But Mom, it's for science!
May 10, 2008 5:15 PM Subscribe
foldit is a new computer game scientists have created that lets YOU help them make science!!
Similar in spirit to the SETI@home project, and expanding on the idea of the Rosetta@Home project (which is also like the SETI project), this one requires active participation to work. And according to its creators, you could win a Nobel Prize!! But you have to get the top score, so start clicking, Nerdlinger!
via digg
Similar in spirit to the SETI@home project, and expanding on the idea of the Rosetta@Home project (which is also like the SETI project), this one requires active participation to work. And according to its creators, you could win a Nobel Prize!! But you have to get the top score, so start clicking, Nerdlinger!
via digg
The site's been acting up a bit today, but keep trying, it'll work.
posted by Koko at 5:23 PM on May 10, 2008
posted by Koko at 5:23 PM on May 10, 2008
Site's folded. Needs more intelligent design.
posted by weapons-grade pandemonium at 5:53 PM on May 10, 2008 [2 favorites]
posted by weapons-grade pandemonium at 5:53 PM on May 10, 2008 [2 favorites]
I'm glad this is out ... I saw Dave Baker give a talk on it a few months back and it looked awesome. He had the whole room of scientists laughing. I love the idea that some kid somewhere might have a knack for folding proteins.
posted by pombe at 6:25 PM on May 10, 2008
posted by pombe at 6:25 PM on May 10, 2008
Thanks! I just visited the site; I'll be back. There was a message on the homepage that the site had gotten a little bit more popular than expected, and that was causing a variety of problems.
posted by Kronos_to_Earth at 6:49 PM on May 10, 2008
posted by Kronos_to_Earth at 6:49 PM on May 10, 2008
Is this different than "Folding@Home"? Since I can't reach the site, and since you haven't told us just what "science" they're creating with this, I can only guess. But "Folding@Home" (which is using distributed computing to calculate how proteins fold) has been running quite successfully for several years now.
posted by Class Goat at 8:00 PM on May 10, 2008
posted by Class Goat at 8:00 PM on May 10, 2008
It's different from F@H in that humans are doing the folding. The computer is still "scoring" the result in the normal way, but instead of doing simulated annealing or whatever to generate candidates, they're asking humans.
posted by hattifattener at 8:08 PM on May 10, 2008
posted by hattifattener at 8:08 PM on May 10, 2008
Oooo, the chemist in the bioassay group just got a new way to annoy his biologist coworkers!
posted by Kid Charlemagne at 10:06 PM on May 10, 2008
posted by Kid Charlemagne at 10:06 PM on May 10, 2008
I like it already from the gameplay movie where the player got "H-Bond Master!!!" and "Hydrophobic Hider!"
Now all they need to do is to integrate this into Puzzle Pirates.
posted by demiurge at 10:30 PM on May 10, 2008
Now all they need to do is to integrate this into Puzzle Pirates.
posted by demiurge at 10:30 PM on May 10, 2008
Can humans really help computers fold proteins?
We’re collecting data to find out if humans' pattern-recognition and puzzle-solving abilities make them more efficient than existing computer programs at pattern-folding tasks. If this turns out to be true, we can then teach human strategies to computers and fold proteins faster than ever!
Somehow I didn't anticipate that last step.
posted by parudox at 10:56 PM on May 10, 2008 [1 favorite]
We’re collecting data to find out if humans' pattern-recognition and puzzle-solving abilities make them more efficient than existing computer programs at pattern-folding tasks. If this turns out to be true, we can then teach human strategies to computers and fold proteins faster than ever!
