01/22/2010: Accelerating the Pace of Drug Discovery using GPUs

張貼者:2010年1月25日 下午9:14Admin strator
By Sumit Gupta, posted Jan 22 2010 at 02:37:39 PM

This post is an entry in The World Isn’t Flat, It’s Parallel series running on nTersect, focused on the GPU’s importance and the future of parallel processing. Today, GPUs can operate faster and more cost-efficiently than CPUs in a range of increasingly important sectors, such as medicine, national security, natural resources and emergency services. For more information on GPUs and their applications, keep your eyes on The World Isn’t Flat, It’s Parallel.

Every time I hear about some little boy or girl who has a life-threatening disease, I get reenergized to promote the opportunities that computers offer in helping reduce disease and finding cures for human ailments. The exciting new developments on this frontier have to do with using parallel computing to speed up research and create the breakthrough discoveries that will save lives.

Finding new drugs is a complex and laborious task. Biochemists have to try millions of compounds before they can figure out which ones are effective against a particular virus or bacteria or which cause a desired reaction in the human body. To narrow the field, scientists use automated tools for high-throughput screening. But at some point, they have to test the remaining biochemical compounds in time-consuming manual experiments in a “wet” laboratory. (See our recent piece on the Tesla BioWorkbench for more information on how we’re helping researchers in the life sciences use GPU computing in their work.)


Testing compounds is an iterative process, and it can take up to more than five years before a new drug is finally discovered. Then there are five or more years of clinical trials and so on as part of the FDA approval process. Obviously, anything that can accelerate the testing process makes a significant difference in the speed with which a potentially lifesaving new drug gets on the market.


Computer simulations of biochemical reactions help point scientists in the right direction and improve their productivity. The problem is, the computer simulations necessary for this type of research are so compute-intensive they’re typically done on supercomputers. It can take weeks or months on a supercomputer to simulate just one biochemical reaction. For example, in order to simulate how tiny cellular mechanisms called ribosomes work to build proteins out of amino acids, it took scientists at Los Alamos National Laboratory more than eight months of supercomputer time to recreate a process that in reality takes about 2 nanoseconds.

With about 15 million to 20 million researchers who could benefit from supercomputer access and roughly a couple thousand supercomputers, there’s just not enough computing resources to go around. Right now biochemists have to request time on these supercomputers a year in advance, and not all researchers even get access.

But if the right tools are available, scientists don’t have to wait to test their new ideas. This is where parallel computing can accelerate innovation, by removing the compute bottleneck represented by traditional supercomputers.

The tremendous parallel processing capability of GPUs accelerates life-sciences applications by an order of magnitude. For example, a popular molecular simulation applications called NAMD (nanoscale molecular dynamics) gets the same performance out of four GPUs as it does out of 16 CPUs. The speed-up is a result of the parallel architecture of GPUs, which enables NAMD developers to port compute-intensive portions of the application using the CUDA C toolkit.

The shift to parallel computing in the life sciences means that biochemists can use a desktop Tesla Personal Supercomputer with four GPUs and outperform a large cluster of CPU servers in a data center. If they need more compute for their simulations, they can easily scale by using GPU-based clusters.

Scientists can thus augment their PCs and departmental servers with GPUs and start doing large biomolecular simulations without requiring a supercomputer. (Of course, GPUs are also being added to the supercomputers soon.)

By giving scientists greater access to computing resources, GPU-based computer simulations are helping to fulfill the promise of using computers to accelerate the pace of drug discovery.