Biologists Find Speed, Imaging Powers Of Supercomputers Key To Research

At an NIH workshop, they discover how advanced computation can improve their vision of life processes PITTSBURGH -- Time was, all that a biologist ever wanted was a new microscope. But times change. A decade after the first supercomputers began to open up new vistas for physicists and engineers, life scientists are beginning to use these powerful machines to work on everything from high-resolution protein reconstructions to detailing how the brain functions. "Anything I can do with a supercomputer, you can do with a pencil and paper," says Scott Berger, a neurobiologist at Cornell University Medical College. "But it would take many lifetimes to do it." For John Gilbert, a biology graduate student at the California Institute of Technology in Pasadena, one supercomputer is worth a million calculations. Gilbert uses an X-ray microscope to study fibroblasts: thin, flat cells that develop into connective tissue. But even with this modern piece of scientific equipment he had a hard time conceptualizing their structure. Any time he wanted to home in on one part of the fibroblast or look at it from a different angle, he had to rely on his imagination or the limited resolution afforded by available light. Enter the supercomputer. It transforms the diffraction pattern from Gilbert's X-ray microscope into a clearer image by using numbers. Each number represents the light intensity at a particular point. In seconds, the computer performs those million calculations and refigures the image the way it best suits Gilbert's investigations. "You often have a large amount of numbers that start off as a picture," Gilbert explains. "But when I want to ask a question about the structure -- like how big is that structure, and does it have parameters that match some model? -- then I need to extract that information so I can use it. Before [on a smaller computer], I've had to spend all my time to get all the data together. That meant that I didn't have any time to go further, to analyze that data." The National Science Foundation supports four supercomputing centers that offer programs or services of interest to biomedical researchers. Here are their names, the equipment on hand, and the person to contact: Cornell Theory Center Cornell University Hoy Road Ithaca, N.Y. 14853-3801 Two IBM 3090/600Es Cindy Harwzinske, (607) 254-8713 National Center for Supercomputing Applications University of Illinois, Urbana/Champaign 605 E. Springfield Ave. Champaign, Ill. 61820 Cray XMP/48, Cray 2 Clint Potter, (217) 244-1958 Pittsburgh Supercomputing Center 4400 Fifth Ave. Pittsburgh, Pa. 15213 Cray YMP/8-32/David Deerfield, (412) 268-6102 San Diego Supercomputer Center University of California, San Diego San Diego, Calif. 92138-5608 Cray XMP/48 Wayne Pfeiffer, (619) 534-5000 Gilbert, 32, and Berger, 28, represent a new generation of computer-savvy life scientists. According to the National Science Foundation, last year biologists made up only about 10 percent of the 5,500 users at the foundation's five supercomputer centers. But that percentage is sure to grow, say computer experts, as life scientists realize how many ways supercomputers can help them. Already, notes Thomas Weber, director of advanced scientific computing at the National Science Foundation, scientists who want to obtain biological simulations appear to be the fastest growing specialty of users at the centers. "Biologists have, both generally and individually, begun to realize that computational biology is very, very important to their science," says Weber. One sign of that change is the three-year, $2.2 million grant that the National Institutes of Health awarded to the Pittsburgh Supercomputing Center, one of the NSF centers. The center is a five-year-old joint project of the University of Pittsburgh and Carnegie Mellon University. "For a lot of reasons, they [biologists] are being forced to modify their system or use math to evaluate their data," says David W. Deerfield II, biomedical staff specialist at the Pittsburgh Center. "Ten years ago, they'd want a new microscope. Now they are buying a PC." Deerfield and his colleague, Hugh Nicholas, Jr., have organized a series of workshops that show life scientists the advantages of machines with far greater computing power. For Berger, the computer's major attraction is its ability to manipulate numbers. In examining how the brain controls blood pressure and heart rate, Berger looks at the interaction between the right and left sides of the brain stem. He measures how the two sides react to different stimuli. A single response -- one side's electrical impulses generated to a stimulus -- represents one data set, 16 million bits large. Each set of calculations takes more than a hour on the Sun computer that he usually uses. "We can do each of those on the Cray in a couple of seconds," says Berger, referring to the machines that helped to create the field of supercomputing with their blinding speed. Add to that data set 10 more that all need to be averaged, then throw in the need to view the image from different angles, and access to a supercomputer becomes the difference between doing and not doing the experiment. The June workshop here taught 18 physicians, biologists, and other scientists how supercomputers could add a new dimension to their work with electron