Somehow I didn't anticipate that last step.
posted by parudox at 10:56 PM on May 10, 2008 [1 favorite]
I think this is brilliant. Humans are great at finding patterns and solving puzzles, and if they can teach computers how to do it like humans, while having humans doing it as well, that's a huge step forward. Great idea, truly, even if it doesn't work out in the end (guess we'll have to wait and see).
posted by gemmy at 11:30 PM on May 10, 2008
posted by gemmy at 11:30 PM on May 10, 2008
I'm thinking a Metafilter group would totally beat the Something Awful group at such a game.
posted by SassHat at 12:39 AM on May 11, 2008 [1 favorite]
posted by SassHat at 12:39 AM on May 11, 2008 [1 favorite]
You will all regret this when Skynet goes live.
posted by basicchannel at 12:47 AM on May 11, 2008 [2 favorites]
posted by basicchannel at 12:47 AM on May 11, 2008 [2 favorites]
so they have multiple players per game, in case someone's going around sabotaging everything? i'm kind of curious but don't want to have to "sign up" just to see what the hell it actually involves, guess i have to.
posted by Dillonlikescookies at 2:12 AM on May 11, 2008
posted by Dillonlikescookies at 2:12 AM on May 11, 2008
Needs Achievements to really addict the poor impulse control crowd.
posted by Talanvor at 10:50 AM on May 11, 2008
posted by Talanvor at 10:50 AM on May 11, 2008
I think it needs a crafting system -- if the proteins could learn enchanting or blacksmithing, you could make totally twink armor and kick ass in PvP!
posted by Koko at 12:12 PM on May 11, 2008 [1 favorite]
posted by Koko at 12:12 PM on May 11, 2008 [1 favorite]
I Am A Protein Biophysicist-
I think it's a wonderful idea, but am skeptical regarding whether you and I will do a better job by pointing and clicking than a computer will by using intelligent algorithms to find the energetic minimum. My guess as to what will come out of this - and I think the Baker folks are leaning the same way - is that we'll find a few ways that humans are able to classify good folds versus bad, and hopefully be able to code these into the next generation of folding programs. One problem with today's programs is that it's possible to find oneself in a local energetic minimum, and thus 'stuck' in the wrong fold, even considering somewhat randomized search methods such as Monte Carlo. It's probably true that your average person who has a very basic sense of what "works" in terms of protein folding will do a better job than a computer in finding their way out of these false minimums.
Even if we're not able to improve our folding algorithms by studying how humans approach the problem, foldit is still a great teaching tool. Ideas like protein folding aren't going to make much sense unless you've been in the field for awhile, or you've played a "game" like foldit. So either way, This Will End Well.
posted by StrangerInAStrainedLand at 1:07 PM on May 11, 2008 [1 favorite]
I think it's a wonderful idea, but am skeptical regarding whether you and I will do a better job by pointing and clicking than a computer will by using intelligent algorithms to find the energetic minimum. My guess as to what will come out of this - and I think the Baker folks are leaning the same way - is that we'll find a few ways that humans are able to classify good folds versus bad, and hopefully be able to code these into the next generation of folding programs. One problem with today's programs is that it's possible to find oneself in a local energetic minimum, and thus 'stuck' in the wrong fold, even considering somewhat randomized search methods such as Monte Carlo. It's probably true that your average person who has a very basic sense of what "works" in terms of protein folding will do a better job than a computer in finding their way out of these false minimums.
Even if we're not able to improve our folding algorithms by studying how humans approach the problem, foldit is still a great teaching tool. Ideas like protein folding aren't going to make much sense unless you've been in the field for awhile, or you've played a "game" like foldit. So either way, This Will End Well.
posted by StrangerInAStrainedLand at 1:07 PM on May 11, 2008 [1 favorite]
Oh, Stranger, I'd argue that protein folding is a computationally intractable problem. Getting stuck in a local minima when doing optimization problems is pretty much the canonically unsolvable problem, currently.
Humans are good at it for certain problems, to certain sizes of N, but thinking that some insightful human constructions will generalize to an arbitrary 'find the 1 element from a space of 2^n elements' is probably going to result in you being disappointed.
Also, Monte Carlo is, I believe, more than 'somewhat randomized'. It is completely random. You just keep sampling until the answer shows itself.
On a related note: Intractability
posted by enkiwa at 3:25 PM on May 11, 2008
Humans are good at it for certain problems, to certain sizes of N, but thinking that some insightful human constructions will generalize to an arbitrary 'find the 1 element from a space of 2^n elements' is probably going to result in you being disappointed.
Also, Monte Carlo is, I believe, more than 'somewhat randomized'. It is completely random. You just keep sampling until the answer shows itself.
