Celia M. Henry
C&EN Washington

The field of bioinformatics is surrounded with so much hype that thinking it's a brand-new field is forgivable. The word itself was coined in the early 1990s, but people have been using databases to manage biological sequence data--which is a part of what bioinformatics encompasses--since the 1960s. Although the field may not be new, it's moving in new directions and creating plenty of job opportunities.

At its root, bioinformatics is the combination of biology and computer science. "Literally, bioinformatics can mean anything along the spectrum of just bar-coding samples to go into a laboratory information management system all the way up to hypothesis-driven research of the most fundamental kind," says Mark S. Boguski, senior vice president of research and development at Rosetta Inpharmatics in Kirkland, Wash. "Usually, it means something between those two extremes."

For most of its existence, what we now call bioinformatics has been synonymous with the management and analysis of protein or nucleic acid sequences. "Now that we're in the era of functional genomics," Boguski says, "bioinformatics is reinventing itself to go well beyond sequence analysis. It's about statistical modeling of the output of functional genomics technologies. Really, it's about biological data management and analysis."

Bioinformatics professionals combine knowledge of science and computers. [Courtesy of InforMax]

Rick Sheridan, vice president for information science at PPGx , a pharmacogenomics company located in Morrisville, N.C., says his company also broadly defines bioinformatics. "The types of bioinformatics systems that we develop are oriented much more toward data warehousing, data visualization, the combining of clinical data with genetic data on very large scales," he says.

At CuraGen , a genomics-based drug discovery and development company in New Haven, Conn., bioinformatics is divided into laboratory information management systems (LIMS) and "discovery informatics," according to Martin D. Leach, director of bioinformatics. "What's crucial for discovery informatics is that we need to know the quality, source, and [location] of all the different types of biological and proteomics data that we handle. This is why we've put LIMS within bioinformatics."

Ultimately, Boguski says, bioinformatics is about "turning data and information into knowledge and insights." However, he also notes that "usually you just generate more information."

Boguski sees bioinformatics following two paths. On one side, much of bioinformatics has "simply been absorbed into the fabric of the way we do things." For example, researchers don't think twice about searching the databases before conducting an experiment. On the other hand, he says, bioinformatics is becoming a discipline in its own right. "There are certain classic algorithms that you study; there are certain tenets to the discipline that make it a real academic discipline apart from just being a tool kit," he notes.

Peter Covitz, vice president of professional services at InforMax , a Rockville, Md.-based bioinformatics software vendor, believes that the field of bioinformatics has opened up and is enjoying greater acceptance within the academic community. "After a curious amount of resistance early on, there's now a complete embracing of the concept of bioinformatics as a unique discipline in the academic community," Covitz says. "That's a sign of maturity, because academia tends to be pretty conservative."

Boguski believes that some aspects of bioinformatics have reached a limit in what they can do. He had his first experience with bioinformatics as a graduate student, when he searched GenBank for a cDNA (complementary DNA) that he had spent a year cloning and sequencing with no knowledge of its function. "When I did that first database search as a graduate student in 1983, there were only 2,500 sequences in GenBank," he recalls. "The chance that I was going to get a match was infinitesimally small. Luckily, I did get a match, or my career might have been quite different. Hits in the old days were rare but carried a lot of information. Today, hits are frequent but carry very little information. I saw the future not in computational biology alone but really in bringing together computational with experimental biology."

That future is being realized now with the ascendance of high-throughput experimental biology techniques. "I think the time has been ripe for a couple of years to rejoin separate trends of computational biology and experimental biology," Boguski continues. "The reason I joined Rosetta and left the National Center for Biotechnology Information at the National Institutes of Health , where I'd done pretty much strictly computational biology for a dozen years, was because I really saw the future at the cutting edge between computational and high-throughput experimental biology."

New technologies have "elbowed their way" into the field of bioinformatics, according to Covitz. He asserts that the best example is the microarray technologies that are used for expression analysis. The technology generates vast amounts of data that have demanded the attention of bioinformatics professionals. "The world of bioinformatics turned its attention toward expression data and is now working quite busily to refine the analysis methods, to refine the data management methods, and to integrate the methods into the existing base of sequence and structure analysis capability," Covitz says.

The explosion of data from high-throughput techniques has led to a concomitant explosion of bioinformatics positions. Bioinformatics is "one of the hottest jobs going," CuraGen's Leach says, "and it's one of the toughest job positions to fill."

