By John Russell
March 10, 2003 | SAN DIEGO – Code jockeys, beware. Bioinformatics is a dead-end specialty soon to be subsumed by the rest of biology. Pay attention, Microsoft wannabes – and, perhaps, Microsoft itself. The life science market is small, and the plethora of popular open-source tools will make achieving commercial success difficult indeed.
So said keynoter Lincoln Stein, somewhat tongue in cheek, at last month’s annual O’Reilly Bioinformatics Technology Conference 2003, which drew roughly 600 hardcore bioinformaticians. Stein is something of a bioinformatics pioneer, which made his talk, “Bioinformatics: Gone in 2012,” all the more intriguing.
Trained as a physician, Stein was a hospital pathologist before becoming the head of bioinformatics in Eric Lander’s lab at MIT’s Whitehead Institute Center for Genome Research. Currently, he is a researcher and associate professor at Cold Springs Harbor Laboratory, specializing in integrated databases for biological sciences.
“The title [of my talk] was facetious,” Stein said. “Really, the opposite will occur. Bioinformatics will be ubiquitous, but I fear people thinking they are getting into a hot field and finding themselves in the equivalent of running a transgenic facility.”
“The [transgenic] technique itself was [once] a hot field of research. Now it’s mundane. You take your plasmid (containing the DNA to be inserted into the target animal) to the transgenic core, and a master’s degree student makes a transgenic mouse for you.” Computational biology is a more accurate description of what bioinformatics is attempting to accomplish, he said. “I actually don’t like the term ‘bioinformatics.’ I think it will sound pretty passé.”
Demand for bioinformatics skills won’t disappear, though. In fact, computer training will become part of mainstream biology education, just as it already is with chemistry and physics. “It will turn into learning the ability to use computers to analyze and integrate large data sets,” he said.
A good model, Stein said, is the graduate genetics program at Washington University in St. Louis. “A major part of the students’ education is algorithm development, computational biology, and software development. It’s not treated as a separate thing. I am concerned about programs where bioinformatics is a separate track.”
Stein predicts that integrating computer science into biology curricula will lead to better design of experiments and more efficient use of wet-lab resources.
Software companies hoping to sell into this market will find the going tough, he said. “We’re likely to have a few successful companies that produce software packages that are widely used, but much of the software will continue to be open-source-style academic packages.”
Developing interoperability standards and deciding how to judge a good piece of open-source software remain challenges for bioinformaticians: for example, what the standards should be governing publication of bioinformatics tools and research results.
“How do you peer review a piece of software?” Stein asked. “What are the academic standards to which software should be held? Does the source code need to be reviewed or made widely available to other researchers for their use as a condition of being published?” Some kind of academic infrastructure for bioinformatics needs to be established, he urges.
Eventually, bioinformatics services will be provided to researchers by support centers, Stein believes. Universities will have people specializing in data mining and databases.
Stein’s advice: “I really think one shouldn’t focus on a particular skill set but on a problem domain. Bioinformatics is becoming mature.”