Wednesday, June 28, 2000

Researchers Turn to Computer Models to Learn What Experiments Can't Teach By FLORENCE OLSEN Klaus Schulten wields an electronic probe, gently pulling and poking the protein molecules displayed on his computer screen until he feels the force of their resistance. This rare interaction, a novelty today, is made possible by new scientific software and high-performance computers, says Mr. Schulten, a professor of physics who is director of the Theoretical Biophysics Group at the University of Illinois at Urbana-Champaign. With powerful computers running programs that model the building blocks of life, Mr. Schulten and other researchers are making new discoveries every day about the mechanics of living cells. "Computational methods are appropriate for trying out various theories that are difficult to come up with experiments for," says John Edward Stone, senior research programmer in the group headed by Mr. Schulten. He adds that computers have the advantage of not "messing up" an experiment in the same way that experimental methods can contaminate whatever is being observed. "You can be an observer of the simulated world without contaminating it." The probe, which resembles a computer joystick, is a "force-feedback" device connected to Mr. Schulten's computer. He can "push" or "pull" on individual atoms or groups of atoms within the computer model -- essentially using the probe to input data into the modeling program. "Just like you try to repair a broken watch," Mr. Schulten says, "you can manipulate a protein" by tinkering with its biomolecular parts. Guided by interactive molecular-dynamics software, scientists develop an intuitive sense of how specific protein molecules behave. With such methods, Mr. Schulten says, scientists are acquiring new information about the molecular mechanics of proteins such as titin, an important component of muscle tissue. They can both see and feel how the protein "unfolds" when several hydrogen bonds in the protein are broken. Such tools and methods will be needed now that researchers say they have nearly completed the map of the human genome, and as scientists rush to discover the structures and functions of all of the proteins encoded in it. Experimental and computational biology are complementary, Mr. Stone says. Computational scientists are happy when the results of their protein simulations match fairly closely the results from experimental studies of proteins. When that happens, he says, "it helps reconcile our theories with reality." Software developed at Illinois to simulate molecular dynamics and display those results graphically undergoes continuous revision and refinement, he adds. The process of improving the accuracy of molecular-modeling software involves thousands of researchers, Mr. Schulten says. And yet the life processes that such software tries to model are too complex for its developers "to claim we know what the exact results should be." The simulation and visualization software that his group has developed is available free, with its source code, to any person or institution that registers to use it. The programmers who wrote the software wanted it to be free, Mr. Schulten says, and he acceded to their wishes. "I couldn't do anything if my programmers weren't really excited about what they do, and many young people today believe in free software." To pay for its software-development activities, the research group uses a fraction of the $1-million a year that it receives from the National Institutes of Health. Oftentimes, academic software gets "stale," because its developers fail to keep up with cutting-edge research in their field, says Mr. Stone, the senior programmer. But he vows that will not happen to the biophysics software developed by the Illinois group, whose members are biophysicists, physicists, computational chemists, computer scientists, and numerical analysts. Just since January, more than 4,500 scientists have registered and downloaded versions of the software. The programs are NAMD, which stands for Not Another Molecular Dynamics program; VMD, the group's Visual Molecular Dynamics program; IMD, its Interactive Molecular Dynamics capability; and BioCoRE, the Biological Collaborative Research Environment, which structural biologists can use when they want to collaborate over the Internet. The software programs, which are written in the languages C++ and Java, run on very large computers with 200 processors, or in some cases, 1,000 processors. Even using such massive computers, it can take scientists months to run the computations required to simulate in a computer what happens in a microsecond inside a protein molecule made up of anywhere from 10,000 to 200,000 atoms. A microsecond, Mr. Schulten says, "is a very short time for a biological cell." Like most biophysicists, he would like to simulate a full second in the life of a protein molecule -- "that's one million times longer." But even the most powerful computers today are not capable of such simulations. Klaus Schulten says he doesn't dwell on such limitations: "It doesn't help me," he says. After all, Mr. Schulten says, he is "not a dreamer." He is a physicist who simply wants to use "the best tools available." Background article from The Chronicle: The 'Open-Source Movement' Turns Its Eye to Science (11/5/1999) Easy-to-print version E-mail this article

A.C.L.U. lawsuit charges that Michigan's merit-scholarship program is discriminatory Donor pledges $80-million to translate genomic map into cures for disease Expanding its reach in higher education, Kaplan buys Quest, a chain of commercial colleges College Board parts ways with the head of its for-profit Web site after just 3 months Hebrew U. of Jerusalem accedes to religious activists and turns over ancient bones for burial In Russia, a student's impertinent letter to the president costs her a career in medicine Researchers turn to computer models to learn what experiments can't teach Studio-art instructors try teaching online, and say it works