Re: scientific publishing task force update

Here, here!

I think Matthias is making a very important point here - one equally  
important to efforts to define biological reality empirically from  
the ground up (semantic web and/or computational linguistic/NLP  
approaches to distilling KR from the literature-base), as it is to  
top-down, highly ontology-centric approaches.

Ultimately, what both efforts are about is creating a formal,  
computable description of reality - biomedical reality when it comes  
to the range of problems being addressed from biomolecular  
informatics on through clinically-oriented medical informatics  
projects.  This computability includes implementations of what I  
would call the "low hanging fruit" of query resolution and data  
integration (highly intertwined problems), as well as more complex  
attempts to create a framework for biomedical reality to support wide- 
field, reasoning systems.  I don't consider any problems across this  
spectrum either more or less useful - or more or less ambitious, it  
just the data integration & query resolution tasks due to their  
requiring slightly less ontological rigor are being addressed now on  
a large scale, whereas the reasoning applications right now to  
maintain a narrow and explicit focus to be effective.

The issue of having a solid ontological foundation to work from and  
the related issue of having a well defined, community-wide collection  
of ontological relations (e.g., the OBO Relation ontology - http:// 
genomebiology.com/2005/6/5/R46) is quite critical to efforts aimed at  
constructing wide-field KR - for whatever purpose.

I completely agree with Philip's statement regarding the effect of  
working with a foundational ontology on biomedical KR efforts:

On Jun 13, 2006, at 5:39 AM, Phillip Lord wrote:
> they tend to complicate some stages of ontology development, mostly  
> notably the first month when you have lots of biologists tearing  
> their hair out trying to work out what a perjurant, continuant,  
> sortal, self-standing kind is.

To my mind, I'm not certain it is either necessary or appropriate to  
drag the research biologist through this process - unless of course  
they want to go through it.  I think at this late date, the field -  
if you will - of biomedical KR has matured to the point where - as in  
bioinformatic programming it is no longer a de facto expectation the  
researcher will need to be an autodidact computer scientist (there  
are degree granting programs targeting the mix of C.S. & biology  
required to be a good bioinformatician) - there is a cadre of  
biologists emerging who have both a penchant and a talent for working  
in the KR realm.  Those folks do need to drag themselves through the  
difficult slog of becoming Natural Philosophers.  These folks can act  
as both effective KR researchers and as
  emissaries to the larger community of research biologists to ensure  
the KR frameworks developed meet the Use Case-defined computational  
needs of the field, as well as helping the researcher "donate" their  
knowledge to the growing KR framework.

I also agree with Philip's statement regarding the usefulness of  
foundational ontologies:

On Jun 13, 2006, at 5:39 AM, Phillip Lord wrote:
> they help to ease the design of an ontology; you have more idea  
> where concepts should go, so you can spend more time worrying about  
> the details of what ever you are
> modeling and less about the big picture.

Here I would agree with the comments made by Matthias, though I would  
take particular issue with the term "modeling" as it can trick some  
people into thinking the act of creating object models (a la OOD  
software design and/or use of UML formalisms or or XML to express the  
models) has anything necessarily to do with KR expressions of  
biological reality.  Modeling and KR tasks can share many goals in  
common, but models tend to be tied to the application domain for  
which they were intended and have no de facto requirement to  
represent reality.  It of course is a good idea to be certain the  
ontological development you do stays in sync with/compatible with the  
relevant data models in the domains you expect to map via ontologies  
- e.g., in the neuroimaging domain, you would want to stay compatible  
with data models and associated formats such as DICOM & NifTi.   
However practical a model may intend to be, however, from a  
philosophical perspective they are much closer to Kant than they are  
to Husserl.  ;-)

Where I would have to disagree with Philip is when he states:

On Jun 13, 2006, at 5:39 AM, Phillip Lord wrote:
> It's not clear that an upper ontology actually brings significant  
> value to the table. The claimed advantage of interoperability  
> between ontologies is, to my mind, somewhat bogus; they only really  
> allow interoperability when you are querying over the concepts in  
> the upper ontology.

