Ontological Design Patterns for the Management of Molecular Biological Knowledge

Bringing Knowledge to Business Processes

Volume

Papers from the 2000 AAAI Spring Symposium

Contents

Molecular biological knowledge has to be represented and managed in such a way that it can be disseminated and applied as an information resource within molecular biology or related areas, such as pharmacology and genetics. Molecular biologists interpret their scientific data by comparing new data against a large amount of existing knowledge in order to predict further conclusions. For instance, the DNA structure of an unknown DNA sequence, its function as a protein, and its evolutionary role are inferred from the degree of similarity with sequences of well-known proteins. The biological function of a protein is defined by both the detailed chemical properties of its surface and its structure, which together depend on the underlying amino acid sequence. One can "predict" function by recognising similarities of different types at the sequence level. The inference of function from protein sequence similarity searches is non-trivial and to be effective requires considerable biological knowledge. For example, to infer function from sequence or pattern database similarities, one must refer to the literature to find the necessary biological evidence to support that inference. Various biology related research areas, such as medicine and molecular biology, started to build ontoiogies to represent parts of their knowledge and terminology, and to provide concise and non-ambiguous biological models (e.g. GALEN for medical language and knowledge (Rector et al. 1997)). Ontoiogies include concepts and instances of information, showing their relationships and possible constraints given by the knowledge area. A concept summarises features of its instances and names their generalisation. For instance, "flowers" becomes the name of the concept about various kinds of flowers, which is described by a definition and attributes about generalised features common to all its concrete instances. An attribute (e.g. leaf-colour) has definition and additional semantic properties. The definitions include information about the context and well-formed use of the terms. Instances (e.g. the rose on my table) are described by values or objects, and belong to their concept.