Kraken offers new research directions in the battle against infection
By Wes Wade
A lack of treatment, cure, or vaccine for both AIDS and its precursor, HIV, is largely the result of a limited understanding of how all the components of the HIV-1 infection pathway operate. Despite the nearly three decades of study since both afflictions entered the public consciousness in the early 1980s, the most successful treatments currently available consist of antiretroviral drug therapies which, at best, merely slow the infection or disease’s progression.
Figure: Key mutations in specific amino acids in the CA domain change the Gag structure in the immature virion, perhaps altering the HIV-1 maturation pathway. The domains with the highest degree of structure are shown with yellow dots. Photo courtesy of Voth group, University of Chicago.
Now, the combination of massively parallel supercomputers that greatly accelerate research and new multi-scale computational simulation approaches that span the time and length-scales from the atomic to near macroscopic scales might soon lend experimental science some new insight in preventing the early stages of HIV-1 infection, and terminating the virus from “hijacking” other healthy cells.
“To me, this is a fundamental challenge for biomolecular simulation,” said Gary Ayton, an associate research professor in the Voth group, a research team led by Gregory A. Voth, formerly of the University of Utah’s Chemistry Department, now of the University of Chicago. “To bridge and span all of these scales [of time and length] such that we can make a meaningful connection with real experiments.”
HIV-1 is a retrovirus. Unlike other viruses that replicate by converting DNA into RNA in a host cell, the HIV particle converts its RNA to DNA. Nine genes are needed to replicate infectious particles and these genes make proteins to help in this process. One of these proteins is called reverse transcriptase and it is exactly this enzyme that current antiretroviral drugs attempt to inhibit. Yet they have only been successful in slowing the replication cycle.
Vaccines which target a protein that attaches to infected cell walls represent another possibility, where the vaccine would then seek to train the body to only attack infected cells.
Voth’s team is trying a different approach. With funding from the National Institutes of Health, and in collaboration with the research groups led by Wes Sundquist and Chris Hill of the University of Utah’s Molecular Biology Department—who have been studying AIDS and HIV since the early 1990s—they have employed their multi-scale approach on the University of Tennessee and the National Institute for Computational Sciences’ Cray XT5 Kraken supercomputer to examine specific stages the virus undertakes to spread throughout the body in hope of ultimately disrupting the process.
Figure: All-atom simulation of six Gag polypeptide CA domains arranged in a hexagonal structure, important to the protein’s overall stability in the immature virion. Red and yellow dots indicate residues important for stabilizing the hexagonal arrangement. Photo courtesy of Voth group, University of Chicago.
A key challenge for the team was to model the Gag polypeptide, a multi-domain protein which plays an important role in the formation of the immature HIV-1 virion. This immature virion later becomes the mature virus particle that will infect additional healthy white blood cells. Using multi-scale simulation approaches, Voth’s team discovered that a few select amino acid residues appear to be crucial in the stability of the entire Gag polypeptide structure. Within the larger context of halting HIV-1 infection, if this structure collapses, the immature virion can never develop and the infection ceases to spread further as the body routinely flushes out older cells in favor of new ones.
Modeling a Virus
Modeling atomic-level behavior can be done using a popular form of computer simulation called Molecular Dynamics (MD). Used for decades, MD codes employ Newton’s Second Law of Motion to model the time evolution of complex biomolecular systems.
For the HIV-1 virion, its size and complexity required a simulation on the scale of 500 million atoms—roughly 3 orders of magnitude greater than any current MD code can handle. Additionally, the team needed to examine the virus at varying time and length scales to gather meaningful results. Fortunately Voth’s team had already developed a highly flexible MD code called TANTALUS. This code employs a method of MD called coarse grained MD, or CG-MD, which is much less computationally intense. With CG-MD, single CG “sites” model entire collections of atoms. The feature of TANTALUS that makes it attractive to the virion problem is its coupling with all-atom simulations, its ability to use both forms of MD, and then make meaningful connections between the two.
“Using the techniques we’ve developed, we can calculate a good approximation to the interaction between the coarse grain sites [the actual behavior of atoms within the virion],” Ayton explained. “You then do MD simulations of those particular regions and either validate or don’t validate that particular interaction. It’s a back and forth thing, and at the end what you hopefully get is predictive capability.”
And it is precisely this predictive capability that Voth’s team is aiming for.
“That’s another new challenge for the simulation approach,” Ayton said. “You can either use it to test a theory or to mirror an experiment. So we’re trying to look in both directions, using it to test experiments and inspire new ones.”
If Voth’s findings are proven accurate through additional research, their results could have a profound impact on the course of future biological experimentation.
This research was supported by the National Institutes of Health (P50-GM082545)