|
|
发表于 2010-7-30 21:28:02
|
显示全部楼层
占座.
原文:
IBM, Volunteers Help Locate Anti-Cancer Drugs
June 17, 2010
By Andy Patrizio
Thanks to 1.5 million volunteers who left their computers running when not in use and a little help from Big Blue, cancer researchers have announced significant progress in looking for new potential drugs in cancer treatment.
The Help Conquer Cancer Project worked with the IBM-supported World Community Grid to send out protein samples for simulation testing on all of the computers. The program, running in the background of the volunteers' computers and harnessing unused compute cycles, simulated a process called crystallization, where proteins crystallize into a solid form.
In this form, the proteins can be further examined by special X-ray to see how they interact with cancer, and whether or not those proteins may cause the disease.
Using the World Community Grid to send out sample after sample to the volunteers, the Help Conquer Cancer Project believes it was able to determine six times as many images per protein for further testing in significantly less time than would be possible under manual human review.
By way of example, if a person looked at one image per second without rest -- which is not humanly possible -- it would take 1,333 days to examine all 12,500 proteins in the study. The World Community Grid did that in a fraction of the time, said Dr. Joseph Jasinski, an IBM distinguished engineer and program director of IBM's Health Care and Life Sciences Institute.
"The World Community Grid is really good at running embarrassingly parallel computation, where you do the same task over and over again. So it's set up for doing many possibilities to try where you want to throw away the ones that are no good quickly," Jasinski told InternetNews.com.
Something like testing for cancer drugs works with a setup like the World Community Grid. The 1.5 million machines work independently and don't communicate with each, hence Jasinski's description of them as "embarrassingly parallel." They perform the same repetitive task over and over -- in this case, testing a protein to see its potential in cancer treatment.
An increasing number of firms are turning to this solution for their own large-scale computing tasks, but they are keeping the process inside the firewall. Distributed computing lends itself well to a "loosely coupled" task like searching through a vast amount of data for a match, Jasinski said.
In a "tightly coupled" scenario, where the program might need the results of one step in order to continue, or processors need to communicate, a company would be better off using IBM's BlueGene servers, where high-speed interconnects enable interprocessor communication, Jasinski said.
IBM consulting helps firms determine whether their computation needs are loosely coupled or tightly coupled, and offers the appropriate solution. Some companies are using loosely coupled computing internally, Jasinski said.
IBM has its own technology and services through its Smarter Planet initiative to help these firms build internal distributed computing systems. Companies like financial, life science and drug research firms put an agent on employee computers and request they leave their computer running at the end of the day, he said.
"We have helped companies and institutions set these things up. It's part of a growing trend around distributed computing, a sort of precursor to cloud computing in a sense, so I think that general trend of trying to harness the horsepower you have and get as much productivity from the infrastructure you have is going to continue," Jasinski said.
"We have a good and growing list of problems people are applying this technology toward, typically in energy, the environment, health care and life sciences. We've also tried to get some stuff going in computational aspects of humanities research," he added.
The World Community Grid launched in 2004 and is the world's largest public scientific research computer network, with 514,000 members offering 1.5 million devices, meaning many people are running more than just their own personal computer.
It runs other tests similar to the Help Conquer Cancer Project, like FightAIDS@home, which looks for a cure for HIV, plus programs to fight influenza, muscular dystrophy, human protein folding and research efforts in the field of clean energy.
In April, the FightAIDS@Home group announced it had found two new potential proteins that could be used in developing protease inhibitors, an effective method of treatment for HIV.
Andy Patrizio is a senior editor at InternetNews.com, the news service of Internet.com, the network for technology professionals. |
|
发表于 2010-7-27 12:05:39