Re: Antwort: RE: Semantic web article in Nature Biotechnology

Hi Melissa and All,
  Thanks for clarifying the "over-arching layer" approach.  From what you 
said definitely the RDF is better than XML.

This means that in SW the following statements are correct?:
1. SW connects ontologies through an over-arching layer
2. The over-arching layers do not change with the change of the underlying 
ontologies.
3. Several layers of over-arching layers could exist for aggregating sets of 
ontologies.
4. An over-arching layer does not contain any details about the sub-domain 
ontologies it is aggregating.
5. A DAG (Directed Acyclic Graph) is used to connect ontolgies in the 
over-arching layer.

IF the above statements are correct maybe the following need clarifications:
 I don't know how the over-arching layer connects two DAG's without any 
details, but through whatever method this results in no integration - only 
aggregation takes place - is this correct??

For example, if I have one ontology has protein structures and another 
ontology has protein activity.  After the over-arching layer connects both 
ontologies will one be able to ask the simple question: "which structure 
belongs to which activity"?.
>From what I am finding from the other emails -- so maybe it is not only 
me -- that it is not clear how the over-arching layer that contains no 
detail information and is not affected by changes in the underlying 
ontologies and can still answer the questions that allow us to navigate 
across ontology boundaries without knowing the other ontology layout.  I am 
assuming that the goal of the over-arching ontology is creating the link 
between ontologies allowing to share.  i.e. biologist and chemist can 
connect.

As you know the relational database model is also based on DAG to connect 
relations.  SW must have a novel unique benefit in using DAG, and how would 
that help the end cause of connecting ontologies to ask the good questions. 
DAG is used to ensure integrity and consistency when joining tables and it 
is used at the table level in the relational model.

As Xiaoshu mentioned, it seems that this is no easier than the schema 
integration as in the relational approach - so far I agree with that.  Hence 
I agree that we need clarity on how linking of ontologies work for all to 
benefit.

Jeremy mentioned that "BioPAX primary goal is to integrate overlapping 
schemas of all the metabolic, molecular interaction, and signaling pathway 
databases into one merged upper level ontology"
Yes, great we all need that but how?  How will it over come changes in the 
underlying ontologies.   I think we all agree to the goal -- the question is 
how and if known state it!!  so everyone in the community benefits.

In the approach brought up by Peter Mork: 
http://genomebiology.com/2005/6/5/R46
This group seem to work on defining a detail to set a standard of primitive 
relations like "is_a" and "part_of".  Will this work if a biologist want to 
add a new primitive relation like "activates_a" in a protein cell as a 
relation?  The approach is about standardizing the ontology language and not 
using an over-arching ontology to link different ontologies.  Once more how 
will the aggregated ontologies react to a change in one of the ontologies.

As Helen Chen mentioned that the monolithic problem might be with us for a 
long time -- which I might tend to agree. This is no different than the many 
databases we have which still don't talk to each other except after the 
painful, costly, and time consuming integration effort.   The hope is SW 
will be able to link them -- I guess how is still not clear to me.

I go back to Eric Neumann's comment, that we still need to understand how to 
apply some of the biological and biomedical ontologies and link them which 
will bring greatest value.

Thanks
-Wafik


----- Original Message ----- 
From: "Melissa Cline" <cline@pasteur.fr>
To: "Wafik Farag" <Wafik@Farag.net>
Cc: <helen.chen@agfa.com>; <public-semweb-lifesci@w3.org>; 
<public-semweb-lifesci-request@w3.org>
Sent: Wednesday, September 28, 2005 9:45 AM
Subject: Re: Antwort: RE: Semantic web article in Nature Biotechnology


