By looking for genes in model organisms that are similar to a given Homo sapiens gene, researchers can learn about the protein the Homo sapiens gene encodes and search for drugs to block it. The MLHI gene, which is associated with colon cancer in Homo sapienss, is used in this example.

It all adds up to good days ahead for bioinformatics, which many assert holds the real promise of genomics. “Genomics without bioinformatics will not have much of a payoff,” states Roland Somogyi, former director of neurobiology at Incyte Genomics who is now at Molecular Mining in Kingston, Ontario.

Michael N. Liebman, head of computational biology at Roche Bioscience in Palo Alto, agrees. “Genomics is not the paradigm shift; it’s understanding how to use it that is the paradigm shift,” he asserts. “In bioinformatics, we’re at the beginning of the revolution.”

The revolution involves many different players, each with a different strategy. Some bioinformatics companies cater to large users, aiming their products and services at genomics, biotechnology and pharmaceutical companies by creating custom software and offering consulting services. Lion Bioscience, based in Heidelberg, Germany, has been particularly successful at selling “enterprise-wide” bioinformatics tools and services. Its $100-million agreement with Bayer to build and manage a bioinformatics capability across all of Bayer’s divisions was at that time the industry’s largest such deal.

Other firms target small or academic users. Web businesses such as Oakland, Calif.-based Double Twist and eBioinformatics, which is headquartered in Pleasanton, Calif., offer one-stop Internet shopping. These on-line portals allow users to access various types of databases and use software to manipulate the data.

In May 2000, DoubleTwist scientists announced they had used their technology to determine that the number of genes in the Homo sapiens genome is roughly 105,000, although they said the final count would probably come in at 100,000. For those who would rather have the software behind their own security fire-walls, Informax in Rockville, Oxford Molecular Group in England, and others sell shrink-wrapped products.

Making Connections

Large pharmaceutical companies“big pharma”have also sought to leverage their genomics efforts with in-house bioinformatics investments. Many have established entire departments to integrate and service computer software and facilitate database access across multiple departments, including new product development, formulation, toxicology and clinical testing. The old model of drug development often compartmentalized these functions, ghettoizing data that might have been useful to other researchers. Bioinformatics allows researchers across a company to see the same thing while still manipulating the data individually.

In addition to making drug discovery more efficient, in-house bioinformatics can also save drug companies money in software support. Glaxo Wellcome in Research Triangle Park, N.C., is replacing individual packages used by various investigators and departments to access and manipulate databases with a single software platform. Robin M. DeMent, U.S. director of bioinformatics at Glaxo Wellcome, estimates that this will save approximately $800,000 in staffing support over a three-to five-year period.

To integrate bioinformatics throughout their companies, pharmaceutical giants also forge strategic alliances, enter into licensing agreements and acquire smaller biotechnology companies. Using partners and vendors not only allows big pharma to fill in the gaps in its bioinformatics capabilities but also gives it the mobility to adapt new technologies as they come onto the market rather than constantly overhauling its own systems. “If a pharmaceutical company had a large enough research budget, they could do it all themselves,” Somogyi says. “But it’s also a question of culture. The field benefits as a whole by providing different businesses with different roles with room to overlap.”

Occupying some of that overlapin resources, products and market capitalizationare companies such as Human Genome Sciences, Celera and Incyte. They straddle the terrain between big pharma and the data integration and mining offered by specialist companies. They have also quickly seized on the degree of automation that bioinformatics has brought to biology.

But with all this variety comes the potential for miscommunication. Getting various databases to talk to one anotherwhat is called interoperabilityis becoming more and more key as users flit among them to fulfill their needs. An obvious solution would be annotationtagging data with names that are cross-referenced across databases and naming systems. This has worked to a degree. “We’ve been successful in bringing databases together by annotation: database A to database B, B to C, C to D,” explains Liebman of Roche Bioscience. “But annotation in A may change, and by the time you get down to D the references may not have changed, especially with a constant stream of new data.” He points out that this problem becomes more acute as the understanding of the biology and the ability to conduct computational analysis becomes more sophisticated. “We’re just starting to identify complexities in these queries, and how we store data becomes critical in the types of questions we can ask,” he states.

Systematic improvements will help, but progressand ultimately profitstill relies on the ingenuity of the end user, according to David J. Lipman, director of NCBI. “It’s about brainware,” he says, “not hardware or software.”

Hooking Up Biologists

Imagine that your co-worker in the next cubicle has some information you need for a report that’s due soon. She e-mails it to you, but the data are from a spreadsheet program, and all you have is a word processor, so there’s no possibility of your cutting and pasting it into your document. Instead you have to print it out and type it in all over again. That’s roughly the situation facing biologists these days. Although databases of biological information aboundespecially in this post-genome-sequencing eramany researchers are like sailors thirsting to death surrounded by an ocean: what they need is all around them, but it’s not in a form they can readily use.

To solve the problem, various groups made up of academic scientists and researchers from biotechnology and pharmaceutical companies are coming together to try to devise computer standards for bioinformatics so that biologists can more easily share data and make the most of the glut of information resulting from the Human Genome Project. Their goal is to enable an investigator not only to float seamlessly between the enormous databases of DNA sequences and those of the three-dimensional protein structures encoded by that DNA. They also want a scientist to be able to search the databases more efficiently so that, to use an automobile metaphor, if someone typed in “Camaro,” the results would include other cars as well because the system would be smart enough to know that a Camaro is another kind of car.

The immediate payoff is expected to be the faster development of new drugs. “Pharmaceutical research is the only industry I know of with declining productivity,” says Tim Clark, vice president of informatics for Millennium Pharmaceuticals in Cambridge, Mass. “The R&D effort is at a primitive craft scale, like cottage weavers, although standardization is one of the first problems that got tackled in the Industrial Revolution, with the invention of interchangeable parts.”

The issue is what standards to use. In a situation reminiscent of the computer industry in the 1970s, everyone advocates standards, as long as they are his or her own. Formal groups have sprung up worldwide with names like the BioPathways Consortium, the Life Sciences Research Domain Task Force of the Object Management Group, and the Bio-Ontologies Consortiumand each has a different idea of how things should be done. Eric Neumann, a member of both the Bio-Ontologies and BioPathways consortia, is a neuroscientist who is now vice president for life science informatics at the consulting firm 3rd Millennium in Cambridge, Mass. (no relation to Millennium Pharmaceuticals). He says Extensible Markup Language (XML) is shaping up to be the standard computer language for bioinformatics. XML is the successor to Hypertext Markup Language (HTML), the current driver of the World Wide Web.

One of XML’s advantages is that it contains tags that identify each kind of information according to its type: “Camaro,” for example, would be tagged as a car. Neumann proposes that XML-based languages will “emphasize the Web-like nature of biological information,” which stretches from DNA to messenger RNA, proteins, protein-protein interactions, biochemical pathways, cellular function and, ultimately, the behavior of a whole organism. Current ways of storing and searching such biological information are centered on single genes, according to Neumann, “but the diseases we want to treat involve more than one gene.”

Clark says the main problems facing bioinformatics that make standard development necessary are the sheer volume of data, the need for advanced pattern recognition (such as within DNA sequences and protein structural domains), the ability to process signals to eliminate “noise” from data, and something called combinatorial optimization, or finding the best path through a maze of molecular interactions. “You can’t build all of it yourself,” he contends.

Neumann thinks combinatorial optimization could be the highest hurdle. “Pathways are a lot more complex than [DNA] sequences,” he states. “If we don’t come up with something, it’s going to be a real mess.”