A new release of BioTop domain top ontology for Biomedicine
Stefan Schulz
a new OWL-DL version of the Biomedical Domain Top Ontology
** BioTop **
is available for download at
http://www.ifomis.uni-saarland.de/biotop/
If interested, subscribe to bio...@googlegroups.com
Kind regards,
Stefan Schulz, Freiburg University Hospital
Holger Stenzhorn, Freiburg University Hospital & IFOMIS Saarbrücken
Elena Beisswanger, Jena University Language & Information Engineering
(JULIE) Lab
Robert Hoehndorf
Hi,
SS> a new OWL-DL version of the Biomedical Domain Top Ontology
SS> ** BioTop **
SS> is available for download at
SS> http://www.ifomis.uni-saarland.de/biotop/
I have some questions and remarks.
You use the class Atom, with six specific subclass. Why choose these
six? Would it not make more sense to import a period table OWL-file at
this point, like http://www.daml.org/2003/01/periodictable/ ?
The same could be done for amino acids, importing
http://www.co-ode.org/ontologies/amino-acid/2005/10/11/amino-acid.owl
For Tissue, the definition may be extended. The cells within the
tissue are probably connected, and somehow have the same function.
Why do you need the hasGrain only EntireMolecularEntity in the
definition?
Having a Virus as a sublcass of Organism seems arguable, at least. I
am certain there are many biologists who do not agree with this
definition.
I do not like the fact that almost every class is modelled twice,
i.e., the whole hierarchy is mirrored under ObjectAggregate. It seems
unnecessary, because whenever needed, it is simple to define these
specific aggregates. And it increases classification time a lot.
What is an Object? And why are Atoms not a subclass of Object?
Biological process could need a definition. Text-definition is a
process specific for biological systems. Maybe introducing a
biological system could help in defining these processes.
CellInVivo is defined as subclass of Cell and subclass of (Organism or
hasProperPart some Organism). It should probably be properPartOf some
Organism.
My last remark is related to Sequences, but you should probably
consult Chris Mungall for more specific information. Defining it as
something which consists of some and only Deoxyribonucleotides seems
to ignore the "sequence" of the sequence. After all, it could be a
collection of these molecules that satisfies this definition. I also
believe that the fact that a sequence is a subclass of a molecule is
not agreed upon. There are biological sequences without
instances. There may be properties that the sequences have, that
cannot be derived from the molecules alone. There may be two different
molecules that have the same sequence.
I think that with a lot of care all these properties of sequences may
at some point be reduced to properties of molecules. But it requires
some effort, esp. for a bioinformatician, who is used to work with
abstract sequences, to make all this explicit.
Rob.
Stefan Schulz
thanks a lot for you for your valuable comments. Here my (short) answers:
> I have some questions and remarks.
> You use the class Atom, with six specific subclass. Why choose these
> six? Would it not make more sense to import a period table OWL-file at
> this point, like http://www.daml.org/2003/01/periodictable/ ?
> The same could be done for amino acids, importing
> http://www.co-ode.org/ontologies/amino-acid/2005/10/11/amino-acid.owl
I choose these six subclasses of atoms because they (H, O, C, N, S, P)
are crucial for further define the main classes of biomolecules. We
are aware of the fact that BioTop overlaps with existing ontologies
especially in the molecular entity branch. In the future this branch
could be separated from BioTop and become the top level of a chemical
ontology.
> For Tissue, the definition may be extended. The cells within the
> tissue are probably connected, and somehow have the same function.
> Why do you need the hasGrain only EntireMolecularEntity in the
> definition?
The description of Tissue is provisional, and it is really a major
enterprise to define what a tissue exactly is. I doubt whether we ever
can agree upon a full definition.
The current description is restricted to a couple of properties all
instances of tissue have in common, such as "hasGrain SOME Cell" and
"hasGrain SOME EntireMolecularEntity" (not ONLY), which means that all
instances of tissue consists of many molecules and many cells.
I do not say anything about connection (so far I do not use this
relation since there is no univocal definition about connection, and
even the notion of connection in the textual deinition of BFO classes
is a bit problematic).
> Having a Virus as a sublcass of Organism seems arguable, at least. I
> am certain there are many biologists who do not agree with this
> definition.
If there is no consensus among biologists about the status of viruses,
the ontologists fight a losing battle...
>
> I do not like the fact that almost every class is modelled twice,
> i.e., the whole hierarchy is mirrored under ObjectAggregate. It seems
> unnecessary, because whenever needed, it is simple to define these
> specific aggregates. And it increases classification time a lot.
We did this to highlight the phenomenon that most occurences of
biological terms in domain specific texts do indeed relate to a
plurality of single objects. Rarely there is one single molecule
adressed. So if, for instance, "Glucose" appears in a text, a
plurality (aggregation) of molecules is meant and not a single one.
> What is an Object? And why are Atoms not a subclass of Object?
It was so in a previous version, but then some BFO discussions arose
(see BFO list) which challenged the fact that the distinction between
the (disjoint) classes Object and FiatObjectPart can be really
maintained across all levels of granularity (the same applies to the
distinction between Object and Object Aggregate).
> Biological process could need a definition. Text-definition is a
> process specific for biological systems. Maybe introducing a
> biological system could help in defining these processes.
The occurrents section is still highly underdeveloped and we welcome
any suggestion.
> CellInVivo is defined as subclass of Cell and subclass of (Organism or
> hasProperPart some Organism). It should probably be properPartOf some
> Organism.
Yes. Thank you ! I fixed it.
>
> My last remark is related to Sequences, but you should probably
> consult Chris Mungall for more specific information. Defining it as
> something which consists of some and only Deoxyribonucleotides seems
> to ignore the "sequence" of the sequence. After all, it could be a
> collection of these molecules that satisfies this definition.
In this case I would have used the relation has-grain. It is again the
notion of connection which comes into the play.
> I also
> believe that the fact that a sequence is a subclass of a molecule is
> not agreed upon. There are biological sequences without
> instances.
> There may be properties that the sequences have, that
> cannot be derived from the molecules alone. There may be two different
> molecules that have the same sequence.
> I think that with a lot of care all these properties of sequences may
> at some point be reduced to properties of molecules. But it requires
> some effort, esp. for a bioinformatician, who is used to work with
> abstract sequences, to make all this explicit.
You adress an important issue, i.e. whether or not to consider certain
entities as "bauplans" rather than material entities.
Buit in this case by which relation should we relate a such a
"bauplan" to the entity that materializes it?
Best regards
Stefan