Two Layer Cakes

The penetration of the Semantic Web into bioinformatics is not like other disciplines. Given critical data resources such as the Gene Ontology (GO) and the challenges associated with the explosion in the number biological databases, biologists, as a community, are already familiar with the advantages of ontologies and the challenges of information integration. Rather than starting from the pedagogical beginnings, penetration will depend on a migration path from the Open Biomedical Ontologies language (OBO) to the Semantic Web. OBO emerged from GO, and is now host to over 60 different ontologies. We have methodically examined each of the constructs of OBO and mapped them to constructs in the Semantic Web stack. We find that most of OBO can be decomposed into layers with direct correspondence to the Semantic Web layer cake. In the process we have enumerated constructs in each system that do not have a simple syntactic equivalent in the other. Elements of OBO “missing” in the semantic web are few, and can still be expressed in OWL. Thus, OBO ontologies may be translated to the Semantic Web. Further, we believe if certain ancillary information is retained during translation, the Semantic Web representation may be translated back to OBO, and the cycle repeated without any loss of knowledge. It is our expectation that tools to automate this process will enable important legacy ontologies onto the Semantic Web. OBO Layer Cake To create a transformation mechanism between OBO flat files and Semantic Web technologies, we find it useful to create a layer cake for OBO, similar to that of the Semantic Web. OBO tags can be partitioned into three layers – OBO Core, OBO Vocabulary, and OBO Ontology Extensions. See Figure 1. The correspondence between the layers of the two systems is direct. In addition, we describe the expressivity of OBO format in terms of various versions of OWL, i.e. OWL Lite, OWL DL and OWL Full.