In the sections below, we describe the nomenclatures and data models that are used to represent genomic data. Often, however, data models and nomenclatures are described as being ontologies. In some very abstract sense, a data model is an ontology with very little ontological commitment (and little in the way of automated inferential mechanisms), and a nomenclature has even less. A nomenclature simply refers to a method of naming or identifying concepts. From the ontological perspective, a nomenclature allows the representation by name of the concepts in an ontology (whether implicit or explicit) without describing the relationships between these various concepts.
A data model defines a hierarchy of classes, attributes and relationships of those attributes without providing the expressivity based on first-order logic. It also does not have the other inferential capabilities that we have mentioned in the desiderata of section 5.2 and the description of the NIKL effort of Haimowitz et al. For instance, a data model describing the components of a microarray (see section 5.4) might be useful for creating a shared structure for storing the results of a microarray experiment. This data structure does confer the capability to provide even simple inferential relationships such as negation or disjunction. However, a data model would be awkward to state that a gene can code for a transfer RNA or an rRNA or a protein but only two of these three classes. Certainly, elaborate data structures can be developed to capture representational detail, but for every disjunction or exception that the ontology architect wishes to add, the complexity of the data model will grow rapidly. That is, while the data model is necessary to provide a structure in which to maintain the properties of the concepts, it is no substitute for the descriptive power and inferential power of the kind of ontologies previously mentioned. This may appear obvious, but at the time of this writing there are several efforts which claim to be ontologies, but which are in fact no more than glorified nomenclatures and data models.
Was this article helpful?