Figure 1.

Comparing biological meaning associated with data and model resource (DMR) elements. A. Free-text labels associated DMR elements that convey human-readable meaning (e.g. text label associated with a data column in a spreadsheet) are a very common method of documentation. Text mining methods can assist with finding relationships between text labels, but may encounter difficulties in identifying closely related concepts expressed using different words: for example the labels "Blood Flow to the Lungs" and "Pulmonary Cardiac Output" have very similar meaning but their textual representation is very divergent. B. Controlled vocabularies provide a standard set of Uniform Resource Identifiers (URIs) with which relevant biomedical concepts may be unambiguously associated. For example, while each of the two elements carries a distinct free-text label, their metadata mappings to the same controlled vocabulary term (with ID#1:1234) makes it explicit that the annotations associated with the two DMR elements are semantically identical (i.e. are synonymous). C. Ontologies provide explicit machine readable knowledge about relationships between terms. The above example illustrates the hierarchy of parts of the heart. By explicitly representing knowledge as well-defined concept nodes and relation edges between such concepts, it is possible to compare DMR metadata associated with concepts from the same ontology precisely and automatically. D. Part of the RICORDO effort is to provide tools for the annotation of DMR metadata with composite ontology structures. A composite term consists of two or more ontology terms in which the relationship between such terms is explicitly represented within the composite knowledge structure. Such composites may be compared on the basis of the terms that compose them - for instance, the two composites depicted in this diagram may be compared, using classification tools, on the basis of the ontology terms for cardiac structure (#2: red) and biological qualities (#3: blue) from which they are derived.