Emad Tajkhorshid is a busy man. The University of Illinois professor and researcher holds appointments in the College of Medicine at Urbana as a professor of Medical Pharmacology, in the departments of bio- chemistry and pharmacology and at the interdiscipli- nary Beckman Institute, where he is deputy director of the National Institutes of Health Resource for Macromolecular Modeling and Bioinformatics. The thread connecting all of these activities is his interest in the fundamental processes of life—how do cells, mem- branes, and molecules function and interact?
“I love to know how biomolecules work, that’s what I really want to know,” he says, brimming with energy and enthusiasm. “More specifically, my research group is mostly interested in the function of membrane pro- teins and membrane-associated phenomena, which is a very fundamental aspect of cellular function.
“Because membranes are separating the cell from the environment, you have to exchange signals and sub- stances back and forth through that membrane. Only really, really small molecules can make it through the membrane itself,” he continues. “Everything else—hormones, sugar, water, vitamins, everything—you have a specialized type of protein that is involved in the transport. Selectivity is the key here–the cell wants to be able to regulate everything independently.”
“About 50 percent of the energy of the cell and maybe 50 percent of the genetic material of the cell is related to membrane transport. It’s a very fundamental, impor- tant process in living cells.”
And because many diseases—such as Alzheimer’s and Parkinson’s—are associated with malfunctions of these transport proteins, understanding them could lead to new treatments.
Tajkhorshid studies these transport phenomena not with a microscope, but with computer modeling and simulation. It’s a technique that Tajkhorshid’s colleague Klaus Schulten often refers to as the “computational microscope.”
“Many of the details we are interested in cannot be captured experimentally,” Tajkhorshid explains. “There is no way for any physical experiment to have physical and temporal resolution simultaneously.” By using powerful supercomputers and simulation software that has been honed over the past two decades, Tajkhorshid’s research group can examine membrane processes with “sub-angstrom special resolution and any temporal resolution we wish!”
Increasing the power of computational resources
Two forces are operating in tandem to advance biomol- ecular research like Tajkhorshid’s: the increasing avail- ability of high-resolution crystal structures for a variety of proteins, and the rapidly increasing power of avail- able computing resources, both in individual research labs and at federally funded national computing cen- ters such as the National Center for Supercomputing Applications (NCSA) at the University of Illinois.
“The number of membrane proteins for which we have a structure is increasing, and this is really exciting,” Tajkhorshid says. “Crystallizing proteins is very diffi- cult—they are floppy and wobbly, so it’s very difficult to crystallize them. But over the last three or four years, there have been quite a few different families of pro- teins crystallized, which gives us a real opportunity.”
Those crystal structures are like static photos of pro- teins. “But of course in order to understand their func- tion you have to know how they move around,” he says. The crystal structure just shows the runner poised at the starting line, but Tajkhorshid wants to watch the runner in motion, accelerating down the track, over the hurdles, and across the finish line.
And that’s where computer simulation comes in.
“We start with solved structures, crystallized structures or structures determined through nuclear magnetic resonance. And we essentially bring them to life,” Tajkhorshid explains. “We bring them to their natural environment of water and the lipid bilayer, and then we bring them to body temperature, and they start moving and relaxing in their own natural environment.”
But even with computer clusters that harness multiple powerful processors, these complex simulations are far from instantaneous.
“Computationally, these things are very expensive,” Tajkhorshid says. “Transporters essentially get engaged with their substrate, and the substrate induces really big changes in the transporter. You have to be able to describe all these stepwise changes that occur. You need to describe all of the tiny bits of the dynamics to describe the dynamics of the entire pro- tein, and that’s why these simulations are so expensive and slow.
“In a recent paper we reported five simulations, each about 5 nanoseconds, and each simulation took about 40 days of 24/7 calculation on 128 processors,” he adds. “When you show this to people in an animation, it lasts about 30 seconds, and you have to really explain how long it took to get to this short animation!”
But even as simulations became more detailed and described longer stretches of time, they still fell short. Just a few years ago, simulations could only capture femtoseconds of protein behavior. “People who were interested in the biology couldn’t take us very seriously, because our time scales were too short,” he says. “We had been up until two or three years ago limited to really limited motions of proteins. You couldn’t really simulate anything meaningful.”
Increased investment in computing hardware and infrastructure has enabled researchers like Tajkhorshid to simulate biomolecular systems for microseconds—still a very short window, but sufficient to provide much more biologically relevant results.
“All these advances in computer science and hardware, and the very nice investment of the National Science Foundation in hardware at supercomputing centers, have made a huge difference in terms of the biological- ly relevant events that we can describe,” he says. “Just this year we have about 10 papers in which we are describing really relevant, really interesting events.”
“We are providing the most detailed view for molecular medicine, in a way,” Tajkhorshid says. “We have all the elements necessary—from a computational point of view, a structural point of view—to describe how these proteins function at atomic detail. So then we can tell what happens with this mutation—say with cystic fibro- sis, what makes this pump not functional. We’re going to provide the most molecular view of physiology and disease in human beings.
“I think this is going to be really exciting for students in biomedical fields and hopefully it provides more insight in developing treatments. It’s going to give a new dimension to how we think about diseases.”