Gene Ontology Annotation Tool (GOAT)

There now exist many biological databases containing enormous quantities of entries of genes and gene products along with descriptions and data about a wide variety of their functional properties. However, the synonymy and polysemy of the descriptive terms and the lack of explicit relationships among them hampers consistent, reliable querying of and interoperability between these databases. In response to this, the Gene Ontology (GO), a structured controlled vocabulary of nearly 17,000 terms, has been (and is being) developed to be used to functionally describe the gene products of various organisms, for which it is becoming the de facto standard. GO is divided into three subontologies of terms (most of which also have natural-language definitions) which may be used to annotate gene products in terms of the molecular functions they possess, the higher-level biological processes in which they are involved, and the cellular locations in which they are active. Each term of each of these subontologies is related to each respective parent term via an is-a or a part-of relationship.

GO has been a success in that its terms are being used to functionally annotate genes and gene products in a number of prominent biological databases. However, as GO continues to increase in size, users find it increasingly difficult to find the terms they wish to use for annotation. Furthermore, although a large vocabulary is provided, the terms have no links to each other apart from those relationships that form the three taxonomic/partonomic hierarchies. Thus, beyond this hierarchical information, there are no constraints within GO that can be used to indicate which terms should or should not be used together in the annotation of a given gene product. It is possible (though unlikely) that an annotator, in describing a protein, could associate the terms "viral life cycle", "amino-acid biosynthesis", and "extracellular matrix" to that protein; it is more likely that he would accidentally do so. In either case, this is likely to be biologically nonsensical. Good annotation relies upon the domain expertise of the annotator and the usability of the annotation tool. We seek to improve upon the latter by creating formal relationships between pairs of GO terms (as well as between GO terms and gene-product types) mined from biological databases and building an application that, relying upon these relationships, can dynamically retrieve and present those GO terms that are most likely to be applicable for a given gene product based on the GO terms and the gene-product type already entered by the user for that gene product. Thus, if an annotator has already selected “viral life cycle” as a biological-process term and then indicated that she wanted to add a molecular-function term, she would be presented with those molecular-function terms that have been used as annotating terms along with “viral life cycle” (as well as those terms’ descendants).

Contact  Mike Bada or Robert Stevens for more information.

Last modified 4 April 2004 by Mike Bada.