On a related note: Intractability
posted by enkiwa at 3:25 PM on May 11, 2008
Go straight to the Youtubes. Must have.
posted by jayCampbell at 6:02 PM on May 11, 2008
posted by jayCampbell at 6:02 PM on May 11, 2008
Let's fold some proteins, Mefites. Later though, when their server is less slammed. Penny Arcade unleashed their linking power upon it today as well, and creating the group took awhile.
posted by Tehanu at 11:29 AM on May 12, 2008
posted by Tehanu at 11:29 AM on May 12, 2008
Oh, Stranger, I'd argue that protein folding is a computationally intractable problem.
I'd tend to agree, if you mean finding the exactly correct native conformation a priori and then trusting your solution as the gospel truth. Evidence however, even if it is recent work out of Baker's own lab, in which they've used a computation solution for a ~100 residue protein as a molecular replacement solution for x-ray diffraction data (for the non-biophysicists among us, this means the predicted structure was good, very very good in fact.), points towards a future in which yes, in at least some if not most cases we can use computers to calculate a *likely* protein fold. Even if, as you correctly state, the combinatorial complexity of the problem makes it practically insoluble in an absolute mathematical sense.
Also, Monte Carlo is, I believe, more than 'somewhat randomized'. It is completely random. You just keep sampling until the answer shows itself.
I'll give you that, but let's be realistic here; anyone using a Monte Carlo method is probably doing so in conjunction with some kind of rejection rules which limit your moves in some way, such as the Metropolis Criterion for starters. Thus, your Monte Carlo search is no longer so random. You're most likely well aware of this, but I've got to defend myself when pickers of nit take up arms.
posted by StrangerInAStrainedLand at 11:41 PM on May 12, 2008
I'd tend to agree, if you mean finding the exactly correct native conformation a priori and then trusting your solution as the gospel truth. Evidence however, even if it is recent work out of Baker's own lab, in which they've used a computation solution for a ~100 residue protein as a molecular replacement solution for x-ray diffraction data (for the non-biophysicists among us, this means the predicted structure was good, very very good in fact.), points towards a future in which yes, in at least some if not most cases we can use computers to calculate a *likely* protein fold. Even if, as you correctly state, the combinatorial complexity of the problem makes it practically insoluble in an absolute mathematical sense.
Also, Monte Carlo is, I believe, more than 'somewhat randomized'. It is completely random. You just keep sampling until the answer shows itself.
I'll give you that, but let's be realistic here; anyone using a Monte Carlo method is probably doing so in conjunction with some kind of rejection rules which limit your moves in some way, such as the Metropolis Criterion for starters. Thus, your Monte Carlo search is no longer so random. You're most likely well aware of this, but I've got to defend myself when pickers of nit take up arms.
posted by StrangerInAStrainedLand at 11:41 PM on May 12, 2008
Hey, that link to results is nifty. And, given the likely search space for n=100 is really large, he must have some really good search heuristics.
Ah, MCMC: it's not a roman numeral! I wasn't trying to pick nits to that degree, I apologize.
I really do wonder if any generally applicable insight can be taken from human action on protein folding. Know of any complexity theorists that are working on this sort of problem? I mean, I could go hit up the internets, but if you know off the top of your head...
posted by enkiwa at 10:53 PM on May 13, 2008
Ah, MCMC: it's not a roman numeral! I wasn't trying to pick nits to that degree, I apologize.
I really do wonder if any generally applicable insight can be taken from human action on protein folding. Know of any complexity theorists that are working on this sort of problem? I mean, I could go hit up the internets, but if you know off the top of your head...
posted by enkiwa at 10:53 PM on May 13, 2008
Been playing this over the weekend - and it's brought out my inner OCD (as if I need much pushing for it to come out.)
It really is a great idea, playing on the idea that humans are generally better at naturally seeing overall patterns.
++do try
posted by mincus at 12:58 PM on May 14, 2008
It really is a great idea, playing on the idea that humans are generally better at naturally seeing overall patterns.
++do try
posted by mincus at 12:58 PM on May 14, 2008
« Older When CyberSquatting Laws go bad.... | An unbiased, informative site for heated issues Newer »
This thread has been archived and is closed to new comments
posted by lumensimus at 5:19 PM on May 10, 2008