Simulations help researchers in the fight against diabetes
by Gregory Scott Jones
Peptides, or chains of various amino acids, are shifty creatures. Their forms move, jump, twist, and change shape, all of which add to the myriad difficulties involved in peptide research (not to mention the fact that the scale on which they operate is minute). They just don’t seem to want to sit still. Yet understanding them is instrumental in putting them to good use—for example, in the treatment of diabetes.
That’s the goal behind recent simulations of exenatide, a peptide consisting of 39 amino acids. Because of its ability to regulate glucose metabolism and insulin secretion, exenatide is highly lauded as a type 2 diabetes treatment option and is currently marketed under the name Byetta. Type 2 diabetes accounts for 90 to 95 percent of all diabetes cases in the United States, and the Centers for Disease Control recently labeled its rise among the American public an epidemic.
To better understand exenatide’s properties and strengthen its effectiveness in combating this disease, a team led by Adrian Roitberg of the University of Florida has used approximately 500,000 hours on the Cray XT supercomputer known as Kraken at the National Institute for Computational Sciences, which is funded by the National Science Foundation, housed at Oak Ridge National Laboratory, and managed by the University of Tennessee. Roitberg and his postdoctoral associate, Gustavo Seabra, have begun to shed light on the workings of this marvelous molecule with AMBER, a well-known molecular dynamics code popular in the modeling of biological systems and one that works well with parallel architectures such as Kraken’s.
The simulation data are still being examined, but they will likely pave the way for more accurate future simulations, providing valuable information for the development of more effective treatments for diabetes, said Roitberg. Because exenatide is a mobile molecule, meaning that it changes form, or conformations, and jumps from place to place rather quickly, it is difficult to investigate. “It is too mobile to capture experimentally,” said Roitberg.
X-ray capture is insufficient because of the small peptide length, and nuclear magnetic resonance, or NMR, isn’t fast enough to capture the peptide’s quickly shifting shapes. Essentially, said Roitberg, capturing exenatide with NMR is like using a camera with a slow shutter speed to capture a speeding race car—the result is a blur.
Hence the need for sophisticated computer simulations that, in this case, are well-suited to Kraken’s scalability. Besides that, Kraken’s sheer size enables the team to simulate what the system is doing at an atomic level, acquiring knowledge beyond, but complementary to, the realm of experiment. The Cray XT’s computing muscle also provides the team with rapid throughput, or fast answers. “This is all that matters at the end of the day,” said Roitberg, adding that a typical day of simulations on Kraken may yield 60 to 70 nanoseconds of exenatide behavior, a far better result than many comparable computer systems.
The team ultimately hopes to make the active conformation, or the shape most effective in combating diabetes, as stable as possible, maximizing the molecule’s effectiveness as a type 2 diabetes drug.
The group previously predicted the structure of the last 20 amino acids in the exenatide chain, known as the tryptophan cage. One of the questions they ultimately hope to answer this time around is what does the tryptophan cage do when joined with the other 19 amino acids that make up exenatide. Thus far, said Roitberg, it seems that the end of the tryptophan cage stays the same, with the other 19 amino acids ultimately stabilizing it. However, most of the time exenatide takes different conformations. Understanding the behavior and structure of this peptide should allow researchers to better manipulate its conformations and up the ante in the fight against type 2 diabetes.
These simulations are just the beginning, however. “We’re certainly going to run more,” said Roitberg, adding that “you never know when you’re done . . . we’ll mutate something, rerun it, see where we’re at.” Given the nature of these simulations and the complicated systems involved, Roitberg’s team was forced to make numerous approximations that will be refined in future simulations, providing researchers with a better picture of exenatide’s structure and behavior.
Besides his research into exenatide, Roitberg has begun to ponder a possibly revolutionary idea in molecular dynamics: Is it possible to fold a peptide using quantum mechanics? Because molecular dynamics depends on various forces pushing and pulling within a system, it just might be. Roitberg admits, however, that it might not, but knowing the answer is vital. This research will require a huge number of processors and the simulations will be extremely complex. Ultimately, systems such as Kraken will provide the solution.
As high-performance computing enters the petascale era, simulations such as these will begin to shed light on previously dark areas of systems of all sizes, many of which don’t lend themselves well to experiment. Exenatide behavior is just one example—a very important one.