Nowhere is the impact of the grand biological unification more evident than in the eukaryotes, where the genomic sequences of three model systems are already available (budding yeast, Saccharomyces cerevisiae, completed in 1996 (ref. 1); the nematode worm Caenorhabditis elegans, completed in 1998 (ref. 2); and the fruitfly Drosophila melanogaster, completed earlier this year3) and two more (the flowering plant Arabidopsis thaliana4 and fission yeast Schizosaccharomyces pombe) are imminent. The complete genomic sequence of the human genome is expected in a year or two, and the sequence of the mouse (Mus musculus) will likely follow shortly thereafter.

The first comparison between two complete eukaryotic genomes (budding yeast and worm5) revealed that a surprisingly large fraction of the genes in these two organisms displayed evidence of orthology. About 12% of the worm genes (~18,000) encode proteins whose biological roles could be inferred from their similarity to their putative orthologues in yeast, comprising about 27% of the yeast genes (~5,700). Most of these proteins have been found to have a role in the ‘core biological processes’ common to all eukaryotic cells, such as DNA replication, transcription and metabolism. A three-way comparison among budding yeast, worm and fruitfly shows that this relationship can be extended; the same subset of yeast genes generally have recognizable homologues in the fly genome6. Estimates of sequence and functional conservation between the genes of these model systems and those of mammals are less reliable, as no mammalian genome sequence is yet known in its entirety. Nevertheless, it is clear that a high level of sequence and functional conservation will extend to all eukaryotes, with the likelihood that genes and proteins that carry out the core biological processes will again be probable orthologues. Furthermore, since the late 1980s, many experimental confirmations of functional conservation between mammals and model organisms (commonly yeast) have been published7-12.

This astonishingly high degree of sequence and functional conservation presents both opportunities and challenges. The main opportunity lies in the possibility of automated transfer of biological annotations from the experimentally tractable model organisms to the less tractable organisms based on gene and protein sequence similarity. Such information can be used to improve human health or agriculture. The challenge lies in meeting the requirements for a largely or entirely computational system for comparing or transferring annotation among different species. Although robust methods for sequence comparison are at hand13-15, many of the other elements for such a system remain to be developed.

Go to:
A dynamic gene ontology
The GO Consortium is a joint project of three model organism databases: FlyBase16, Mouse Genome Informatics17,18 (MGI) and the Saccharomyces Genome Database19 (SGD). It is expected that other organism databases will join in the near future. The goal of the Consortium is to produce a structured, precisely defined, common, controlled vocabulary for describing the roles of genes and gene products in any organism. Early considerations of the problems posed by the diversity of activities that characterize the cells of yeast, flies and mice made it clear that extensions of standard indexing methods (for example, keywords) are likely to be both unwieldy and, in the end, unworkable. Although these resources remain essential, and our proposed system will continue to link to and depend on them, they are not sufficient in themselves to allow automatic transfers of annotation.

Each node in the GO ontologies will be linked to other kinds of information, including the many gene and protein keyword databases such as SwissPROT (ref. 20), Gen-Bank (ref. 21), EMBL (ref. 22), DDBJ (ref. 23), PIR (ref. 24), MIPS (ref. 25), YPD & WormPD (ref. 26), Pfam (ref. 27), SCOP (ref. 28) and ENZYME (ref. 29). One reason for this is that the state of biological knowledge of what genes and proteins do is very incomplete and changing rapidly. Discoveries that change our understanding of the roles of gene products in cells are published on a daily basis. To illustrate this, consider annotating two different proteins. One is known to be a transmembrane receptor serine/threonine kinase involved in p53-induced apoptosis; the other is known only to be a membrane-bound protein. In one case, the knowledge about the protein is substantial, whereas in the other it is minimal. We need to be able to organize, describe, query and visualize biological knowledge at vastly different stages of completeness. Any system must be flexible and tolerant of this constantly changing level of knowledge and allow updates on a continuing basis.

Similar considerations suggested that a static hierarchical system, such as the Enzyme Commission30 (EC) hierarchy, although computationally tractable, was also likely to be inadequate to describe the role of a gene or a protein in biology in a manner that would be either intuitive or helpful for biologists. The hierarchical EC numbering system for enzymes is the standard resource for classifying enzymatic chemical reactions. The EC system does not address the classification of non-enzymatic proteins or the ability to describe the role of a gene product within a cell; also, the system has little facility for describing diverse protein interactions. The vagueness of the term ‘function’ when applied to genes or proteins emerged as a particular problem, as this term is colloquially used to describe biochemical activities, biological goals and cellular structure. It is commonplace today to refer to the function of a protein such as tubulin as ‘GTPase’ or ‘constituent of the mitotic spindle’. For all these reasons, we are constructing three independent ontologies.