First attempt of converting data to BioTop, some remarks.
Matthias Samwald
I have made a first attempt at converting existing biomedical data to
the BioTop ontology format. I used code for the conversion of BioPAX
to bio-zen, and adapted it as a converter from BioPAX to BioTop. The
data is taken from the Reactome database, and there are several other
interaction databases that can be converted in a similar manner. It
can be downloaded from
http://neuroscientific.net/res/biotop/reactome-in-biotop-example.zip
The data is represented as individuals, mainly as 'aggregate' classes,
because I think molecular pathways are more accurately described on
the level of populations/aggregates. I have not thoroughly checked
ontological consistency yet. Only a part of the Reactome database was
converted.
Some remarks:
The current version of biotop lacks classes for biological occurents
like "binding", "conversion" and so on. Therefore, I have made all
processes individuals belonging to the generic "biological process"
class. Users of Protégé 3.x should note that the participants of
biological processes and part-of relations are not shown in the GUI,
because the BFO/RO still lack domain and range restrictions on
properties. I see that this has been discussed in another thread in
the BioTop discussion group, as well as the BFO discussion groups.
Will subclasses of "biological process" be added soon? I could make
some suggestions based on BioPAX and bio-zen, if you would be
interested.
Will subclasses or individuals be added to "MolecularFunction" and
"ProteinFunction", or does such an extension already exist? Will the
'concepts' from the Gene Ontology be represented as classes or
individuals?
Do you plan on adding constructs (or import constructs from other
ontologies) for the representation of data like protein sequences,
molecular weights, reaction equilibria etc.? How would you advise that
things like bibliographical annotations or references to database
entries should be represented? Maybe OBI could be useful for this when
it matures?
biotop:AminoAcidMonomer should probably be a subclass of
biotop:Monomer
Why are classes for molecules etc. not subclasses of "Object"?
If these classes are made subclasses of "Object", I think
"ObjectAggregate" should have a restriction like "hasGrain some
Object".
The "aggregate" classes are not structured in class/subclass relations
like their grain counterparts are. I think this should be done, if it
does not collide with any formal restrictions of BFO.
The only direct link to the biotop OWL file points to a SVN, so the
contents of this file are probably subject to ongoing change. It might
be useful to have a stable URL for importing each version of biotop.
Do you have any suggestions for further developments and improvement?
Cheers,
Matthias Samwald
Stefan Schulz
nice to meet you in the BioTop list.
>
> The current version of biotop lacks classes for biological occurents
> like "binding", "conversion" and so on. Therefore, I have made all
> processes individuals belonging to the generic "biological process"
> class.
This is true, and it has two explanations.
1. When we started to build BioTop we took the classes of the GENIA
ontology as a starting point. In GENIA there aren't biological occurrents.
2. BioTop avoids the overlap with other bioontologies such as the GO,
where biological processes and molecular functions are included.
Nevertheless we see the need for precisely and formally define biological
states, events, processes, as well as qualities, dispositions, and functions.
We plan to do this in the future but we know that this is not a trivial thing.
> Users of Protégé 3.x should note that the participants of
> biological processes and part-of relations are not shown in the GUI,
> because the BFO/RO still lack domain and range restrictions on
> properties. I see that this has been discussed in another thread in
> the BioTop discussion group, as well as the BFO discussion groups.
The BFO / RO people are aware of this.
>
> Will subclasses of "biological process" be added soon? I could make
> some suggestions based on BioPAX and bio-zen, if you would be
> interested.
>
Your suggestions are highly welcome. Is BioPAX and bio-zen compatible
with GO in terms of "biological process" ? Is there a clear distinction between
a process (an occurrent) and a function (a dependent continuant) ?
What about states and events? Is an event a process ? Or a part of a process?
Can a process be part of an event ?I tend to the position that events
are related to processes such as count objects are related to masses.
This has some roots in philosophy (cf. Antony Galton: On What Goes On:
the Ontology of Processes and Events, FOIS 2006). All these are
interesting questions. The application of these concepts to
bio-ontologies still require a good deal of basic ontological
research.
Nevertheless we plan to extend our framework to processual entities,
but these should fit to existing domain ontologies (BioPAX, GO) on the
one hand, and to top-level ontologies (BFO, DOLCE) on the other hand.
>
> Will subclasses or individuals be added to "MolecularFunction" and
> "ProteinFunction", or does such an extension already exist? Will the
> 'concepts' from the Gene Ontology be represented as classes or
> individuals?
The GO concepts are considered classes (universals) by GO. To
interpret them as individuals would make no sense. BioTop does not
claim to re-invent the wheel, and its overlap with existing domain
ontologies is only required (1) where the classes under scrutiny are
non sufficiently well defined in the doimain ontology, or (2) where
more specific classes are necessary for defining the central BioTop
classes.
>
> Do you plan on adding constructs (or import constructs from other
> ontologies) for the representation of data like protein sequences,
> molecular weights, reaction equilibria etc.? How would you advise that
> things like bibliographical annotations or references to database
> entries should be represented? Maybe OBI could be useful for this when
> it matures?
> biotop:AminoAcidMonomer should probably be a subclass of
> biotop:Monomer
In the unclassified version of BioTop those taxonomic links that can
be inferred by the classifier are deliberately missing. The subclass
relation you mentioned should be present after successful classifying
the ontology.
> Why are classes for molecules etc. not subclasses of "Object"?
>
> If these classes are made subclasses of "Object", I think
> "ObjectAggregate" should have a restriction like "hasGrain some
> Object".
This was exactly what we had in an earlier version, but then we read
the BFO definitions and became hesitant. Isn't any object in biology
an "object aggregate" of elementary particles, and even those are
aggregates of quarks etc... ??
Pierre Grenon who designed the subdivision of the category
"independent continuant" admits that there is a commitment to a
"mesoscopic" granularity level, whereas I argue that the categories in
an upper-level ontology should be granularity-independent.
A problem with "object aggregate" is that the objects must not share a
boundary, this is maybe to strong a restriction. Cells in a tissue,
e.g. share boundary parts.
> The "aggregate" classes are not structured in class/subclass relations
> like their grain counterparts are. I think this should be done, if it
> does not collide with any formal restrictions of BFO.
This occurs when you run the classifyer.
> The only direct link to the biotop OWL file points to a SVN, so the
> contents of this file are probably subject to ongoing change. It might
> be useful to have a stable URL for importing each version of biotop.
>
This is an issue for Holger Stenzhorn, who cares about the technical details.
>
> Do you have any suggestions for further developments and improvement?
>
Yes, help maintain the mailing list alive and keep on reporting your
experiences.
Thank you a lot for your valuable comments.
Stefan
--
PD Dr. Stefan SCHULZ [stsc...@uni-freiburg.de]
Universitätsklinikum - Abt.Medizinische Informatik
Stefan-Meier-Strasse 26 D-79104 Freiburg
[home: Eschholzstr. 70 D-79115 Freiburg]
+49 (0)761 2033252, 2036702, 2049089, FAX 2033251
www.imbi.uni-freiburg.de/medinf/schulz
Skype: stschulz