The Genom of Homo sapiens.pdf

Ontologies for Biologists: A Community Model for theAnnotation of Genomic DataM. ASHBURNER,* C.J. MUNGALL, †‡ AND S.E. LEWIS ‡*Department of Genetics, University of Cambridge & EMBL - EBI, Hinxton, Cambridge, United Kingdom;† Howard Hughes Medical Institute, and ‡ University of California, Berkeley, CaliforniaGenomics has made biology into an information science.David Botstein, June, 2003We celebrate the 50th anniversary of the discovery ofthe structure of DNA in 1953. The nature of the geneticcode became the theoretical preoccupation of the ensuingdecade, and immediately the language of information scienceentered biology (see Kay 2000). The discoveries ofthe genetic code, both its structure (syntax) and content(semantics), and of mechanisms for error recovery duringits transmission, were followed, in the next two decades,by the development of methods to readily determineDNA sequences, to clone specific DNA sequences inbacterial hosts, and to amplify sequences in vitro by thepolymerase chain reaction. Within 40 years or so of Watsonand Crick’s discovery, the first complete sequencesof bacterial genomes were determined; today we celebratethe completion of the human genome. These advancesbrought a revolution that has affected even themost conservative fields of biology (see, e.g., Hebert etal. 2003).None of these advances, except the very first, wouldhave been possible without the application of methodsfrom computer science, a field whose growth and maturityhave closely paralleled that of molecular biology.Computational methods were introduced to biology in theearly 1950s, for the calculation of Fourier summations inprotein crystallography (Bennett and Kendrew 1952;Huxley 1990). However, it was the need to capture, store,assemble, and analyze DNA sequence data (Staden 1980)and to correlate them with other biological knowledgethat motivated a new field of science—bioinformatics.The first attempts to collect protein and nucleic acid sequencedata were published as slim printed books (Dayhoffet al. 1965; Croft 1973; Barrell and Clark 1974). Thedevelopment of public computer files of sequence datasoon followed, with the establishment of the PIRDatabase in 1980, the EMBL Data Library and Genbankin 1982, and the DDBJ in 1986.Few can doubt that, without the international nucleicsequence data library as a common, freely available, depositoryof all public sequence data, neither the achievementof sequencing the human genome, nor its analysis,would have been possible. MEDLINE, Swiss-Prot, andPDB are equally core databases that also are very broadin content (“horizontal”) and play an absolutely centralrole enabling genomic research. In addition, there are anincreasing number of “vertical” databases, narrow in content,but covering this content in great detail. These includethe model organism databases and databases for arestricted class of objects; for example, transcription factorsor eukaryotic promoters. In the last decade or so,model organism databases have been developed as communityprojects for all of the biological organisms commonlyused in research, and the number of specialistdatabases for biological objects has mushroomed. This isillustrated by the growth, from 24 in 1993 to 129 in 2003,in the number of databases that have contributed to theannual database issue of Nucleic Acids Research.The proliferation of databases in this general field ledmany, in the early 1990s, to agonize about questions ofdatabase “interoperability” and to attempt to develop“federations” of databases (see, e.g., Fasman 1994) ordatabase warehouses (e.g., the Integrated Genome Database[Ritter 1994]; see Davidson et al. 1995) or to dictateto database builders a common technical solution todatabase design. These attempts failed, with the only possibleexception being the success of ACeDB in a numberof communities (Durbin and Thierry-Mieg 1991). Yet theproblems that the multiplicity of databases posed to bothbench biologists and computational biologists have notgone away; indeed, they have worsened. They haveworsened for two reasons: One is that the number of databases(and the amount of data) has simply increased, theother is that biologists have increasingly realized the importanceof knowledge from a variety of organisms otherthan their own favorite model. Mouse biologists foundthat they needed information about genes from Drosophilaor yeast; human geneticists needed information aboutgenes in Caenorhabditis elegans or zebrafish.One strategy to overcome some of the problems resultingfrom the dispersion of biological knowledge is toexploit insights from related fields. This is the approachtaken by the many different databases participating inthe Gene Ontology project. This project began in mid-1998 as a collaboration of three model organismdatabases (those for Saccharomyces cerevisiae, Drosophila,and mouse) to solve a very well defined problem:“How can gene products be described in a biologi-Cold Spring Harbor Symposia on Quantitative Biology, Volume LXVIII. © 2003 Cold Spring Harbor Laboratory Press 0-87969-709-1/04. 227