When a Rose Must Be Called a Rose

Douglas Brutlag challenges students in his computational biology classes at Stanford University to search the large proteomics databases for yeast membrane proteins. Without knowledge of the database lexicons, the students generally come up well short of the mark. "They find 20 to 200," says Brutlag, professor of biochemistry and medicine at Stanford's School of Medicine. "In fact, there are almost 2,000 proteins." The problem: linguistics. "These are controlled vocabularies," Brutlag explains. "The key words for membrane proteins are trans membrane, inner membrane, and outer membrane, and unless you have synonyms for all of those, you miss them when you search the data." Multiply the graduate students' challenge in searching databases by orders of magnitude, and this represents the trials of researchers in academia and industry worldwide. There is a Babel of different computer languages, imagery systems, and software programs that use unique symbols and store their biological treasures in distinctive file types. The Molecular Biology Database Collection (2002 update) of Nucleic Acids Research includes nearly 400 databases;1 some experts estimate that these computer warehouses use 400 different file types, and an uncountable number of different names for the same objects. "There's just so many IDs for everything," says J. Mike Cherry, an associate professor at Stanford who runs the Saccharomyces Genome Database. "Part of [the challenge] is coming up with a standard linking system. People have talked about this with the human genome; there's a lot of evolution. One program calls a gene [one name] and in another it's called another." These colliding terms and databases that have vexed academia for years --since before the Human Genome Project opened thousands of new sequences and millions of single nucleotide polymorphisms to study--have now become a critical obstacle to industry. Both biotechnology and Big Pharma companies want to plumb the databases to find new genes and proteins. But they also want to link the information to other databases, as well as to the journal articles that assess their value. The goal is to condense costly drug discovery times and reduce failures in clinical trials. To do this, multiple databases have to share the same platform, use similar file systems, and employ controlled vocabularies. But the sheer enormity and diversity of the information slows the research. "The biggest problem we have is massive data and little integration," said biotechnology investment banker G. Steven Burrill at the Biotechnology Industry Organization's 2002 annual convention in Toronto. Don't look for biotechs to solve integration problems soon. The success of private database companies and integrators hinges on their ability to attract financing, and bioinformatics companies were the first to feel the effects of a decline in biotech stocks in 2001. The market forced pure data systems such as Celera Genomics Group in Rockville, Md., to prove the value of its information to the drug pipeline. Now, rather than marketing data systems to Big Pharma, many companies are touting their own drug discovery capabilities. "Many of the information providers are downstream- integrating," says Gary Pinkus, a partner in the San Francisco office of McKinsey & Co. "They are selling access to their databases to create enough cash ... to use for their own drug discovery." Many small informatics companies bank on Big Pharma purchasing their software integration systems or on acquiring the companies. But the pharmaceutical industry has yet to be convinced of the downstream value of database integration. "The pharma companies are saying, 'We want to be the integrators, we will pick and pull best-of-breed from different information-based companies and then we'll define the platform,'" says Pinkus. "That makes it more challenging for anyone who is trying to be an integrator." AstraZeneca, for example, has taken an architecture approach to integration, and regularly surveys informatics experts to determine how much the pharmaceutical company leaders want to purchase and how much they will construct in-house. "To second-guess what is the next [platform]...is hard," says Jason Swift, the company's global information system program manager, based in Alderley Park, UK. "But we are trying to react to that with an architecture with which we will be able to integrate new systems." Bioinformaticians have launched multiple standards organizations, including the Interoperable Informatics Infrastructure Consortium (I3C), and, while companies still market data tools, integration remains an obstacle to building a new generation of drugs and to eking out profits from a changing industry. "We still lack the necessary linkages--standards and integration--to extract real value and efficiency from information technology," according to the Burrill & Co. 2002 report on the biotechnology industry.2 The entrance of giant information technology companies like IBM, Hewlett Packard, Sun Microsystems, and Oracle into the campaign for universal information architectures has encouraged bioinformaticians. The three companies joined several others in creating and financing I3C in June 2001. That organization has already built both standards and a lexicon for biological terms, and expects to evaluate new standards submitted by software engineers. In turn, the industry heavyweights propose that by creating common platforms, they can profit through sales of components, training, and linked services.3 References 1. A.D. Baxevanis, "The molecular biology database collection: 2002 update," Nucleic Acids Research, 30:1-12, Nov. 20, 2001. 2. G.S. Burrill, "Biotech 2002: The 16th annual report on the industry," San Francisco: Burrill & Co., 2002. 3. S. Jaffe, "Enter the heavyweights," The Scientist, 16[13]:44, June 24, 2002">The Scientist, 16[13]:44, June 24, 2002.

When a Rose Must Be Called a Rose

Interested in reading more?

Become a Member of