For the first time, researchers simulated a complete, functioning organism at the atomic level. How NCSA's SGI Altix system helped them make history.

Picture an object with a million moving parts. Now imagine building a model to simulate that object - one that reveals how each part interacts with the other, every one-millionth-of-a-billionth of a second.

No, it's not rocket science. But it just might be harder.

For nearly as long as biophysicists have relied on computers to study the dynamics of molecules, they've hoped to replicate a functioning life form at the atomic level. Yet even the simplest of these organisms proved too complex to simulate in total. But now, thanks to the efforts of three Illinois researchers, a highly scalable molecular dynamics application, and one very powerful computer, all that is history.

At the University of Illinois at Urbana-Champaign, two graduate students and their professor have achieved what until now had been patently unachievable. They generated the world's first atomic-level simulation of a simple, but complete, functioning organism. And they did it on an SGI® Altix® system.

A research team led by Professor Klaus Schulten, head of the Theoretical and Computational Biophysics Group at Illinois, simulated a plant virus with as many as 1 million moving atoms. Dr. Schulten worked with grad students Peter Freddolino and Anton Arkhipov on the year-long project, which was reported in the March 2006 issue of the scientific journal Structure.

Intricate mechanisms

Satellite Tobacco Mosaic Virus
The world’s first atomic-level simulation of a functioning organism.

The smallest natural organisms known, viruses contain intricate mechanisms for infecting host cells. The Illinois team simulated one of the tiniest and most primitive viruses in an attempt to recreate the process of infection and propagation. The satellite tobacco mosaic virus - it's called a satellite virus because it relies on a host cell and a primary virus to reproduce - attacks tomato and tobacco plants throughout the US, leading to mosaic-like discolorations.

While the virus attacks plants, the researchers predict that someday, drugs for animals or even humans may be designed and refined with the help of computer-based simulations like the one developed in Illinois.

The satellite tobacco mosaic virus is so small that biologists refer to it as a particle. Nevertheless, simulating the organism and how it functions holds tremendous promise for medical research. "It allows us to see how the virus assembles and disassembles," notes team member Peter Freddolino. "Because assembly and disassembly are two of the key steps in the viral life cycle, understanding these events could lead to the development of drugs designed to attack them at these vulnerable points."

After several months of careful planning and preliminary research, the team secured time on Cobalt, an SGI® Altix® 3700 Bx2 system located at the National Center for Supercomputing Applications (NCSA). It proved a powerful, efficient resource. Despite the complexity of the project, the researchers needed just a fraction of the NCSA's 1,024-processor Altix system: Most simulations used 256 Intel® Itanium® 2 processors and 128GB of total memory, leaving the rest of the NCSA system available for other projects. The team's scalable molecular dynamics code, a Gordon Bell Award-winning application called NAMD, segments tasks across processors and memory, enabling simulations to draw the most of as many of processors and as much memory as they require.

The entire simulation was completed in 50 days. Had the researchers relied on today's desktop computer systems, they wouldn't have finished until 2041.

Biological reverse-engineering

Tobacco Virus Simulation
Simulations of the capsid without the RNA core and the whole virus.

The project is also the first successful case of biological reverse-engineering of a complete virus. "This is on the highest end of what is feasible today," says Schulten. "The approach is something that we learned from engineers: Reverse engineer the subjects you're interested in and test fly them in the computer to see if they work in silico (or simulated on a computer) the way they do in vivo (in the body). Naturally, deeper understanding of the mechanistic properties of other, more complicated viruses will eventually contribute to public health and medicine."

The entire simulation covers just over 50 nanoseconds of time, or 50 billionths of a second. Still, it was enough time for Schulten's team to make several conclusions about the virus. Among them: although the organism appears symmetrical, it actually pulses in and out in an asymmetrical pattern.

As it turns out, the Illinois team's simulated findings support observations made by other researchers in traditional "wet lab" work. Those earlier observations, however, left scientists wondering what caused the viral behavior - something that remained a mystery until today.

They accomplished this by studying the virus not just as a whole, but as the sum of its separate components - its protein shell and nucleic acid core. "The going theory was that the protein shell pulled the entire thing together and was responsible for its assembly," explains Freddolino. "But the simulation showed that the nucleic acid was even more important for keeping the virus together."

When simulated as a whole, the virus appeared stable and symmetrical. But after removing a component, the researchers saw how vital the nucleic acid was to the structure of the virus. "The protein shell just fell out without the nucleic acid," recalls Freddolino. "That told us a lot."

"The analysis of the results required originality," says Arkhipov, who leveraged his physics training to analyze the diffusion of ions and other phenomena. "It's interesting to see how each part of the structure moves a little bit on its own, and how that affects its symmetry."

An invaluable resource

NCSA's large Altix system allowed the team to focus more on science than computing. "The ideal situation is to work with a powerful computing platform that provides output quickly and with minimal disturbance. In this way, the underlying science is the focus of the effort. NCSA provided exactly that," added Schulten, a long-time NCSA user.

"The Altix platform has excellent single node performance and a very efficient MPI implementation," said Freddolino, "and both of these factors made the system very useful in performing our work efficiently."

"It's incumbent upon centers like NCSA to make the most of the federal government's technology investments by working closely with scientists like Professor Schulten as well as entire communities of scientists," said Thom Dunning, director, NCSA. "We have developed new ways to allocate supercomputing resources, such as our large-scale SGI Altix deployment, to give scientists what they need in order to make incredible breakthroughs like the simulation of an entire living thing."

Image courtesy of NCSA and University of Illinois at Urbana-Champaign