Researchers at the University of Illinois Urbana-Champaign have successfully modeled the interior of the HIV-1 virus using the new Blue Waters supercomputer at that university. The work has the potential to open the door to new antiretroviral drugs that target HIV-1, the virus that causes AIDS.
When the researchers at university’s Beckman Institute for Advanced Science and Technology began the project, they didn’t have the capability to model the entirety of the capsid, the protein-based, polymorphic structure inside each HIV virus that contains the RNA that is responsible for disrupting healthy human cells. The average capsid structure contains about 64 million atoms, including about 1,300 different proteins, making it a difficult computational problem.
“Information was known about isolated parts” of the capsid, says Juan Perilla, a postdoctoral research associate at the Beckman Institute, in a Beckman video about the project. “We didn’t have the whole structure together.”
Their luck changed when the National Center for Supercomputing Applications (NCSA) took deliver of Blue Waters, the new petascale system installed on the Urbana-Champaign campus. Blue Waters was originally supposed to run on IBM Power7 chips. When IBM cancelled the contract, Cray stepped in, and delivered a system composed of 49,000 NVIDIA GPUs, 500 PB of storage, and a 300Gbps wide area connection with a peak capacity of 11.5 petaflops
The capsid protein modeling project was one of the first to run on Blue Waters, which didn’t officially go online until March 2013. According to Klaus Schulten, head of the Beckman Institute’s Theoretical and Computational Biophysics research group, Blue Waters has the capability to model a structure with 100 million atoms, giving researchers a little bit of headroom above the 64 million atoms in the capsid.
Just the same, it took Blue Waters a month to complete the modeling and create the world’s first detailed rendering of the HIV-1 capsid, a major scientific achievement. “We could now look, protein by protein, how these over 1,000 proteins are capable of realizing a rather wide distribution of local shapes along the surface of the HIV-1 capsid,” Schulten said.
“It also gave us the first insights into how one might approach this capsid through pharmacological warfare, by developing a drug that might fit into some of the nooks and crannies, and thereby shutting the capsid so that it couldn’t release its genetic content,” he continues in the video.
Blue Waters was “essential to the success of this project because it’s the only machine where we could have run these simulations,” Perilla says. The researchers published their findings in a May 30 article in the journal Nature.