I was going to try and pull some ideas together and send them as and
email; I'll do that, and as long as people aren't horrified by it (I
expect discussion) I'll set up a wiki page.
Matt
Eric Neumann wrote:
Bill,
Thanks for sending out the urls-- always good for a discussion
thread-group to have the same common references! It may be necessary
to identify not one, but a few definitions of evidence to be used by
different groups (e.g., researchers def vs. HC compliance forms--
Dirk's point)-- remember, its about namspaces and the ontological
structures associated with each!
On the subject of 'evidence' has anyone started such a esw-wiki page
for HCLS? There's enough good input from several people over the last
few days, that I hope someone is willing to distill the ideas, and
list them on such a page.
If no one is willing, I will try and do so, but I cannot guarantee it
will be done right away (i.e., people with more invested interest on
this topic might wish to start a page sooner and post its location to
the group)...
Eric
-----Original Message-----
From: [EMAIL PROTECTED] on behalf of William Bug
Sent: Wed 6/13/2007 9:02 AM
To: public-semweb-lifesci hcls
Cc: Waclaw Kusnierczyk; Barry Smith; Matthias Samwald
Subject: Re: Evidence
Sorry I've been out of touch on this and other HCLS IG activities,
but I've been - and will continue to be for some time - tied up with
other tasks.
I believe both issues as originally raised by Matthias aer extremely
important:
a) creating a cogent and concise means of inter-relating entities
that is - as best we can implement it - tied to a realist view of
biomedical reality
b) dealing in a consistent and - as much as is practical - formal
way with evidence - which includes dealing in a consistent manner
with "information" entities.
I think Vipul, Matt Williams, Chimezie, Daniel and others have all
raised important issues in regards to evidence. I would also cite
two active threads in the HCLS IG that have direct bearing on this
issue:
1) Again beating the old (maybe not quite dead) horse of the
experiment we began in BioONT back last September, I would cite the
following HCLS IG Wiki page:
http://esw.w3.org/topic/HCLS/OntologyTaskForce/
OboPhenotypeSyntaxExperiment
If the top of the page is familiar (or too dense), just
jump to
the section starting roughly 1/3 down the page entitled:
"The OBO Phenotype Syntax + PATO Quality way to represent
experimental observations/research statements/claims"
This "experiment" draws on a significant body of work both in
the GO/OBO community, as well as ongoing community ontology
development seeking to apply BFO to this issue of providing a
consistent and coherent representation of biological reality - most
especially - in this context - OBO-RO (http://www.obofoundry.org/
ro/), PATO (http://www.obofoundry.org/cgi-bin/detail.cgi?id=quality),
and OBI (http://www.obofoundry.org/cgi-bin/detail.cgi?id=obi)
2) SWAN
http://www.ncbi.nlm.nih.gov/sites/entrez?
Db=PubMed&Cmd=ShowDetailView&TermToSearch=17493287&ordinalpos=1&itool=En
trezSystem2.PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_RVDocSum
http://www.mind-informatics.org:8081/swan/
My sense is these two efforts are both very relevant to this
discussion. SWAN obviously encompasses a complete, functional system
currently in use by the AlzForum designed to describe hypotheses in
the context of "evidentiary" statements. The "experiment" Wiki page
takes a more constrained approach than SWAN to describing evidence
for "experimental assertions" drawing from the community biomedical
ontology efforts defined above (as well as other resources). I see
this approach and the SWAN approach as very much complimentary and
synergistic, each bearing their own advantages and disadvantages. In
this experiment, there are still many details to be worked through
more explicitly, some of which relate directly to this issue Matthias
raised initially (how and when should we reference RDBMS-based
records for bio-molecular entities). Still, there is much more there
beyond this single issue of citing RDBMS records - as is true in SWAN
- that addresses issues related to providing a formal framework for
"experimental evidentiary assertions". Note too that though the
example on this Wiki page draws from an existing publication (very
much a kin to the publication evidence used by GO annotators and
other informatics projects such as NeuronDB at Yale), the approach is
intended for use directly in annotating data repositories as well.
I would also note there is currently an ongoing discussion on the obo-
phenotype list of this very topic - i.e., how to reference a UniProt
record in a biomedical ontological framework - a thread Alan, and OBO
investigators have all been contributing to (see the "Phenote for
expression" thread at http://sourceforge.net/mailarchive/forum.php?
forum_name=obo-phenotype).
I agree with Matt W. and Adrian's suggestion we must consider the
extensive and long standing body of work related to "evidence-based
science". As Vipul and Daniel have both remarked, we must seek to
use such approaches in a manner that can accommodate the way evidence
is established in thin clinic. However, whether your requirement is
to include/exclude classified individuals, or doubt/question
interpretations of rules for deriving "evidence" from experimental or
clinical observation, it still will be necessary to provide a shared
(hopefully formal) definition of the relevant entities - e.g., in the
context of Karen's query such entities (considered in a BFO context)
as "Smoking Behavior", "Assessment for Smoking Behavior" - which may
include nominalized and qualified, numeric restrictions in the OWL
sense (which certainly can be used to represent the required
classification requirements). To then give a "name" to such sub-
types - as is done in when applying a diagnostic label to a specific
EKG waveform or blood sample data point ("high sodium"), can
certainly be done in OWL.
In regards to information entities, as Waclaw pointed out, there is
an ongoing collaboration between the BFO developers and BFO users/
developers such as those working on the OBI ontology to provide a
means to characterize such entities in a BFO context. As has been
mentioned, this is still a work-in-progress, and one in which we -
the HCLS IG - can actively participate.
Finally, to extend Daniel's radiological evidentiary statement
example, in the biomedical imaging domain (both in the clinic and in
research domain), often we are relying on algorithmic means to first
identify biologically-relevant objects in the digital images. These
algorithms also bring with them many caveats and assumptions, which
also need to be addressed when expressing this "evidence" in a formal
context. This latter issue is one we are seeking to address in the
BIRN project using BFO, OBO-RO, and OBI to establish as best we can a
formal means of expressing the experimental observations (both "raw"
and "derived") upon which one can build more complex assertions.
Cheers,
Bill
On Jun 12, 2007, at 3:53 PM, [EMAIL PROTECTED] wrote:
>
> Hi Waclaw,
>
>
>> Matthias, if you look carefully at BFO, you'll see that roles are
>> entities. This means that evidences, as roles, are entities.
>
> Of course. I just wanted to differentiate that an experiment is not
> an instance of any class called 'evidence' (in other words, an
> experiment 'is not' evidence). Instead, it should be associated
> with an 'evidence-role'.
>
> cheers,
> Matthias
>
> cheers,
> Matthias Samwald
>
> ----------
>
> Yale Center for Medical Informatics, New Haven /
> Section on Medical Expert and Knowledge-Based Systems, Vienna /
> http://neuroscientific.net
> --
> Psssst! Schon vom neuen GMX MultiMessenger gehört?
> Der kanns mit allen: http://www.gmx.net/de/go/multimessenger
>
Bill Bug
Senior Research Analyst/Ontological Engineer
Laboratory for Bioimaging & Anatomical Informatics
www.neuroterrain.org
Department of Neurobiology & Anatomy
Drexel University College of Medicine
2900 Queen Lane
Philadelphia, PA 19129
215 991 8430 (ph)
610 457 0443 (mobile)
215 843 9367 (fax)
Please Note: I now have a new email - [EMAIL PROTECTED]