General Formal Ontology (GFO)
Various domain-specific ontologies have been developed within the
biomedical domain over the last years. For example, Open
Biomedical Ontologies33 (OBO) is an
umbrella organization for various ontologies covering domains such as
the anatomy of individual species, celltypes (6), or
molecular functions of genes and gene products (3).
The rapid growth of biomedical ontologies in size and number leads to
the problem of ontology and data integration. How is it possible
for different ontologies to interoperate? How can the content of
different ontologies be retrieved in a single query?
In contrast to GFO, most biomedical ontologies are represented using
a weak formalism. They can be represented as a directed acyclic graph
(DAG). In a DAG, the categories are represented as nodes, and
the relations between the categories are represented as edges. For
example, the relation that ``nucleus'' is part of a ``cell'' is
represented by two nodes, ``nucleus'' and ``cell'', which are linked by
a directed edge, which is labeled ``part-of''. These graphs are
commonly used in conjunction with a minimal set of axioms, such as
transitivity or symmetry.
Many of the relationships used in these biomedical ontologies can
be defined in GFO. For example, the mereological relations, like
part-of, are already present in GFO. It is often the case that
the semantics of relations using the same name differ between different
biomedical ontologies. Aligning two ontologies that use a relation
with the same name in different ways requires a formalism that will allow
for a representation of the differences in the two ontologies used. These
differences are beyond the expressiveness of DAGs, but can be made
precise within first order logic using the conceptualization that is
provided by the GFO.
In (14), GFO has already been used to
represent knowledge about biological functions in the Gene
Ontology(3), the Celltype Ontology (6)
and the Chemical Entities of Biological Interest (ChEBI) Ontology
(12). As shown in (13,14), the
GFO's method for describing functions using requirements, goals and a role
universal leads to greater expressiveness, and the possibility for more
fine-grained analyses of biological phenomena. In addition, it is
possible to use this analysis to re-analyze the so-called annotation relation34.
Furthermore, GFO plays a role in a curation framework for
biomedical ontologies, which is currently under
development35. This framework is based on a semantic
wiki, and it allows for the formal representation of relations between
concepts within the wiki. These relations are typed, in the sense that
their arguments are restricted to categories, and these are based on
GFO. In a sense, a core ontology is derived from the content
upon which the semantic wiki is based.
GFO, however, provides the possibility for further uses in
biomedical ontologies. In (14), the construction of a
domain ontology based on GFO's treatment of functions is
proposed. GFO can provide the conceptual means to ease the
construction of additional domain-specific ontologies, and provide a
common framework that will be compatible with a majority of the biomedical
ontologies, in order to assist in the integration of different
ontologies, and to make them amenable for automated reasoning.