DNA microarrays yield large data sets for bioinformatics professionals to deal with. [Courtesy of Rosetta Inpharmatics]

Rainer Fuchs, director of research informatics at Biogen in Cambridge, Mass., believes that one of the hurdles to creating a top-notch bioinformatics group is providing sufficient challenges and effectively integrating bioinformatics into the company's overall research effort. "If you aren't able to really make your bioinformatics program an integral part of the research strategy, then people will begin to look elsewhere because they don't really see how they're contributing to the overall success of the company," he says. "One of the goals for us here at Biogen is to make sure that our bioinformatics scientists feel like equal partners in our discovery program and feel that they're making a real contribution to the success of the company.

"It's definitely a marketplace where the demand is much higher than the supply," Fuchs continues. "We're all struggling to find people. We're all competing for a small pool of experts."

Actually, many people say there's no shortage of applicants for these positions. The shortage is in qualified applicants.

"We're looking for people who have a strong background in two areas: the biology and the informatics components," Fuchs says. "If you look at the typical résumé that you receive for an ad for a bioinformatics position, it's a biologist who has learned to use computational tools."

However, just using the tools is often not enough. "They need to have a very strong understanding of the methods they're using, of the strengths and weaknesses of the tools," Fuchs adds. "A bioinformatics person has to understand what is the best solution to a given problem. The typical applicant doesn't have that knowledge. A good bioinformatics person is someone who can clearly grasp the biological complexity of problems, has a good understanding of the whole palette of possible computational solutions out here, and is also able to implement and apply the best possible solution to that problem."

Leach agrees. "An ideal bioinformatics person would be someone who has strong biological training and strong computer science training, whether it's for genomics or proteomics or other types of biological data. To find either a computer scientist trained with that degree of biology or a biologist trained with that degree of computer science and software development skills is a rare commodity," he says.

Katharine Giacalone, vice president for human resources at Rockville, Md.-based Celera Genomics , says Celera is "constantly looking for highly experienced bioinformatics professionals." She says they have trouble finding people with appropriate experience. "There are many candidates out there with Ph.D.s who've gone back to complete a master's degree in computer science, but they don't necessarily have the actual work experience using the tools or even building tools or using them against large-scale data sets that we have here at Celera," Giacalone says. "We want to make sure we have the right talent to help us move the business to the next step of researching and analyzing the data."

Steve Chamberlin, the director of bioinformatics at EraGen Biosciences in Alachua, Fla., also bemoans candidates' lack of experience, particularly industrial experience. Most of his company's applicants were originally trained as scientists and received a second degree in computer science. "Their programming experience is usually based on rather simple course projects, which do not prepare them for working in a department that has a large internal informatics development environment," Chamberlin says. He asserts that it takes six months before such individuals' productivity justifies the time spent in training them.

Thodoros Topaloglou, director of gene expression data management at Gene Logic's Berkeley, Calif., location, tries not to use the word bioinformatics when he advertises positions because it attracts too broad a range of applicants. "We try to be more specific in what we say we want," he says. "My group is mostly data management. We are dealing with large amounts of data, so we are looking for strong database and programming skills. My colleagues who work in the front-end development need software development, visualization, and user-interface skills. My colleagues who work in the interpretation of the science in the data need more in-depth understanding of biology."

Covitz says he receives "floods of résumés" when he posts bioinformatics positions. However, he says that most applicants don't qualify under his selective criteria. "Someone who has gotten a Ph.D. in molecular biology and took a few computer science classes on the side does not qualify in my selection process. There are a lot of people like that. That's a perfectly valid first step to take if you want to enter the realm of bioinformatics. The distinguishing feature that I look for is someone who has actually used that computer science ability that they've acquired to solve a biological problem."

Covitz believes that scientific training should be the foundation of any bioinformatics professional, because it fosters critical thinking skills. "There's a certain basic skill set you have to have--mainly critical thinking ability--which comes with scientific training, and that's generally applicable to bioinformatics as well as laboratory-based biology," he says.

"Some people claim that you have to start with biology," Fuchs says. "I'm not convinced about that. I think you can start with either biology or computer science. You have to build a solid foundation in one particular area and then complement it with an almost equal level of skills in the other area. What I'm looking for in résumés is clear evidence that somebody has identified a biological problem and has been able to identify and implement a working computational solution to that problem. The major shortcoming I see when I look at résumés is that there's just no evidence that people have the ability to go out, really break down a problem, and identify the best possible solution."