microscopes. First the computer converts the electron microscope image into a matrix of numbers. Each number represents the light intensity at a particular grid point, or pixel. Then the computer can manipulate the matrix to create an enhancement of the original. Theoretically, a less powerful computer could complete the necessary calculations to achieve the same high resolution. But the task would take far too long to be practical. "In order to work, a project has to have a pace [of progress]," says Ross Smith, a cell biologist at New York University Medical Center. The supercomputer offers other advantages as well. To the scientists, the image generated is much more than just a pretty picture. "Ultimately, you can look at it [a problem] with your eyes or make numbers," says Dean Hillman, a professor of physiology at New York University, who has worked in image processing for about 20 years. "But by seeing it, you can begin to conceive of how a particular molecule can act. [A picture] allows you to integrate information which is, in effect, multidimensional, where with numbers, the mind can handle only one dimension at a time." Frank Harris, executive officer in biological, behavioral, and social sciences at NSF, cites the Human Genome Project as another area in which advances in biology will spur developments in computational science and vice versa. "The science is rushing up from behind and, in fact, will force further development in computer technology," he adds. NSF is phasing out the von Neumann Center at Princeton University, one of the five orginal supercomputing centers. But the four others are being upgraded and will have twice the capacity of the original five. Plus, there's been a 90-fold increase in programs available to researchers since the centers were formed in 1985. But scientists far from those centers still need better access to that computing power, which can be daunting to a first-time user. "The skepticism for scientists is unfamiliarity and being a very large distance from it," says NYU's Hillman, who thinks scientists are intimidated, but only because they don't know what a supercomputer has to offer. "Once they are introduced to it, you'll find it being used." Bridget Carragher, a neurobiologist from the University of California, San Diego, thinks it will take more than an introduction. Fellow scientists would be happy to hook up with a supercomputer "because they'll use what is there that makes their research easier," she says. "But it has to be made accessible. Supercomputer centers have not been accessible before." The centers say that has changed. Now scientists need only be able to sign onto Internet, a computer networking system available at universities and medical centers. Then they request computer time from one of the four active centers: the Pittsburgh Center; the National Center of Supercomputer Applications in Champaign, Ill.; Cornell University's Center for Theory and Simulation in Science and Engineering in Ithaca, N.Y.; or the San Diego Supercomputer Center. The centers themselves choose among requests for small amounts of time, although there is talk that NSF may begin to parcel out computer time as part of an individual grant. According to Deerfield, it is "relatively easy" for others to get a few hours per year on a supercomputer by applying to one of the centers. Bigger requests are more difficult to honor, he says. But he also points out that a Cray can accomplish in one second what the typical personal computer requires a day and a half to do, so researchers may require far less time then they realize. There are caveats with such sophisticated number and image handling, however. The vivid, three- dimensional pictures generated with the Cray can be at once a researcher's powerful ally and an alluring deception, says Deerfield. "One question is whether the pictures are better than the scientific content, whether the picture is actually better resolution than the data," he says. Such caveats may some day curb the enthusiasm for supercomputer usage. But for now, the benefits of these machines are hard to pass up. As typical of its other workshops, the Pittsburgh Supercomputing Center had five applicants for each of the June class's 18 slots. The hope is that the 18 workshop participants will collaborate with others. One such collaboration, between Cesar Fermin, a cell biologist at Tulane University, and workshop participant E. Loren Buhle, a cancer researcher at the University of Pennsylvania, will harness the super-computer's imaging prowess to understand otoconia, microscopic structures in the inner ear. Before, Fermin says, he was limited to what he could do with his PC. "It was Little League, and he [Buhle] is in the big leagues," says Fermin. "I would like to take advantage of what he knows. Those of us who grew up without computers need to re-educate ourselves." It's collaborations like that between Fermin and Buhle that make all the work put into setting up such workshops worthwhile, say the organizers. "[Biologists] know we're out there, but they are not sure how they could use us," Deerfield says. "That's one of the biggest problems -- just getting people to try it for the first time." Steve Twedt reports on science for the Pittsburgh Press.

Biologists Find Speed, Imaging Powers Of Supercomputers Key To Research

Interested in reading more?

Become a Member of