Stanford gives distributed computing an A

A university team shows that thousands of low-end PCs wrangled together into a loose network can be used effectively on problems scientists encounter in the lab.

Evan Hansen Staff Writer, CNET News.com
Department Editor Evan Hansen runs the Media section at CNET News.com. Before joining CNET he reported on business, technology and the law at American Lawyer Media.
Evan Hansen
2 min read
Scientists at Stanford University have demonstrated tangible proof that scientific experiments can be conducted using thousands of low-end PCs wrangled together into loosely linked networks. A group of chemists, including Stanford assistant professor Vijay Pande, said they successfully predicted the folding rate of a protein using calculations worked out on a so-called distributed computing network. Their research, conducted last year, was published this week in the science journal Nature.

In an interview, Pande said the demonstration was an important proof of concept for the use of distributed computing in the lab. Distributed computing involves spreading computing tasks across hundreds or thousands of computers on the Internet or private networks that would otherwise be sitting idle.

"For most scientific problems, it's not obvious how to throw more computers at a problem in a way that is helpful," Pande said. The protein experiment, however, provided a solid example.

The idea of tapping spare processing power from ordinary PCs has seized the imagination of the computing industry, luring thousands of volunteers into donating their spare CPU cycles to the search for extraterrestrial life and to the battle against diseases such as cancer. But none of those efforts has contributed to publishable scientific results, Pande said.

And though distributed computing projects have had some success when it comes to mathematical challenges--one group found the largest prime numbers, while another group deciphered a message encoded with RSA Security's RC5-64 encryption algorithm--the Stanford experiment illustrates that distributed computing can be applied to problems scientists encounter in the laboratory.

The Stanford chemists have been working to create mathematical models to predict the behavior of complex biological molecules. The group hopes to learn more about how certain proteins become folded, a key problem in understanding illnesses such as Alzheimer's disease, Pande said.

In order to test the accuracy of their models, the Stanford chemists devised an experiment aimed at producing a measurable, quantitative result. Pande said they chose a well-understood protein known as BBA-5 and turned over the job of calculating the rate of folding in the molecule to some 200,000 PCs connected through a distributed computing project known as Folding@home, formed in 2000.

The predictions were then compared with physical tests conducted on the same protein using a laser.

The experiment did not produce new information about the protein, but it did provide some corroboration for the group's algorithms.

Although Pande called the experiment a success, he cautioned that it also underscored the difficulty of crafting useful work for distributed computer networks.

"It's one thing to tell people they're fighting cancer, and another to say what's being done that will lead to publishable research," Pande said. "I would like to think that we will one day have something that is not insignificant."

News.com's Stephen Shankland contributed to this report.