The screen savers of thousands of volunteer PCs enrolled in the Folding@Home project have not flickered in vain. The project has yielded new insights into how to attack the thorny biological problem of protein folding, researchers report online 20 October in Nature.
Proteins begin as a long chain of amino acid building blocks inside a cell, then they contort and twist into their characteristic shape. This folding can take anywhere from a few nanoseconds to tens of microseconds, depending on what steps the protein goes through on the way to its final shape. Even though the whole process is incredibly fast from a human perspective, a typical modern computer needs years to simulate even a microsecond of the action.
To get around this problem, a team of researchers tried a different approach. Rather than simulating wiggling proteins for several microseconds, chemical biologist Vijay Pande and colleagues at Stanford University aimed for much shorter glimpses--about 5 to 10 nanoseconds--of a 23-amino-acid chain called BBA5. The chances of a chain folding into its final configuration in this amount of time are slim, so hoping for it to happen is like playing the lottery, Pande says. But instead of buying tickets one at a time and simulating the folding on a single computer, the team farmed out tickets to more than 30,000 volunteer computers around the world, which ran the short simulations.
Back at Stanford, the team accumulated about 2000 years of volunteer computing time and predicted that BBA5 should fold in an average of about 10 microseconds. To test this prediction, a group led by Martin Gruebele of the University of Illinois, Urbana-Champaign, took BBA5 into the lab and used a laser to shock the proteins into unfolding. The researchers then measured the time they took to refold by tracking a fluorescing amino acid that glows when the protein is stretched out but is hidden once it folds. Their estimated folding time of 7.5 microseconds agrees well with the computer predictions.
This is the first time a protein folding simulation has predicted and matched with a lab result, says biophysicist Charles Brooks of the Scripps Research Institute in La Jolla, California. Plus, he says, "it's a neat way of getting the calculations outside the scientific community--the public at large is doing the science."