Friday, August 6, 2010

Can You Fold It?

I've known about the Foldit protein folding game for a while now, but a new Nature paper, and a bunch of associated press releases, have recently brought it to light in a big way.

Briefly: proteins. Proteins are the workhorses of the body and every living cell in it. No matter the cell type, proteins allow cells to do what they do, whether it be breaking down food, sending nervous signals, or whatever. They come in thousands of varieties but are all made up of the same stuff: Amino acids. There are 20 different amino acids; they are small molecules composed of carbon, oxygen, nitrogen, sulfur, and hydrogen. To build a protein, amino acids are joined together into an unbranched chain. Each amino acid has a small group of atoms called a sidechain that sticks off of the mainchain, or backbone. But an unbranched, straight-line formation is not an ideal formation. Proteins fold up, keeping some amino acids towards the center and others to the outside, some amino acids close together and some far apart. Each kind of protein folds up, usually on their own, into a very specific shape, the same shape every time. This folding is very unique to the protein, allowing it to exist in its most stable state, a state and shape that also specifies the function of the protein.

Perhaps now you see the where the name Foldit comes from. Because there are so many different ways a protein could possibly fold, figuring out which structure is the best one is one of the hardest problems in biology. Foldit is a game that allows users to predict the structure of a protein using their puzzle-solving intuitions and competitive natures. The game algorithm creates some potential starting protein structures, and then gives users a set of controls (like "shake," "wiggle," and "rebuild") and lets them manipulate the protein's structure in three dimensions as the game gives them feedback on the energy of a configuration. The designers of Foldit are smart in that they used the same conventions as other computer games and borrowed from aspects from online gaming communities such as leaderboards, team and individual challenges, user forums, etc.

Taking a look at the users of the game, even those with no significant background in biology or biochemistry were better at folding proteins than the game's algorithm. In a series of ten challenges, they beat the algorithms on five and drew even on another three. The authors of the study looked at how human pattern recognition gave them an edge over the computer. One example in many is that humans were able to detect hydrophobic ('water-hating') amino acid sidechains and rearrange them so that they were tucked inside the protein structure rather than outside. The best players, though, tended to have different strengths. Some were good at big adjustments while others were better at fine-scale optimization. And that's where the team-play function really shined; different team members handled tasks they were most adept at. That isn't to say that humans did better than the computer in every way. The algorithm was shown to be superior in generating the original unbranching sequence, before folding.

In the past few years scientists have become very interested in harnessing the power of the masses, and the masses' interest in science. In addition to Foldit take a look at Galaxy Zoo.

In case the in-text links didn't work, here's the Foldit and Galaxy Zoo sites:
http://fold.it/portal/
http://www.galaxyzoo.org/

This is the new Nature paper:
Cooper, Seth, et al. (2010) Predicting protein structures with a multiplayer online game. Nature: 466, 756–760. (DOI: 10.1038/nature09304)

And some story links:
http://arstechnica.com/science/news/2010/08/gamers-beat-algorithms-for-finding-protein-structures
http://www.nytimes.com/2010/08/05/science/05protein.html
http://www.wired.com/wiredscience/2010/08/crowdsourced-protein-folding/

No comments:

Related Posts with Thumbnails