Foldit—an online video game that allows players to compete against one another to find the best protein structure—has been a successful experiment, according to a group from the University of Washington (UW).
Foldit players use tools to twist, pull, freeze, and otherwise manipulate the structural dimensions of protein models to create those with the highest possible degree of stability. Based on the gamers’ moves, the game calculates the potential energy of the structure. The lower the potential energy, the greater the stability and the more points are awarded to the player.
Writing in Nature, the game’s creators reported that the combination of computer software and human intuition has been synergistic. Since it was created in 2008 it has produced a wealth of pertinent data about protein structure beyond what computers alone could amass. This structural information will be available to scientists seeking to create advanced pharmaceuticals targeting protein-based diseases such as cancer, HIV, and Alzheimer’s disease.
“We discovered that people through game play are able to explore the space in a way very different from computers,” Zoran Popovic, associate professor of computer science and engineering at UW and principle investigator on the project, told BioTechniques. “For some particularly hard problems where computers fall flat, we have shown that people are capable of surpassing computers and producing the best-known protein predictions.”
The structure, or native conformation, of each protein determines its biology function. Predicting this native conformation is necessary before researchers can begin engineering proteins for specific activities. But the overwhelmingly high number of potential structural possibilities has confounded the scientific field for the past 40 years.
“We had a world-class domain expert in proteomics, David Baker, who was up for the challenge of trying something completely outside of the biochemistry research norm,” said Popovic. Baker is professor of biochemistry at the UW who had previously developed [email protected], a program that uses the processor power from dormant computers within their online network to predict protein folding using various algorithms.
Working with Popovic and David Salesin, Baker was able to take this project to the next level by substituting human decision-making for the random-based programming algorithms. The group repackaged deterministic algorithms as user-friendly tools and added new tools that enabled players to directly alter aspects of the protein.
To attract players, the researchers incorporated elements of competition, social networking, and collaboration. Foldit incorporates discussion forums, player groups, strategy sharing, solo players, and a multi-angled point system.
The more people working on these problems, the more powerful the game becomes as a tool for folding prediction, according to Popovic, and Foldit does its part to tap into gamer fever. It’s not just being played in graduate school laboratories or elite research institutes; Foldit has drawn a diverse following by enticing regular gamers—interested in games like World of Warcraft or God of War—to solve science problems. Even nonscientists are invited to play, so long as a few simple guidelines are followed:
1. Pack the protein: make it as small and compact as possible.
2. Hide the hydrophobics: surround the orange side chains with as many atoms as you can.
3. Clear the clashes: when you see a red spiky ball, move the side chains farther apart.
Since the Nature paper, a number of scientists with their own unsolved scientific frustrations have approached the Foldit group. Now, the group’s challenge is to select which scientific dilemma will benefit the most from such a human-computer approach, and then to design a game around that problem. Popovic’s team has begun designing games for drug and vaccine design, nanotechnology, and biofuels.
“In my view as a computer scientist, the big question is, "how can we solve very hard problems of high-importance that neither computer nor people alone can solve?” said Popovic.
The paper, "Predicting protein structures with a multiplayer online game," was published Aug. 5, 2010 in Nature.