Gaming for good: human thought beats computer algorithms at solving protein structures

Considering my fascination of late with unusual author lists in science papers, you can guess how excited I was to see an article in Nature that credited online gamers. I was especially amused to see that citation services like PubMed abbreviate “Foldit players” as “Players F.”

A screenshot from my own training as a protein folder

Now on to the actual story. We all know that playing video games can require serious problem-solving skills. Gamers sometimes spend hours each day solving puzzles and honing their spatial reasoning abilities. Did you ever wonder if those efforts could be applied directly to real-world problems? A research team based at the University of Washington developed a computer game called Foldit, which engages people’s ingenuity (and free time) to predict protein structures.

The idea for Foldit came out of an existing project called Rosetta@home, which uses the spare processing power on people’s computers to solve protein structure by minimizing the energy of the amino acid chain. This type of project is called distributed computing, because it uses many computers in parallel to execute a task which would take much longer on a single processor. So whenever a Rosetta@home user steps away from the computer, a screensaver turns on and shows a protein as it folds and searches for the best structure.

One of the most popular distributed computing projects is SETI@home, which analyzes signals from a radio telescope for any pattern which might come from alien civilizations (SETI stands for Search for Extra-Terrestrial Intelligence).  UC Berkeley’s own Space Sciences Laboratory was the force originally behind SETI@home; their software platform is used today by many other projects, solving a wide range of problems (scientific and otherwise).

Rosetta@home is great at optimizing protein structure, as long as that structure is somewhat close to the real one. The problem comes when the protein is stuck in a non-ideal structure (a local energy minimum) and it has to go through a much higher energy transition state to find the global minimum. This is where human reasoning steps in. As the authors put it, people excel at solving those structures that require “substantial backbone rearrangements.” Foldit players actually outperformed the Rosetta algorithms at most of the challenges posed. They call this approach “crowd-sourcing,” i.e. recruitment of the public to complete large-scale tasks (see our beloved Wikipedia as a prime example).

I love this idea. The authors realized that distributed computing was taking advantage of the idle processor power in our home computers, but what about the idle processor power in our brains? People spend a lot of time solving puzzles for fun, and there are many puzzles that need to be solved to advance scientific research. The best thing about it is that you don’t need to have any knowledge of biochemistry whatsoever. So give it a try! And when someone asks you why you’re playing computer games at work, politely inform them that you are working toward your next publication (as Players F.).

Further reading:
Excellent coverage in ars technica

ResearchBlogging.orgCooper S, Khatib F, Treuille A, Barbero J, Lee J, Beenen M, Leaver-Fay A, Baker D, Popović Z, & Players F (2010). Predicting protein structures with a multiplayer online game. Nature, 466 (7307), 756-60 PMID: 20686574

Leave a Reply