Issue Date: April 2, 2007
Science By And for The People
THE PUBLIC SPLASH surrounding SETI@home's search for extraterrestrial life nearly a decade ago signaled the promise of using home computers to tackle gargantuan computing jobs. Now, it's as easy as a button click on your PlayStation 3 (PS3).
Announced just last month, Stanford University chemistry professor Vijay S. Pande's popular volunteer protein-folding dynamics project Folding@home has been integrated into the software of Sony's ultrahip video game console. With the system's lightning-fast cell processor, Pande's group hopes to tap into even greater computing power of the people. Already, 14,000 PS3 users have signed up for Folding@home.
The Folding@home liaison with PS3 is just one sign that the notion of harnessing home computer systems for science is flourishing. Around the world, hundreds of thousands of PCs are crunching calculations of protein structures, near-Earth asteroid orbits, and climate change, and the results are being aired in publications like Science and Nature. Though useful only for certain types of problems, volunteer computing has emerged as a way not only to handle huge volumes of data but also to connect a wary public with real science.
The concept of "distributed computing" has been obvious to many scientists since the dawn of computer networking. Most of the time, computers sit idly on desks, their endlessly cycling processors waiting to be used. A huge computing job parceled in manageable tasks could involve not only computers within a company but also the anonymous computers sitting on desktops in dorm rooms, garages, and home offices connected via the Internet.
In 1999, the University of California, Berkeley, Space Sciences Laboratory (SSL) launched SETI@home, a program in which home computers were used to scan radio signals from outer space for signs that they could have been generated by intelligent life. The project, one of the first volunteer-computing projects, was wildly successful: More than 5 million people have participated in SETI@home.
SSL computer scientist David Anderson, who manages SETI@home, further evolved the software infrastructure to make it generally usable for any research group that wanted to set up a volunteer-computing project. Thus was born the Berkeley Open Infrastructure for Network Computing, or BOINC.
Since then, BOINC has become the most popular software platform for volunteer computing, hosting dozens of projects, from Spinhenge@home, which calculates spin dynamics of magnetic molecules, to Einstein@home, which searches for exotic objects called gravitational pulsars. Want to help track the orbits of near-Earth asteroids? Just download Orbit@home, and while you're out getting coffee or skiing in Aspen, your computer will be hard at work.
A number of labs, like Pande's, have developed their own platforms. IBM has established the World Community Grid, which hosts several projects, including FightAIDS@home, which tests docking abilities of anti-HIV drug candidates, and Help Defeat Cancer, which scans tissue microarrays for signs of cancer, using both BOINC and a software package from United Devices called GridMP.
Though the volunteer-computing strategy has spread from hard-core math and physics to economics and games, projects in the chemical sciences, particularly involving biomolecules such as proteins, DNA, or RNA, attract users because of their importance in understanding diseases such as Alzheimer's or muscular dystrophy. A Folding@home user from the U.K. with the screen name Rich99million says the significance of the project's objectives helped convince him to switch from SETI@home. "The idea of potentially helping to cure diseases or just helping to develop a better understanding of what goes wrong, along with the more varied work, won me over," he says.
A U.S. user with the screen name gwildperson, who says she's a social worker, says both she and her boyfriend use Folding@home. "We've both had relatives with Alzheimer's disease, which makes the project personal," she says. "I'm particularly impressed with the scientific papers that they've published."
The types of problems that can be solved this way need to be split into small segments that can be handled without communication with other processors, such as scanning chunks of data for extraterrestrial signals as in SETI@home. Collections of very repetitive tasks work well, such as running a simulation over and over with different starting conditions, in the style of the wildly successful climateprediction.net from Oxford University.
QMC@home is a case in point. Quantum Monte Carlo algorithms are statistical methods relatively new to quantum chemistry for calculating the electronic structures and reactivity of molecules. They can be split into small, individual tasks. "QMC has the opportunity for nearly perfect parallelization," notes Stefan Grimme, chemistry professor at the University of Münster, in Germany.
When that group, including graduate student Martin Korth, began studying QMC a few years ago, the potential for using volunteer computing was not lost on them. "The idea of a SETI@home-like project went around as kind of a joke for some time," Korth says. But then he read about BOINC. Now, some 14,000 volunteers have run calculations of noncovalently bound systems and interaction energies of DNA base pairs. The group is now wrestling with unstopping a methodological bottleneck in QMC known as the fixed-node error. To help do that, volunteers are calculating the ground-state energy of molecules, mostly organics.
Though Pande's Folding@home tackles the molecular dynamics of protein folding, a number of other projects focus on protein structures themselves, using various computational strategies. More than 100,000 volunteered computers are predicting protein structures at Rosetta@home using Rosetta, a software package developed by David Baker, biochemistry professor at the University of Washington.
Similarly, Charles L. Brooks III, a molecular biology professor at Scripps Research Institute in La Jolla, Calif., runs Predictor@home, and Brooks' former postdoc Michela Taufer, now a computer science professor at the University of Texas, El Paso, has started Docking@Home in collaboration with Brooks and others to study protein-ligand interactions. Thomas Simonson, biochemistry professor at the Ecole Polytechnique in Palaiseau, France, started proteins@home. In Japan, Tadashi Ando at Tokyo University of Science runs TANPAKU, also for predicting protein structures. All run on BOINC.
Perhaps most compelling about volunteer computing is the social phenomenon it has engendered. Hundreds of thousands of users organize themselves into teams with names such as Dutch Power Cows and Fires of Heaven and compete for "credits," or hours spent contributing to a project. These online communities are an important and, some say, integral part of the project. Moderators host forums and message boards on which users post their experiences, air grievances, and ask questions.
According to the statistics compiled by software of the various projects, volunteers are typically male and likely computer buffs with a strong interest in science. They frequently run numerous computers in their homes for the sole purpose of volunteer computing.
"I've seen pictures of people standing next to 20 computers like they used to stand next to their hot rods," Pande says.
Users frequently develop a "brand loyalty" to a project, whether it be Rosetta@home or climateprediction.net, Anderson notes. As one user from the U.S. who has a computer science background sums it up, "All Rosetta, all the time." Others participate in multiple projects.
But volunteer computing is not going to replace lab science or supercomputers. "Computers really are better than people for a lot of things," says Andrew J. Westphal, an associate director of SSL. He also directs Stardust@home, a variant of volunteer computing that uses volunteers to search for traces of comet dust particles in images (C&EN, Jan. 22, page 42). "A lot of science is interpretation, and you need somebody with expertise to do it right."
Predictor@home's Brooks points out that maintaining a volunteer-computing project is a lot of work. "Volunteers have to be kept happy," he says. "The social aspect makes it difficult to turn off the project for a while. People get angry and send e-mails."
Yet volunteer computing could go a long way toward ameliorating a general public hostility toward science that's arisen because many people don't know what science is, Anderson believes. "We're trying to create an open door into the lab," he says. "People can come inside and see what science is all about: trial and error, hard work, and unraveling real-life mysteries."
- Chemical & Engineering News
- ISSN 0009-2347
- Copyright © American Chemical Society