Slashdot Mirror
Folding@Home 2.0 - An Online Protein Folding Game
a boy named woo writes "Tired of justifying your gaming addiction? Now you can really help accomplish something while you play... thanks to Howard Hughes Medical Institute researcher David Baker at the University of Washington." In collaboration with others, Baker has designed a game, called "Foldit," with a practical outcome: players manipulate on-screen images of protein chains and attempt to predict their folding patterns. From the article:
"'Our main goal was to make sure that anyone could do it, even if they didn't know what biochemistry or protein folding was,' says [co-creator Zoran] Popovic. At the moment, the game only uses proteins whose three-dimensional structures have been solved by researchers. But, says Popovic, 'soon we'll be introducing puzzles for which we don't know the solution.'"
All that's there is the windows executable and the mac executable.
DNA is the ultimate spaghetti code.
http://fold.it/
Technoli
Baker already heads up Rosetta@Home , a BOINC project that has your computer fold proteins in its spare time. He's appreciated for keeping his journal up-to-date and being responsive to participants; Folding@Home is somewhat less responsive (and doesn't provide the BOINC option).
If you look at known sequence patterns (motifs) then you can assume it will fold the same, try to fit it and compare the fold with other proteins with the same motifs whose structures have been solved experimentally.
Protein folding is a very interesting subject that has seen some big advances over the years but we're still at the very beginning.
http://www.mininova.org/tor/1392891
Part of the project actually is to determine what can make the game more fun. For me the funnest part is competing against others.
Some drink at the fountain of knowledge. Others just gargle.
I work in a protein engineering/structure lab that has strong connections to the Baker lab, both in people and in scientific collaboration. The biggest project to come out the Baker lab is a protein structural modeling, simulation and prediction suite known as ROSETTA. While I'll gloss over some of the nitty-gritty about the methodology, suffice it to say that ROSETTA, through a combination of knowledge based and physics based modeling, has knocked the pants off of just about every other program out there used to simulate, design, and fold proteins. (Quantum-based physics models can be much better than ROSETTA, at the expense of a few extra superclusters and months of simulation time).
I no longer work as a ROSETTA developer or the "protein folding problem", but many of my lab mates do. They struggle with ROSETTA sometimes, as it comes close to predicting the real structure of a protein, and then falls away and wanders into another structure far from reality. If only it could 'see' the best structure when it came close!
The problem can be analogized with surveying a landscape. Imagine every square feet of dirt you can see is one possible protein structure, and you want to find the lowest elevation square foot. For a human, the visual search process is fairly quick and rapid. You can see a few hills out in the distance, but a much lower valley on the other side, where the land is lowest. It takes only a few seconds. On the other hand,a computer with no prior knowledge of the landscape can take a very long time to find that global minimum. The computer essentially has to drop a ball on the landscape and watch where it rolls, then pick it up, put it somewhere else and watch again (Physics and computer modelers forgive me!). It may never pick the right starting point to get over that far away hill.
Perhaps the brain can be as good at finding great protein structures as we are at finding lowest elevation points. Perhaps intuition about how a protein 'should' look can get us places a computer program never can without a ton of time and power. That's what this game is all about. The baker lab has done a fantastic job of turning a very hard scientific problem into a competitive game that is simultaneously fun, provides possible scientific information, and represents something of a human experiment on how our brains work.
This could be the next leap forward if it turns out some people have an innate knack for folding. It should be interesting to watch.
-Ryan
AUWYHSTOT (Acronyms are Useless When You Have to Spell Them Out Too)
This game is an offshoot of the rosetta@home effort to model protein folding. Folding@home is a separate effort.
Yeah, I've been playing it.. it's pretty awesome. I have no idea what I'm doing but my score goes up.
As such, I'm pretty sure genetic algorithms could give similar performance.
How we know is more important than what we know.
Sheer genius. I don't normally get into puzzle games but this one had me playing for 45 min straight before forcing myself to stop. It's well designed and fun to play.
With tetris it was time wasted down the tubes. At least with this you are doing something useful (and it might save somebody's life).
I can just see this thing going to cell phones, PDA's, etc.
This is academia, not industry. We are publicly-funded. Most researchers do not make any money off their discoveries and just get paid from grants. So the poor Indonesian kid probably cannot expect to get rich quick, but at least can be accorded his due share of fame and the benefits arising from that.
http://uwnews.org/article.asp?articleID=41558
Reason the account manager functionality wasn't baked into the server-side part of the platform is because David A. wanted to stress the BOINC stack as a technology anybody could use for a distributed computing platform. Basically anybody can throw a project together to solve whatever problem they want to solve, without having to clear it with David A. or myself.
Whoever runs an account manager has to make a judgment call about which projects they are going to expose, and by virtue of exposing it, they are stating that they trust the project. So if you sign up with that account manager you are trusting the person who runs the account manager to expose projects that are not going to bork your system.
It is a variation of the PGP signing model with the various levels of trust based on who you trust.
----- Rom
BOINC Development Team