This post is by Deena Gould, graduate student fellow with the Center for Games & Impact.

Foldit is the classic discovery game. An intriguing scientific problem is turned into a playful challenge and suddenly thousands of ordinary people begin contributing to the solution.   Even before Foldit was released in 2008, I was intrigued by the puzzle of protein folding. Since proteins are involved in almost all cellular processes, it seemed that their importance and prevalence would have led us to know more about how they’re formed? Why is it so hard to figure out the shape of the protein if you know the sequence of its amino acids? Would the Foldit game help me understand more about protein folding? Could I contribute to solving puzzles that had eluded scientists?

Foldit player screen with protein puzzle

Game Play

In Foldit, the challenge of figuring out the shape of the protein given the constituent components was turned into a 3D digital jigsaw puzzle. The player turns and twists a protein model to simulate and find the most favorable interactions among the chemical groups. Overall, Foldit reminds me of the Rubric’s cube puzzles that my mother used to love. In Foldit, you quickly learn to make the protein as compact as possible, avoid empty space, and avoid having components clash. Tools with funky names like “shake” and “wiggle” show you how to fix the clashes, improve the backbone, and reassemble the side chains. Juicy feedback in the beginning stages keeps you believing that after just a few more puzzles, and with a little luck, you might be able to unlock some secrets and save mankind from horrible diseases.

Feedback in Foldit

The immersive problem-solving environment enables you to continue improving your skills and coaxes you into trying more challenging puzzles. When you learn that the models you manipulate represent actual proteins that scientists have posted for the community to solve, the game feels real. There are competitions and there are collaborations. I used the in-game chat feature and found that even on a Saturday night fellow players were happy to help me learn to use the “rubber band” feature to change the strength of atomic repulsion. There is a chat tool that allows players to upload a screen-shot useful to discussions between mentor and mentee or peer collaborations.

Game Design

How does someone design a puzzle game if they don’t know the solution to the puzzle? Foldit uses a molecular modeling program based on current knowledge about biochemistry to provide feedback to the players and score the puzzle solutions. Feedback is based on measurements of how chemically stable the folded structure would be based on having the lowest free energy or most favorable set of chemical interactions. Higher scores are awarded for keeping repulsive forces apart, compacting the molecule, burying the hydrophobic chains, and creating or maintaining hydrogen bonds. This seemed like an interesting way to design a game with “real challenges”.

Game Impact and Research

The opportunity to be a “research collaborator” is a powerful motivator. In Foldit, players can actually help create new knowledge that scientists use to build drugs that benefit humanity. When scientists know how a protein folds, they know its structure and can begin to understand its chemical processes that cause or prevent cell malfunctions. So when Foldit players contribute solutions to protein puzzles, they may be generating new knowledge useful to the creation of drugs that interact with the protein and alter the chemical processes to prevent or reverse diseases.

I had heard about the 57,000 Foldit players credited with scientific authorship in the publication Nature, so I decided to look it up. The Nature publication reports that the collective power of many players’ protein puzzle-solving provided useful results that were equal to, or better than, the computer generated simulations scientists had been relying on previously. Since there is a huge landscape to search for optimal protein folding, using the crowd-sourcing contributions of many people has been genuinely beneficial.

Prior to reading the Nature publication, I was a bit skeptical about the authenticity of the impact of Foldit’s crowdsourcing. How were the players able to do something that the computer algorithms couldn’t do, or do it better than the algorithms, if the game rules that gave the players feedback were based on these computer algorithms? The explanation is that human players were much better than computers at generating solutions that require divergent thinking about how to work through molecular instability in unique and creative ways. The publication also reports that good Foldit players have a greater intuitive sense for 3 dimensions and spatial reasoning than the computer. This information about human learning, human cognition, and artificial intelligence is as interesting and useful for advances in society as the players’ scientific contributions to protein folding.

Systems Thinking

The idea of using a lot of people to solve protein-folding problems can be just as intriguing as actually solving the biochemical riddles. Foldit has demonstrated that a multiplayer video game can be useful for scientific problem solving.   How far could the intelligence of collective human game playing take us as the corpus builds over time? Will new systems-level properties emerge? How will new technologies change the way we think about the creation of knowledge?

Cooper, S., Khatib, F., Treuille, A., Barbero, J., Lee, J., Beenen, M., Leaver-Fay, A., Baker, D., Popovic, Z., & Foldit players. (2010). Predicting protein structures with a multiplayer online game. Nature, 466: 756 – 760.