SANTA CRUZ, Calif.--Are institutions of higher education to blame for the shortage of skilled bioinformatics employees? Some bioinformatics managers in industry who talked to BioInform attribute at least part of the struggle to fill job openings to universities that they said aren't teaching students the right skills.
At the same time, many concede that the demands of the rapidly evolving genomics business are such that academic programs have a hard time keeping up.
Michael Liebman, director of bioinformatics at Wyeth-Ayerst in Radnor, Pa., said education programs in bioinformatics aren't producing graduates with appropriate skills. "In the effort to put out numbers, programs are being formed that aren't aligned with company needs," Liebman said. He noted that university programs seem to emphasize sequencing and genomic data, overlooking other important areas in bioinformatics.
Internships are one way to prepare students for industry jobs, Liebman suggested. "They have no idea what they are getting themselves into if they don't do internships," he said.
Bioinformatics programs that teach biology and computer science skills don't necessarily show students how to solve complex problems. They might understand the science, but have a hard time formulating that understanding into a solution, Liebman observed. Workplace experience teaches students to use their academic knowledge to solve real-world problems, he contended.
THE NEXT GREAT ALGORITHM
Will Gilbert, a principal scientist at Astra Research Center Boston, agreed that academic bioinformatics programs come up short in preparing students for industry. "They aren't producing what the industry needs," he commented.
Gilbert worried that students focus on writing "the next great algorithm" instead of building tools industry can use. A theoretical focus works well in academics, he said, but bioinformatics students looking for industry jobs should be more pragmatic. He suggested that a good training program would require students to perform a joint project with another discipline to introduce them to working across different fields.
Academic programs also tend to focus on students at the doctorate level. Gilbert argued that industry needs more people at the lower levels who can use the current bioinformatics tools effectively. "I get so many resumes that look the same," he said, noting that many undergraduates have a strong molecular biology background but no understanding of the bioinformatics tools pharmaceutical companies rely on.
Gilbert said more undergraduate bioinformatics courses, such as the one he teaches at the University of New Hampshire, would prepare students for industry positions.
David Searls, director of research and development at SmithKline Beecham in King of Prussia, Pa., hasn't hired anyone straight out of an academic bioinformatics program. Bioinformatics requires a broad range of skills that are hard to teach in one program, he observed. "It is important not just to have individual skills, but to have a combination of skills," Searls commented. "My suspicion is that there may be some shortfalls in terms of comprehensiveness."
Celera, the new company established by Craig Venter and Perkin Elmer to sequence the human genome in three years, is in the midst of a large bioinformatics recruiting campaign. Tony Kerlavage, director of gene discovery there, described Celera's ideal bioinformatics candidates as "people who have an understanding of two different worlds."
On one hand, Kerlavage said, bioinformatics hires must understand computers. "That doesn't necessarily mean they have to be a computer scientist or a software engineer, but they have to have some programming skills, particularly in Perl. It's immensely useful to be able to write your own scripts, to be able to manipulate data, because inevitably you never have the tools at your fingertips that you need to move things around," he explained.
At the same time, Kerlavage continued, new bioinformatics staff at Celera must know biology. "It doesn't do any good for someone to look at a sequence alignment without knowing what it means. What does that probability value really mean? A lot of people use tools like Blast and say what's the top hit in here and think that's their answer without thinking more deeply about what it means," he observed.
"A good bioinformaticist has to understand the underlying theory of how these algorithms work," Kerlavage continued. "They don't need to be able to write them, but they have to understand how they work and how changing the parameters will change the outcome."
Kerlavage asserted that students can become overly dependent on bioinformatics tools available on the internet. "There are all these services on the internet and you just go out and send your sequence to this server and that one, and the answer comes back. Certainly small labs looking at one protein or one gene can get the information they need that way," he said.
But he argued that's not enough in a high-throughput environment. "You need to really understand the intricate details of what all the results from all these tools mean. I would hope that's one of the things these programs are teaching people. What these algorithms are really doing under the hood is really important for people to understand," he said.
Kerlavage admitted that in Celera's current search, experience is often a more important factor than a job candidate's degree. "The people we're interviewing are the ones who have that broad experience and have worked on a problem and used a variety of tools to solve it. People who have gone out and done that sort of thing are the ones who impress us the most. We're just going to have to do a lot of training as they come in," he concluded.
The perspective of bioinformatics tools vendors is somewhat different. Myra Williams, president and CEO of Molecular Applications Group, disagreed with those who said bioinformatics programs aren't turning out fully trained employees.
Williams said the programs are producing good students, but not enough of them. "There are so few people who have had formal training in bioinformatics that we haven't hired one," she told BioInform. Instead, Molecular Applications has chosen to train biologists in bioinformatics on its own, she said.
Candidates with strong biology backgrounds design software that better meets drug discovery needs, Williams commented. "People who are in computer science think in terms of jargon," she said. "They require users to know that jargon." Most biologists who use the tools don't know the jargon and don't want to. They build tools that are easy for other biologists to use, she claimed.