ENERGY | WIRELESS | NANOTECH | MEMS | OPTICS | QUANTUM | 3D | CHIPS | ALGORITHMS

Monday, February 20, 2012

#ALGORITHMS: "Video Processors Aim for Cancer Cure"

Researchers are repurposing the graphics processing units (GPUs) that power video gaming consoles for ultra-detailed simulations of how cancer cells reproduce, unlocking new protein folding techniques that could lead to a cure.



When Wake Forest University touted its latest medical research effort, it described the breakthrough as "the cure for cancer comes down to this: video games." That sensational assertion, when investigated, relates to professor Samuel Cho's repurposing, for medical simulations, the multi-core graphics processing units (GPUs) which were originally intended for video games.

Graphics processing units like Nividia's Tesla, with up to 512 on-chip high-speed cores, were intended to accelerate the processing of pixels for video games. With each core assigned to a sub-set of screen real estate, the GPU can divide-and-conquer detailed rendering problems in short order--offering thousand-fold speed-ups of video gaming tasks compared to traditional CPUs that must labor though updating each pixel on the screen in serial fashion.

What scientists like Cho discovered, in cooperation with colleagues at the University of Maryland and Zhejiang University in China, was that GPUs are also perfect for simultaneously emulating all the various processes that go on inside a living cell. A traditional CPU has to simulate these "goings on" in serial fashion, updating progress with each simulated process, storing its results in memory, then moving on to the next process. After all the processes are updated "time" is moved forward, and the single CPU then goes through the list of processes updating them all again.

A GPU, on the other hand, can emulate all processes in parallel--assigning each process to a separate core. The emulation can then proceed in real-time, with each process calculating its next state in lock step with all the other processes, thus more accurately imitating all the simultaneous processes that are going on inside a living cell. Without the parallel processing capabilities of GPUs, Cho estimates his detailed simulation would have taken 40 years to run.

As a result, Cho and colleagues claim to be able to "see exactly how the cells live, divide and die." And it’s the dying part that has raised hopes for a cure to cancer. In particular, when simulating ribonucleic acid (RNA), Cho noticed that hitherto hidden states in its folding and unfolding had previously masked the means by which the telomerase enzyme--found only in cancer cells--makes them immortal. All other cells live, die, and are replaced with fresh copies. Cancer cells, on the other hand, become immortal by adding telomeres to the ends of copies, causing the uncontrolled growth characteristic of cancer. Now cancer drugs can be sought that block the telomeres from being added, potentially curing cancer.