July 11, 2002 | In a keynote presentation at the O'Reilly Bioinformatics conference held earlier this year, entitled Bioinformatics: Building a Nation from a Land of City States, Lincoln D. Stein, a noted researcher at the Cold Spring Harbor Laboratory on Long Island, addressed one of the most critical issues in bioinformatics research today — the profusion of data and the incompatibility of databases. An edited version of his talk was recently published in Nature magazine.
Stein likens the current state of bioinformatics to the appearance of Italy in the Middle Ages, a country characterized by an assortment of family-run city-states, each with different currencies, dialects, legal systems, and so on. Italy produced its share of brilliant scientists but lagged in other areas, such as technology and industrialization.
Similarly, the lack of standardization and other impediments to data sharing are proving detrimental to biomedical research, ironically at a time when it should be in the throes of its own renaissance. To illustrate his point, Stein displayed a map of 12th-century Italy, depicting numerous multicolored territories and regions, and proceeded to superimpose the logos of the world's
leading bioinformatics organizations — NCBI, Ensembl, FlyBase, WormBase, the University of California at Santa Cruz, and others. This profusion of databases and resources has fueled an explosion in genomic research, but it also creates some serious problems. "Unarguably," writes Stein, "this environment of creative chaos has greatly enriched the field. But it has also created a significant hindrance to researchers wishing to exploit the wealth of genome data to its fullest. ... This inconvenience for the bench biologist is disastrous for the bioinformaticist, who typically needs to aggregate data from many online sources to create a data set for further analysis."
As an example, consider a researcher who wishes to examine the complete repertoire of genes from the nematode Caenorhabditis elegans in order to find those that affect neuronal function and putatively encode a secreted protein. What would be a tedious process by hand, involving multiple steps of downloading, reformatting, cutting, pasting, and e-mailing, can be automated with the use of Perl, a popular scripting language that is particularly adept at sequence comparisons. The problems become trickier, however, when sequences from different organisms are compared, as each genome database typically has its own format and user interface. As a result, bioinformaticists tend to rely on short-lived, screen-scraping scripts to extract only relevant textual information, but these are frequently rendered useless by even trivial changes on a target Web page.
Stein is not arguing in favor of a homogeneous system of data representation, because as he says, "such conformity would destroy the creative aspect of online databases." However, in the short term, he does advocate a seven-point "code of conduct" to facilitate bioinformatics research and collaboration (see above right).
The Long Haul
As a long-term solution, Stein proposes the use of Web services — a new distributed computing architecture building on existing distributed efforts, such as CORBA (Common Object Request Broker Architecture), that use existing Internet communications and data exchange standards to make databases and other applications more accessible to researchers. In this scenario, individual online databases would still exist, but a single script could interrogate multiple
|Code of Conduct for Bioinformaticists
|1 A Web page is an interface: Facilitate access of Web pages to scripts.
2 An interface is a contract: Give advance warning of interface changes to minimize inconvenience; use new URLs and provide temporary legacy interface.
3 Choice is good: SOAP/XML is "sweet!" — HTML is discouraged.
4 Allow batch downloads: Break data sets into bite-sized pieces to facilitate downloading.
5 Use existing file formats: Do not reinvent the wheel when multiple workable formats already exist.
6 Design sensible formats: Use common sense and simplicity.
7 Allow ad hoc queries: Make copies of databases available for direct access using native query language.
databases, allowing databases to continue to operate in characteristically distinct ways.
Stein maintains that the necessary technical infrastructure to implement this vision is almost in place. There already exists a dynamic common vocabulary of genes, produced by the Gene Ontology Consortium. XML (Extensible Markup Language) is the preferred transfer format, and XSD (XML Schema Definition) the data definition language. SOAP (Simple Object Access Protocol) is the XML-based wire protocol for exchanging information between computers. WSDL (Web Services Description Language) provides the public interface for a Web service. Finally, UDDI (Universal Description, Discovery, and Integration), which was originally created by Ariba, IBM, and Microsoft, is a registry for publishing and finding businesses and Web services.
A good example is the XEMBL service from the European Bioinformatics Institute (EBI) in Hinxton, England. EBI customarily distributes genome sequence data as a series of flatfiles produced from an Oracle database, but over the past couple of years has developed a CORBA infrastructure to provide a more efficient and accessible means of distributing data files, in response to strong user demand, particularly for XML. One of the XML standards in use is BSML (Bioinformatic Sequence Markup Language), provided by LabBook Inc., which has a free Genomic XML Viewer available for download.
Another EBI resource is the Open Bibliographic Query System, designed to provide access to MedLine and other literature repositories. Although currently available via CORBA, EBI intends to add a SOAP/XML layer to enable access to the repositories as Web services.
Stein wrapped up his O'Reilly presentation by contrasting the multicolored map of Renaissance Italy with today's European Union, depicted as a uniform shade of blue as if to underline its common currency, parliament, and trade policies. Stein hopes that this could be a metaphor for the maturation of the bioinformatics industry in the next decade.