Science is full of hard problems. One hard problem is protein folding. Indeed vast amounts of computer power have been thrown at this problem. So one wouldn’t think that the computer we’ve got sitting on top of our body would be much use for this problem. But is this true? Can humans fold proteins better than computers? Enter onto the scene foldit developed by a group of researchers here at the University of Washington.
An Economist article explains:
The existing program uses trial and error, and pre-programmed mathematical rules that govern folding as understood today. But users of the screensaver told David Baker, a biochemist at the University of Washington and lead scientist on Rosetta@home, they could do better. So Dr Baker, Zoran Popovic, a computer scientist at the University of Washington, and graduate students Seth Cooper and Adrien Treuille set about creating a compelling computer game.
Players use their computers to fold proteins. The more chemically stable the folded protein becomes, the more points the players are awarded. In trials of the game hundreds of players were given 40 protein puzzles to solve (for the trials, the folding solutions were already known). Many of the best players were not scientists but were able to find the correct structure faster than computers.
The next big step will be to give players proteins for which the optimal folding is not known. They will then be doing cutting-edge research in protein-structure prediction. If all goes well, the game will move on to protein design this summer, by including options that allow players to modify sections of the protein. This will allow them to design a protein that blocks the action of a virus.
Dude, I totally want a game to beat D-waves future Kwantum Komputer!
When you block the action of the virus, can you hit X, B, B, X and use a kewl finishing move, like throwing the virus in a spiky crystalline pit?
That is awesome. I’ve wondered if you could do something like this ever since the day I found out Tetris was NP-complete…
Interesting do human realy solving protein folding much better than computer?
“Go: Go is perhaps the largest and most complex game that humans have tried to solve, with a 19×19 board that results in a whopping 10,170 possible positions (InWap). Whereas an average chess position allows for 15 to 25 moves, Go positions allow approximately 250 moves. While the strongest Go computer programs are competitive with champion Go players on modified nine-by-nine boards, the complexity of the regulation boards is such that the programs can be beaten easily by even moderately intelligent children (AI Horizons).”
http://www.gelfmagazine.com/gelflog/archives/what_games_can_humans_still_win.php
If Dwave will continue researche and will do universal quantum computer then for them will need error correctio, becouse how they would know that answer is good? Maybe acording some optimal checking… But what they can check if they don’t know do realy protein folding or somthing behave according to they simulation and not according to some over more chaotic schem?
The biggest problem I can see with this is that it’s going to provide ammunition to the Intelligent Design crowd. No, it doesn’t make sense, but it does make good headlines:
That’s the kind of thing we may see coming.
Computers don’t use first principles to design proteins or predict their folds. I mean maybe they use physical first principles, but there are other factors at play. Timing can be critical – some but not all proteins fold as they are being made, so the entire sequence and therefore possible lower energy conformations are not available. Some proteins require energy to adopt their most stable conformation, while others take advantage of an energy barrier to NOT adopt their lowest energy conformation. Solvation is important and until recently has not been possible to model with any degree of accuracy. And even if you have a structure, it is just a snapshot of what the protein is doing.
I’m not surprised that people can arrive at solutions quickly. Random searching, even with dead end elimination and other algorithmic approaches to reduce the size of search space, does not allow for intuition to be injected. More effective would probably be a computer providing a panel of possible conformations, then a mechanical turk type process where people could select and play with certain molecular configurations. And then to iterate this process, so that when a ‘player’ went to bed (or the bathroom or whatever) the computer would then take the player’s current structure and tweak it, providing another panel of options (derivatives from the parent structure that the player left).
Anyhow, when you give a biological system a chance to solve a problem in protein folding or molecular recognition, it will. Look at phage display and immunological affinity maturation processes. These technologies are methods that exploit biological diversity to arrive at one or many solutions to a problem. The rules that are used by those systems are much more closely aligned to biological reality than the models that we understand today and can program into a computer.