GPUs Model Maturation of HIV
By Allison Proffitt
December 4, 2012 | Thanks to a distributed computing platform of NVIDIA GPUs, bioinformaticians at IMIM (Hospital del Mar Medical Research Institute) and UPF (Pompeu Fabra University) have used molecular simulation to explain a specific step in the maturation of HIV. The results were published in the most recent issue of PNAS.
When the HIV virus enters a cell, the viral RNA is released, the corresponding DNA is synthesized, and polyprotein chains are produced. But for that cell to infect other cells, the polyprotein chain must be divided into many individual proteins capable of reconstructing the virus. HIV protease does the work of snipping the long chain into distinct proteins, but where does HIV protease come from?
“One of the most intriguing aspects of the whole HIV maturation process is how free HIV protease, i.e. the ‘scissors protein,’ appears for the first time, since it is also initially part of the long poly-protein chains that make up new HIV virions," said the study authors.
It turns out, HIV protease cuts itself out of the center of the long protein chain by binding one of its connected ends (the N-terminus) to its own active site and then cutting the chemical bond that connects it to the rest of the chain. Once free, HIV protease makes quick work of the rest of the chain and many infectious virus particles are born.
But figuring this mechanism out wasn’t easy. “When you go to the structural/atomistic level it is difficult to capture transient dynamics with any experimental technique,” study author Gianni De Fabritiis told Bio-IT World.
ACEMD, a molecular dynamic software, “computes the atoms dynamics for protein, water and ions as in a real experiment but in a computer,” De Fabritiis explained. Using the software, “we could visualize the moment in which the protease bind itself for cutting. There were NMR experiments setting constraints on how this process should occur, but this is the first time that it is seen at the atomic level.”
De Fabritiis’ group ran ACEMD on GPUGRID, a network that harnesses the processing power of thousands of NVIDIA GPU—graphics processing unit—accelerators from household computers made available by the public for research purposes. Individuals download a software client to their home computers that takes advantage of computer downtime to work on projects registered at GPUGRID.
“The advantages for the scientist are enormous,” De Fabritiis said. “It's enough to have a few thousand Euro server to handle a small project. Graphics cards, space and electricity are handled by users and continuously updated. It is great.
“The limitation is that it can practically only run in a high-throughput mode—many runs not faster runs,” he continued. “However, in our opinion with GPUs this is not even a real limitation as the runs are long enough and it is not much faster on a supercomputer. It also requires time and dedication. People want the server to be always up. I don't remember last time than GPUGRID was down.”
“GPUs accelerate these computer simulations by 5 to 10 times,” said Sumit Gupta, general manager of the Tesla Accelerated Computing Division at NVIDIA. “This means that on an average PC, you can get 5 to 10 times more work done (or get the same work done in 1/10th the time).”
For DeFabritiis’ team, the research lasted about two years, though he guesses that the actual compute time would be 3 to 6 months.
Having a structural picture of this step in HIV maturation is important because it facilitates—at least in principle—HIV drug design, said De Fabritiis. Though developing a treatment would require a good deal more research and development, arresting HIV protease could stop viral replication.