Lots of computing power, as demonstrated by this computer bank at Celera Genomics, is required in bioinformatics.

Topaloglou places the emphasis on computational skills. He says biological scientists express the problems, which then need to be formulated as computational problems. "A good understanding of biological language is important for someone to get up to speed, but the core expertise in my view is a good understanding of computational principles--that is, algorithms, databases, statistics, data visualization, and solid programming methodology," he claims.

PPGx's Sheridan believes that two types of preparation are needed. People who have "domain expertise," or an understanding of the science, are still needed. "We're finding a lot of people who are really hybrids. They've got the science background, but they also have an intense interest in computer science. They've taken courses in both disciplines." In addition, he says, people are being drawn directly from computer science without a natural sciences background. "We're not asking them to become domain experts in the science. They don't need to. We're relying on them to bring their expertise in how you handle data, display data, visualize data, and report data."

Sheridan agrees that finding people who appreciate both biology and computer science is difficult. He doesn't think that the proper background can be gained simply by scientists "dabbling in computer science." He suggests that people who are interested in bioinformatics get formal training in both science and computer science.

However, Sheridan does not advocate a narrow training in bioinformatics. "You're talking to me about bioinformatics because it's now a hot area. Ten years ago, we wouldn't have been having this conversation," he says. "Anybody who is going to train for a career in bioinformatics specifically might be in trouble 10 years from now. I'm saying that you need to get a very broad-based training so you can respond to what the market wants you to do."

Covitz admits to being biased by his own background; he has a Ph.D. in molecular biology. During his postdoctoral training, he made a conscious decision to acquire computer skills and apply computer-based approaches to biological problems. He says students today have opportunities to make such decisions even earlier in their education.

However, Covitz is skeptical of some of the educational programs being offered. "What I'm generally not as enthused by are people who have a shallow background in both computers and biology," Covitz says. "There are some programs out there touting themselves as bioinformatics programs that do that. They do a lot of coursework in biology but not a lot of lab work. They do some coursework in basic computer science and computer programming, but they don't go too deeply into it. What you end up with is someone who's mediocre in both fields."

Leach says CuraGen's bioinformatics group is composed of biologists, biophysicists, computer scientists, and software developers. The group has a "critical mass of biologists," Leach says, so finding cross-trained individuals is not as important as it once was. Members of the bioinformatics support team have put together a course to teach biology to computer scientists. "We can take raw software developers who have no knowledge of biology, teach them about the human genome, teach them about DNA, teach them about transcription and translation and proteins and PCR--the basic crude set of techniques--so they can understand where they fit in terms of biological data, their role in what they're doing in terms of analyzing it."

Sheridan says the bioinformatics effort at PPGx involves employees with a wide range of backgrounds. For example, PPGx hires people who understand biology and can speak the language of bench scientists. But it also employs people with experience in developing software for large-scale inventory management. "The old days of some scientists building software in the bioinformatics arena are long past. The problems are too large, the amount of data is too large for the approach to bioinformatics software to be dealt with by people who don't have a deep understanding of the way to optimize computer systems."

People with training in both biology and computer engineering are important for making data accessible. "That's where their value comes in: being able to work with the scientists to help them organize the data, to work with the people who are mining things in the public domain, to bring the data in and load it in a fashion that it's retrievable in a useful way," Sheridan says. "If you just relied on engineers who didn't really understand the problem, you could wind up with what we call a data prison."

For people who will be mining data rather than generating software, CuraGen's Leach believes that familiarity with the various biological and chemical databases is the appropriate preparation, including their structure, the information they contain, and how they cross-reference each other. In addition, he says a crude understanding of the algorithms used to analyze data is desirable--not just the output but how they work.

With all this talk of biologists, is there a role for chemists? Yes, according to Sheridan. "A lot of the things we look at, especially in drug development, are the interactions of complex molecules. Bioinformatics has an incredible role in terms of developing tools for predicting toxicology and other types of capabilities that would certainly be a highly attractive career for somebody with a background in chemistry," he says. "You're looking at complex interacting systems at the molecular level and trying to predict--based on structure--function and ultimately alterations in biological effect. I absolutely think there would be an ongoing role for people trained in chemistry and biochemistry."

