database about a specific case or particular (the data about Jane, >> the data about Jane's oophorectomy). >> >> 3. Sometimes ontology people use 'instance' to mean 'leaf node' in an is >> 4. In the philosophical ontology world, however, and increasingly in >> the GO/OBO world, 'instance' is used to mean: a particular >> (individual, spatiotemporal, normally observable) entity which >> instantiates a given type. Thus a particular organism, cell, tumor, >> pill, surgical intervention, .... >> Did I miss anything out? >> I think we need a term to cover meaning 4., though I would be happy >> to substitute another term for this if one is suggested. >> With greetings >> Barry >> At 10:58 PM 5/26/2006, Ida Sim wrote: >> Chris, >> Thanks for this. It's great to see the trial bank schema and instance data >> imported into OBD as a first step. >> We are indeed interested in a reference ontology for clinical trials but I >> don't think we've resolved the issue of types vs. instances. This requires >> much more discussion. >> Simona Carini has just joined our project and will work with UVic to push >> the visualization project further along. That will start to provide concrete >> uses for OBD and we can then work towards some more implementation >> questions. >> Best, >> Ida >> On 5/24/06 5:14 PM, "chris mungall" <cjm@fruitfly.org> wrote: >>> >>> Sorry for the lack of subject line in the previous email! I hate it >>> when people do that. Here it is again, with subject line. >>> >>> On May 24, 2006, at 5:06 PM, chris mungall wrote: >>> >>>> I've added a new wiki page: >>>> >> <http://smi.stanford.edu/projects/cbio/mwiki-internal/index.php/>http://smi.s >> tanford.edu/projects/cbio/mwiki-internal/index.php/ >> >>>> Clinical_Trials_in_OBD >>>> >>>> This is reachable from the OBD wiki page >>>> >>>> The page contains a link to the demo and a discussion of where to go >>>> next for representing clinical trials in OBD. In particular OBD, will >>>> require that terms are represented as types and not as instances, >>>> which will require a shift away from the current trialbank >>>> representation. From Ida's slides at the last all hands, it seems >>>> that trialbank may be moving in this way anyway, towards a reference >>>> ontology for clinical trials? >>>> >>>> Cheers >>>> Chris >>>> --++**==--++**==--++**==--++**==--++**==--++**==--++**== >>>> cbio-developers mailing list >>>> cbio-developers@lists.stanford.edu >>>> https://mailman.stanford.edu/mailman/listinfo/cbio-developers >>>> >>> > > Ref: ontology for clinical trial


Ricardo Pietrobon co-chairs the MSI Ontology Working Group that represents the metabolomics domain in FuGO.


Ricardo is Assistant Professor at Duke University and his current work is primarily in clinical and translational research. He uses an ontology to encode randomized trials.


In particular, his group seeks to expand existing ontologies, by including 'quantitative terms' (e.g. such as risk ratio, confidence intervals, variables) , so that these would work with statistical packages such as R/bioconductor.


Ricardo has also been in contact with Ida Sim.


Therefore, also as MSI Ontology Working Group, we will be very interested to know more about the plans of this ontology for clinical trial and contribute where possible.

Agenda

Participation

Requests for participation are welcome. Please send a brief statement of your expertise to Barry Smith.

Participants (First Tentative List]

Russ Altman -- Stanford University

Ted Grasela -- Cognigen Corporation, Amherst, NY

Mark Musen -- NCBO, Stanford University

Fabian Neuhaus -– NCBO, University at Buffalo

Nigam Shah -– NCBO, Stanford Medical Informatics

Susanna Sansone -- EBI, Hinxton, Cambridge

Ida Sim -- NCBO, University of San Francisco Medical Center

Barry Smith -– NCBO, University at Buffalo

Venue

Retrieved from "https://www.bioontology.org//mediawiki/index.php?title=Workshop_on_Clinical_Trial_Ontology&oldid=3559"