Here I would agree with the sentiment I believe both Robert and  
Matthias have expressed.  There are practical aspects of putting a  
(or THE) foundational ontology (and foundational relations) in place  
that can make a very big difference in the computability of the  
resulting knowledge maps (aka association files, annotations, etc.).   
On the BIRN project, our work toward creating a KR framework for all  
of the neuroscience data we need to accommodate it has proven  
extremely liberating and effective to fix on the BFO.  I believe same  
is also true for both the Gene Ontology curation effort, as well as  
the work on FuGO and PATO (the latter two being two ontologies we are  
investing heavily in on the BIRN KR efforts).  I wouldn't want to put  
words in the mouths of those much more experienced contributors to  
GO, FuGO, or PATO (please chime in if you have corrections or  
qualifications on this issue), and it certainly is true fixing on a  
foundational ontology can cause a lot of grief to curators,  
programmers, and researcher-users of the knowledge resource, but I'm  
pretty certain the long-term gains of doing so will be significant.

Some may think this has little to do with the more bottom-up approach  
to KR implicit in Semantic Web projects.  I really don't believe this  
to be the case.  Though I completely agree SW approaches to KR will  
help to provide a more dynamic and fine-grained accurate  
representation of biological knowledge, I would maintain the  
resulting triplet repositories will have much more longevity and  
applicability to the field, if they are constructed in such as way  
that promotes a convergence between the top-down and bottom-up  
approach.  Each approach can - and should - inform the KR framework  
constructed by the other.

By way of examples relevant to the issue of neuroscience data KR, as  
I mentioned, we on the BIRN Ontology Task force have begun to  
converge on use of BFO (and the OBO Relation ontology) not only for  
use in the ontological efforts originating from BIRN, but also in our  
selection of external ontologies we draw on (e.g., Neuro-FMA, FuGO,  
PATO, GO - not that all of these are yet fully BFO compliant).

There are also neuro-oriented KR projects making productive use of  
DOLCE and SKOS.  Matthias's Semantic Synapse project, for instance,  
uses both.

Just my $0.02

Cheers,
Bill




On Jun 13, 2006, at 8:29 AM, Matthias Samwald wrote:

>
>
>>  One small, but significant, dislike of the bio-ontology community
>>  for SUMO (as used by Solditova and King) is that it isn't really
>>  only an upper level. It strays into, for instance, stating a
>>  protein is a foodstuff. this, as you might suppose, causes
>>  biologists to laugh.
>
> That is very true, and I think that the importance of having huge  
> top-level ontologies like SUMO or maybe Cyc is largely overrated.
> On the other hand, having very small and basic foundational  
> ontologies (e.g. the most basic ontologies of the DOLCE lite  
> ontology, BFO or SKOS) is more important than most developers of  
> ontologies seem to think. It is a great aid to the development of  
> interoperable ontologies to have a common, basic framework of  
> classes (e.g. physical-object, perdurant, quality) and properties  
> (e.g. part-of, participant-in).
> These basic ontologies do not need to be large or complicated to be  
> useful (around 20 classes and properties are sufficient, I guess).  
> Quite to the contrary, making these foundational ontologies too  
> complicated would significantly decrease their usefulness.
>
>
> //Matthias Samwald
>
>

Bill Bug
Senior 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 - William.Bug@DrexelMed.edu







<p>This email and any accompanying attachments are confidential. This information is intended solely for the use of the individual to whom it is addressed. Any review, disclosure, copying, distribution, or use of this email communication by others is strictly prohibited. If you are not the intended recipient please notify us immediately by returning this message to the sender and delete all copies. Thank you for your cooperation.</p>

Received on Tuesday, 13 June 2006 13:55:27 UTC