The demand for bioinformatics professionals can be expected to remain high for the foreseeable future. Kal Ramnarayan, vice president and chief scientific officer at Structural Bioinformatics in San Diego, says: "My own feeling is that, for the next two to four years at least, the demand is going to be consistent. But after about two years or so, what might happen is that there might be a lot of standards in handling data that will have arisen, resulting in the availability of turnkey solutions. At that point, the field will be looking for bioinformatics people who have strong expertise in identifying targets--exceptional biology experience but not that much programming skill. This will be the result of two factors: tools that have matured, resulting in companies forced to adapt available tools rather than develop their own, and the pressure on companies to produce results (for example, identify novel targets and pathways) rather than tools."

Celera's Giacalone expects that the demand will increase in the future. "Computers aren't going away," she says. The amount of data generated by biology will continue to require mining with a computer. "I see the field as continually growing and developing," she says.

[Previous Story] [Next Story]

Top

Educational programs in bioinformatics or computational biology are cropping up at many universities. A search of the World Wide Web yielded the following list, which is not intended to be comprehensive.

Boston University offers M.S. and Ph.D. degrees in bioinformatics (http://www.bu.edu/bioinformatics).

Carnegie Mellon University, Pittsburgh, offers a B.S. in computational biology (http://www.bio.cmu.edu/Programs/Undergraduate/compbio.html).

George Mason University, Fairfax, Va., offers a Ph.D. in computational sciences and informatics (http://www.scs.gmu.edu/Academics/PHD.html) and a master's of new professional studies in bioinformatics (http://www.scs.gmu.edu/Academics/Master.html) through the School of Computational Sciences.

Georgia Institute of Technology offers Ph.D. and professional master's degrees in bioinformatics through the School of Biology (http://www.biology.gatech.edu/bioinformatics/welcome.html).

Johns Hopkins University offers a Ph.D. through the Program in Computational Biology in the Institute for Biophysical Research (http://www.jhu.edu/~ibr/bwf.html).

Keck Graduate Institute, Claremont, Calif., offers a professional master's degree in applied bioscience (http://www.kgi.edu).

Rensselaer Polytechnic Institute, Troy, N.Y., offers B.S., M.S., and Ph.D. degrees in bioinformatics and molecular biology through the department of biology (http://www.rpi.edu/dept/bio/info/bioinformatics.html).

Stanford University offers M.S. and Ph.D. degrees in medical information sciences, including bioinformatics, through its interdepartmental program in the Stanford University School of Medicine (http://camis.stanford.edu/).

University of California, Irvine, offers M.S. and Ph.D. degrees in biomedical informatics through the department of information and computer science (http://www.ics.uci.edu/~biomed/).

University of California, Los Angeles, offers an undergraduate concentration in bioinformatics through its Cybernetics Interdepartmental Program and a certificate in bioinformatics from the Bioinformatics Interdepartmental Program for candidates pursuing a Ph.D. in a participating department (http://www.bioinformatics.ucla.edu).

University of California, San Francisco, offers M.S. and Ph.D. degrees through its Graduate Program in Biological and Medical Informatics (http://www.mis.ucsf.edu/).

University of Nebraska Medical Centerand University of Nebraska, Omaha,jointly offer M.S. and Ph.D. degrees in bioinformatics (http://www.isqa.unomaha.edu/bioinformatics/).

University of Pennsylvania offers B.S. (department of biology, department of computer and information science, or department of mathematics), M.S. (Program in Biotechnology), and Ph.D. (department of computer and information science, department of biology, or Biomedical Graduate Studies Program) degrees in computational biology (http://www.pcbi.upenn.edu/).

University of Southern California offers a professional master of science in computational molecular biology and a Ph.D. in computational biology and bioinformatics (http://www.hto.usc.edu).

University of Washington and Fred Hutchinson Cancer Research Center, Seattle, offer a graduate program in computational molecular biology (http://bozeman.genome.washington.edu/compbio/).

W. M. Keck Center for Advanced Training in Computational Biology (affiliated with Carnegie Mellon University, University of Pittsburgh, and Pittsburgh Supercomputing Center) offers a Ph.D. in computational biology (http://www.cs.pitt.edu/keck/Frames/Welcome.html).

[Previous Story] [Next Story]

Top

Chemical & Engineering News
Copyright © 2001 American Chemical Society