The problem of calculating the way proteins fold is one that requires immense computational power, but can be helped along by humanity's innate ability to solve 3D puzzles. A crowd-sourced game called Foldit enables human players to contribute to defeating some truly awful diseases by figuring out how the proteins fit together. In some cases they even beat the scientists themselves to figuring out the structure of viruses — and they've made computers much better at it in their wake.

Proteins are long strings of amino acids, and much like the earbuds in your pocket can potentially fold up in any of a myriad of ways — but unlike the cable of your headphones, a protein can only ever fold in a single configuration. Predicting how a protein is going to fold requires a huge amount of processing power because of the millions of possible ways it can fit together, but it turns out humans are much more adept at figuring this out than computers. Foldit provides players a protein which they try and tangle, and then scores them based on how well folded it is, information which the scientists then parse for usefulness in the real world. Earlier this year, players even managed to pick apart how an AIDs related Mason-Pfizer monkey virus (M-PMV) virus fit together, a puzzle which had baffled scientists for decades. Out of this tremendous efficiency, the question arose if the humans skills could somehow be ported into an algorithm that computers could use.

To do this, the researchers added a new set of tools to the game which allowed players to codify their strategies as macro "recipes" that they could share among other players, who could then modify them further. These recipes spread like wildfire, as soon there were 5,400 available, both originals and variants, but two stood above the others as the most efficient: Quake and Blue Fuse. This latter recipe eventually was crowned king, and the one that players used the most as an efficient automation tool to help them.

Yet at the same time and completely independently of the Foldit players, scientists were working on their own algorithm using a far more advanced set of tools than the ones provided to the game-players. It was dubbed Fast Relax, and was significantly more efficient than its predecessor from 2002, Slow Relax. Bizarrely and coincidentally, Fast Relax and Blue Fuse were extremely similar, even though the makers had no contact with one another. Produced under incredibly different circumstances and with wildly different tools, the two algorithms are an example of convergent evolution. While the scientist created Fast Relax was overall more efficient than Blue Fuse, that's only because Blue was optimized to run in Foldit rather than the real world.

The researchers are now working on expanding the Foldit tools to allow for more control over the process, and hopefully make bigger and better algorithms. This process shows that citizen science can make some remarkable strides for research, but that it needs more than just your spare CPU cycles, it needs your brain, too.