Working on an SGI Altix system located at NCSA, the Stony Brook team recently achieved computer simulations that offer insight into the mechanics of HIV protease, a molecule that slices the pre-HIV protein chain into pieces that ultimately evolve into a mature virus. By modelling how HIV protease works across time, researchers hope to determine how best to target it with medicines that could stop the molecule from doing its job and thus prevent the HIV virus from developing altogether.
Dr. Carlos Simmerling, associate professor at Stony Brook's Center for Structural Biology, guided the research team, which successfully simulated how the HIV protease changes between two forms that already have been determined through experiments. More important, however, is that the group was able to capture the protease in a third, fully open state - one that had previously been hypothesized but never directly observed.
When the structure is open, it is vulnerable to inhibitor drugs that can bind with the molecule and render it harmless. But pristine, crystallized examples of the molecule have only shown the molecule as either closed or barely open, leaving no room for inhibitors to enter. Until now, this lack of knowledge about the open form has hindered the progress of HIV drug developers, even as many current drugs specifically target HIV protease.
"We determined that if we knew how HIV protease opened, we could better identify a new and potentially more sensitive drug target", stated Dr. Simmerling. "The challenge has always been simulating it long enough to see the transformation at work. The structure just doesn't open frequently enough to easily measure it. And when it happens, it happens fast." Though no specific measurements yet exist, scientists estimate that the structure remains open for less than a millionth of a second.
The Stony Brook researchers' simulations are the most extensive ever done on HIV protease. Individual simulations modelled only 50 nanoseconds of behaviour - less time than it takes a beam of light to travel 50 feet - but they still last long enough for Dr. Simmerling and his colleagues to model HIV protease in unprecedented detail. "We can model the full change between the known structures with very high accuracy", Dr. Simmerling explained. "We can also see how it opens, and where a drug molecule binds to the protease and causes it to close. And then we can reverse that process, and the protease opens again. These are all things that experiments have not been able to show us."
Such reliability suggests the simulation will prove helpful in testing the potential efficacy of new drugs, and in understanding how variations of HIV seen between different patients can change a drug's behaviour. "HIV is a very adaptive virus that mutates easily, which can reduce the benefit of a drug", Dr. Simmerling stated. The Stony Brook research, however, focuses not solely on the structure of HIV protease but on its mechanics. "If we target its shape with a drug, then any change in shape can diminish the drug's effectiveness. But if we can target its job, the shape doesn't matter as much and it will have a harder time evading the drugs."
Dr. Simmerling's group developed the simulations using AMBER, a molecular dynamics application developed in part by Dr. Simmerling's lab. The team typically used 64 processors of NCSA's SGI Altix 3700 Bx2 system for each of the simulations, leaving the remainder of the 1024-processor resource available for other projects. The work, which continues to this day, began last fall. Earlier this year, the breakthroughs were chronicled in the Proceedings of the National Academy of Sciences and the Journal of the American Chemical Society.
The team also was assisted by Dr. Roberto Gomperts, principal scientist at SGI, who helped optimize AMBER code to accelerate the simulations, and even contributed code for the project. "Roberto's contributions made the simulations a lot faster", stated Dr. Simmerling. For the researchers, achieving such a detailed picture of HIV protease dynamics was a milestone that was a long time coming, and the broad availability of SGI Altix systems made it easier. "I don't know that I would have tried this a couple of years ago", stated Dr. Simmerling. "The resources were out there, but they were too precious."
But today, thousands of SGI Altix systems worldwide save scientists weeks, months and even years by accelerating some of the most daunting computational problems ever attempted. The Stony Brook HIV protease simulations took 20.000 CPU hours on NCSA's Altix system; in real time, the project lasted about three months. In contrast, Dr. Simmerling estimates that it would have taken more than a year to complete the work using the Simmerling lab's Linux cluster. "It would have been at least six to seven times slower than the Altix, and the cluster doesn't scale as well", he stated. "I wouldn't have done it for such a risky project."
SGI Altix is one of the most powerful platforms for running high-performance computation applications such as AMBER. For instance, in tests designed to assess the viability of the project, SGI's Dr. Gomperts and Dr. Simmerling's team determined that the same AMBER simulation running on a 128-processor Xeon cluster at NCSA and on 128 processors of an Altix Bx2 system would take more than three times longer to complete on the cluster. Relying on the cluster, the team would need 26 days to simulate 50 nanoseconds of the test model. But on the Altix system, they would need just 8 days.
"Demanding computational chemistry applications like AMBER require exceptionally low latencies to derive the most effective benefit from CPUs, and this is precisely what SGI Altix provides", stated Dr. Gomperts, who also supported the Stony Brook research project by running large calculations on one of SGI's own 512-processor Altix systems. "The Altix platform saves valuable time for researchers by allowing them to map massive data sets in memory, or to keep them in file buffers for ready access. When combined with the floating-point power of Intel Itanium 2 processors, these advantages mean that even the largest calculations can be performed as a matter of routine on Altix."
The Altix family leverages the built-in SGI NUMAlink interconnect fabric, which allows global addressing of all memory in the system and delivers data up to 200 times faster than conventional interconnects. For the first time, more complex data sets and complete work flows can be driven entirely out of memory, enabling productivity breakthroughs that traditional Linux clusters or repurposed UNIX servers can't achieve.
Altix systems offer breakthrough flexibility and configurability, scaling to up to 512 processors per node. Based on a 64-bit Linux operating environment, the Altix family is uniquely capable of independently scaling Intel Itanium 2 processors, shared memory and/or I/O on a single, standard chassis with different expansion modules, providing optimal resource usage for demanding technical applications. More news on the SGI Altix can be found in the VMW June 2006 article First-ever simulation of functioning organism spawned by ingenuity of Illinois researchers and power of SGI Altix.