Bioinformatics Boom?

PHOTO CREDIT: Thomas Gottin BIOINFORMATICS PROGRAMMING: Rensselaer's John Salerno (shown with biologist Jane Koretz) has been piecing together courses for the institute's new bioinformatics offerings. "Bioinformatics," the buzzword, seems unavoidable. But, for life scientists, is learning the related skills unavoidable, too? Acquiring that software savvy may not be absolutely necessary, notes W. Graham Richards, professor of physical and theoretical chemistry at Oxford University. But it may be prudent. "If you look in the back pages of Nature or Science, almost every week, there's job after job." Richards suspects that the ocean of data being generated by the Human Genome Project, as well as by a growing number of private sequencing ventures, will provide work for years. "There are masses of data that have to be handled," Richards says. "There's a huge shortage of people to do this work." The shortage of skilled people results in part because few have been formally trained in the relatively new field--which keeps expanding to include genomic comparisons, molecular modeling, and statistical analysis. Most computational biologists, as many prefer to call themselves, have taught themselves and built many of their own software tools. For example, Richards, along with then-graduate student Garrett Morris, now at Scripps Research Institute in La Jolla, Calif., wrote Chameleon, one of the first pieces of bioinformatics software. Only recently have a variety of commercial software packages and formal training programs emerged. Richards expects that midcareer scientists, as well as graduate and undergraduate students, will recognize the opportunities and acquire the necessary skills: "I think there's a huge demand to learn." Carol J. Bult, a member of the mouse genome informatics group at the Jackson Laboratory in Bar Harbor, Maine, agrees: "If you train yourself and if you can learn how to access information and to work on multiple platforms as a biologist, you have a whole world open to you that you wouldn't have otherwise." Scott D. Kahn, senior director of life sciences marketing at Molecular Simulations Inc., of San Diego, is less bullish about bioinformatics job opportunities for mid-career life scientists considering a job change. "A pretty limited number of people will be able to make a midcareer switch," he predicts, adding that the people capable of doing so have already seized the opportunity. "I think a large number will come from graduate schools." Perhaps more important than learning bioinformatics software, he suggests that molecular biologists learn more structural chemistry. "In general, biologists could benefit from having a better understanding of protein structure. Chemists know biology, and biologists don't have a good understanding of 3-D structure of proteins. I think part of the bioinformatics transition is going to be a blurring of that traditional line between chemistry and biology." Mark S. Boguski, senior investigator at the computational biology branch of the National Center for Biotechnology Information, agrees that self-taught bioinformatics experts have already made the transition. "Those relatively few people who are self-taught and are really competent in this area have either been hired by pharmaceutical and biotech companies years ago or they already have jobs that they like." However, Boguski thinks too few of such individuals exist to fill the growing demand (M.S. Boguski, Trends Guide to Bioinformatics, Trends Supplement 1998, Cambridge, U.K., ier Science, 1998). He agrees that current and future crops of graduate students will provide some expertise. "The way universities are responding to this in the short term is they identify one or two of the postdocs or grad students who have a knack for this." In some instances, department chairs snatch up postdocs and grad students with good biological research potential and a knack for computers. Then they send the young scientists to year-long training programs--with the understanding that they will accept a tenure-track position upon their return. But the crop of undergraduates and graduate students now being trained may still not represent enough labor. In addition, it may be years before the new graduate and undergraduate programs being developed yield trained professionals. However, Boguski isn't sure how long the shortage will last. "I think the supply of data to analyze for the foreseeable future will be so great that more and more people will become computational biologists, but perhaps once some of the hype wears off, and once there are enough competent people in the area, it will still be a very good career path with good employment opportunities." And even if midcareer scientists don't want to make a complete switch, some bioinformatics knowledge could help. Boguski suspects that scientists with at least a working knowledge of computational biology, some research background, and management experience could be in line for more high-level positions. Boguski concludes: "Even if you don't want to become a full-time computational biologist, everyone is going to have to have that [skill] on his utility belt." MOLECULAR SIMULATIONS: Software from Molecular Simulations helps predict protein structure based on sequence data. better software. W. Graham Richards "The problem of going from the sequence to the 3-D structure is ... unsolved," W. Graham Richards, professor of physical and theoretical chemistry at Oxford University, notes. "No one, honestly, can predict the structure. There is clearly a need for a breakthrough." Much of the earlier software warehouses sequence data and compares different segments of DNA. But it stops short. "The Human Genome Project and sequencing mean you've got the DNA sequence of lots of proteins; hence, you have the amino acid sequence," Richards explains. "So you know the primary structure, that is, the order of amino acids in a protein. What you would like to know, if you're a pharmaceutical company, is how they fold up and what their three-dimensional shape is." Existing X-ray crystal structures can help some, Richards notes. Some software can compare two similar sequences--one of which has a known crystal structure--and make predictions about the shape of the protein for which no crystallography exists. Still, even those tools have limitations. Carol J. Bult, a member of the mouse genome informatics group at the Jackson Laboratory in Bar Harbor, Maine, adds that other software advances would be helpful as well. "What we lack are really robust tools for visualizing that information and for trying to piece out the patterns and the contextual nature of gene expression." Existing software could also be improved so that biologists who aren't computational whizzes can master it, Bult adds. "Most of the systems are developed for high-throughput sequencing labs, and they're developed by UNIX-savvy people. Different components are developed as experimental tools, so they don't have user-friendly interfaces." And because the existing tools have been developed over time by so many people, with so many diverse research needs, piecing together an analysis package from a combination of commercial and Web-based tools--many designed to run on different operating platforms--can be daunting. "It's not like you have shrink-wrapped software," Bult says. "It's not like you will be able to go to Computer World any time soon and pull off-the-shelf tools." Until programmers develop more user-friendly bioinformatics software and integrate a wide variety of tools into one package, the field will remain dominated by people with a computational background, rather than by biologists with little formal training in bioinformatics, but a large need to analyze the deluge of data. "It will take time for a lot of the stuff being developed now to filter down into the general community," Bult predicts. --P.S.

Bioinformatics Boom?

Interested in reading more?

Become a Member of