>
> Hi Wafik,
>
> One beauty of SW is that the "over-arching layer" doesn't really need to
> know any details about the sub-domain ontologies.  This is the classic
> contrast between XML and RDF: XML data is tree-structured, while RDF data
> is structured as a DAG.  If you combine two different DAGs, the result is
> still a DAG.  But if you combine two trees, the result is not necessarily
> a tree: if you join on the leaves, you end up with a structure with two
> roots.  To turn this into a tree again, you have to go through "XML hell"
> and rebuild the whole structure.
>
> The more subtle beauty of SW is that it helps foster over-arching layers
> that aren't big and cumbersome.  If data aggregation is difficult, then
> data tends to be kept in "monster tables": full of every quantity you
> might want, and several you didn't. Except, there's always something else
> that want, so you end up joining with another monster table to get it.
> The result: a real monster.  And worse, the data may be inconsistent or
> outdated, because when data aggregation is difficult, it tends to be done
> infrequently, with the results kept for a long time.  SW helps avoid this
> through a framework flexible enough to contain combined ontologies (RDF),
> and mechanisms such as LSIDs to serve as unique global identifiers.  So it
> becomes easier for us to keep the data distributed, and in small,
> lightweight ontologies that we can combine when we need.
>
> Cheers,
>
> Melissa
>
>
>> Hi Helen,
>>   I thought the power of the SW is linking those monolithic fragmented
>> domains or sub-domains into one with no over-arching layer.  Maybe I
>> am wrong -- please someone correct me here.  The idea of creating
>> another over arching layer to connect the sub-domains might seem
>> defeating the purpose.  How will this over arching layer react to any
>> change in any of the underlying sub-domains -- will it need to be
>> changed/adjusted every time one of the underlying sub-domains is
>> changed???
>>
>> I think the power of the SW is in linking ontologies -- creating RDF
>> statements in each ontology to establish a link. Best -Wafik Farag
>>
>>
>>   ----- Original Message ----- 
>>   From: helen.chen@agfa.com
>>   To: Phillip.Lord@newcastle.ac.uk ; wangxiao@musc.edu
>>   Cc: public-semweb-lifesci@w3.org ; public-semweb-lifesci-request@w3.org 
>> ; wangxiao@musc.edu
>>   Sent: Tuesday, September 27, 2005 12:02 PM
>>   Subject: RE: Antwort: RE: Semantic web article in Nature Biotechnology
>>
>>
>>
>>   Hi, Phil and Xiaoshu
>>
>>
>>
>>   >> (3) All these ontologies are still developed in a monolithic manner.
>>   >> No consideration about the granuality and practicality has been
>>   >> emphasized.
>>
>>   >There have been many considerations of practicality made. Monolithic
>>   >development is a problem, but both OBO and the MGED ontologies have
>>   >been attemping to ensure consistency between ontologies as well as
>>   >minimise overlap.
>>
>>   Even with many considerations and increasing community efforts in 
>> developing ontologies to be consumed by semantic web, I am afraid the 
>> "monolithic development" problem will be with us for a long time. 
>> Ontologies developed within different specialties and sub-domains will 
>> always be fragmented and rather "monolithic" in their own right.
>>
>>   In Healthcare domain, different regulatory bodies may develop 
>> ontologies for their practice guidelines, and disease management centers 
>> develop their own care plans and protocols.   It is not realistic to hope 
>> for a well-coordinated ontology that covers everything nicely under the 
>> hood.
>>
>>   What I understand of the power of semantic web technology lays the 
>> connecting and inference capability between those "fragmented" knowledge 
>> bases.  This connection is to be reached by a thin layer of 
>> "over-arching" ontology and a set of basic rules. We have limited 
>> experience in linking (mapping) our rather "monolithically developed" 
>> RPGOntology (ontology for EU-radiation protection guideline) with SNOMED 
>> CT(http://www.snomed.org/snomedct/).  The benefit of such connection can 
>> not be over-stated.
>>
>>
>>   Helen
>>
>>
>>
>>   ---------------------------------------------------------------------------------------------
>>   Helen Chen, Ph.D
>>   Research and Innovation Center for Healthcare
>>   Global Architecture and Design Group
>>   Agfa Healthcare
>>
>>   http://www.agfa.com/healthcare
>>
>>
>>
>>
>
> 

Received on Wednesday, 28 September 2005 19:08:52 UTC