February 17, 2001

5:03 PM

Well, if SETI@HOME is too much of a long-shot for you, then how about something absolutely certain to result in important findings? GENOME@HOME is trying to search the results of the human genome sequencing project to find the place in the genome where certain important proteins are encoded, and FOLDING@HOME is trying to figure out how proteins are folded to become enzymes, where shape is more important than chemistry.

FOLDING@HOME is working on some of the critical proteins of HIV, among other things. HIV has been sequenced and from that they know the amino acid sequences of the enzymes it makes. But without understanding their shapes it's not possible to figure out how they work. This represents one of the best applications of volunteer distributed computing I can think of. With 40,000 participants, FOLDING@HOME has already had successes, including one of the HIV enzymes. (Courtesy of Firing Squad)
posted by Steven Den Beste (14 comments total)

If SETI didn't excite you, maybe Entropia will light you up. Like SETI, Entropia offers software to convert your unused computing power into potential solutions to worldwide problems. You can help medical researchers with Fight AIDS@home, protect the environment, or explore the limits of human knowledge, and all you have to do is download the free software. The Entropia community has generated more than 700 million research hours, so here's your opportunity to use existing resources to help save the world. These distributed computing models are becoming quite popular.
posted by netbros at 7:20 PM on February 17, 2001

The problem with Entropia is that it's not what it appears to be. They're using the AIDS project as a way of getting people to run their client, but in fact they can change the app which runs on the client at any time. Once they have a large group of volunteers, they intend to sell time on the system. They can send a message to the client on your computer at any time and replace the AIDS computation with something else perhaps more sordid. I'm not impressed with that project.

That's why the ones I cited are better; you know what you'll be doing. With Entropia you don't; all you know is that you're giving time on your computer to them so that they can sell it -- and not give you any of the money. (But to find out what I just said, you have to read the fine print on their site. They certainly don't go out of their way to tell you about it.) This is not an area where I trust private industry; I'd rather work with a university if I'm going to let someone use my computer.
posted by Steven Den Beste at 8:56 PM on February 17, 2001

Blast from the past. ;)
posted by hobbes at 11:31 PM on February 17, 2001

Also, Entropia was discussed here and here, but not over here.
posted by gluechunk at 11:41 PM on February 17, 2001

When it comes to devoting cycles to great causes, my favorite continues to be the Great Internet Mersenne Prime Search. They've already found 4 primes, including the current and previous largest known prime, and will likely find more. As I look at the site, which I haven't visited in a while, I see that it's affiliated with Entropia, but I'm not sure that makes it any less cool.
posted by dcodea at 8:26 AM on February 18, 2001

I guess this is why there are horseraces. I contributed a thousand work units (about a year and a half of CPU time) to SETI@HOME and then decided to move on.

Searching for primes is intellectually interesting but rather less important from my point of view than doing basic research which will lead to medical advances. It's hard to see just exactly how finding another handful of Mersenne Primes will substantively change the human condition. (And while cracking RC5 does have political significance, I don't find it as interesting. Anyway, that political battle is largely won now so RC5 is less important than before.) On the other hand, solving the protein folding problem could lead to major medical advances, given that certain diseases are thought now to be caused by misfolded proteins ("prions"). If the folding problem can be understood, it might be possible to create enzymes which refold them into their correct form, which might lead to a treatment for Alzheimer's disease.

It might also give us a leg up on understanding certain critical viruses (e.g. HIV, influenza) by making it possible to understand exactly how the enzymes they encode work. Determining the amino acid sequence is easy with current technology, but without understanding the shape of the proteins it's difficult to figure out exactly how they function, nor how to block their activity.

So I downloaded and installed FOLDING@HOME last night, and it ran for about six hours while I was sleeping (since my computer stays on 24/7). It will get my computer from now on when I'm not doing anything else, at least until the next time Michael Colicos updates VIM.

The one thing I do know is that I won't be contributing free computer time to any commercial corporation irrespective of what compute payload they're running now. Universities, yes. Entropia, no. They can buy their own computers.
posted by Steven Den Beste at 12:30 PM on February 18, 2001

I wonder if Bill Joy runs folding@home... ;-)
posted by muppetboy at 1:29 PM on February 18, 2001

Any of these called CALIFORNIAPOWERCRISIS@HOME?

No, the client doesn't do anything. It just keeps your computer from falling asleep.
posted by dhartung at 2:14 PM on February 18, 2001

What irks me about the Folding@home page is that there is no description of their method of 'solving' folding. Sorry, but one success does not impress. There's a guy in the field who can predict folding just by looking at a protein; no program has bested him yet. My instinct is to automatically distrust anyone who claims to have solved it.
posted by methylsalicylate at 8:43 AM on February 19, 2001

They're planning on revealing the math, it's just that they haven't finished their paper yet. For that matter, they really can't publish until after the program has been running for a while, because that's how they'll check their algorithm.
posted by Steven Den Beste at 12:29 PM on February 19, 2001

I'll believe it when I see it. Word is that Pandi is little more than a charlatan. If he has indeed solved folding, my hat's off to him, but my own (largely uninformed) guess is that we don't yet know enough about the fundamentals of molecular biochemistry (and, dare I say, biophysics) to do this.
posted by methylsalicylate at 6:00 AM on February 20, 2001

perhaps its not surprising that (some) people here think "folding@home" is neat. after all, protein folding has been a "grand" challenge in science for 15 years (and haven't we all done well). there are plenty of scientists willing to pretend it would have a great impact if it were "solved" because hey, their grants depend upon it. the facts are somewhat different.

Fifteen years ago it would have been amazing. Today its a yawn. Much of protein structure space has been covered by experimentalists, not by physicists thinking this is an "easy" problem to solve with their superior intellect. Today homology modelling will get you better models than "ab initio" models will in the next 5-10 years and they STILL aren't very useful. As for structure itself, while it can have impact on drug design (especially second generation drugs) it's effect is often marginal. i know of companies that have structures of their targets and barely look at them.

So given that its not a big deal even if everything is on the level, lets look at this particular case. Pande has no reputation in the field beyond the esoteric application of lattice models, something that has proved spectacularly useless, even from the usual level of incompetence of "bio"-physicists. my sources tell me (and they could be wrong) that the code being spun out is actually Jay Ponders' "TINKER" which is open source. While good solid code there is no "novel" physics here. Let's look at their website shall we: What have they actually done? Fold a small (20 amino acid) helix which is neither novel or impressive. Fold some non-biological polymer. ditto. "Folding" of HIV (oooh, i wonder why they chose that protein) Integrase. Look carefully, it is actually the UNFOLDING of Integrase. Similarly for the "designed" peptide. the link says "Folding" the page says "unfolding". for those of you who don't know the field the difference between being able to unfold a protein and folding a protein is roughly the difference between breaking a fine piece of crystal into 1000 pieces and putting all those pieces together. to call this man a charlatan is really unfair to honest-to-god rascals. its appalling.

But i'm not even warmed up. protein folding is an intrinsically bad task for distributed processing beyond a relatively small number of processors. thats why Kollman had to go the traditional "Supercomputer" route for his villin headpiece simulation. you think peter kollman couldnt have got as much distributed processing as he wanted? the man is a god. and his simulation did a bad job of folding, i.e. it wasnt that it didnt fold, it just did a poor job. this suggests not a lack of simulation time but inadequate force fields. long time simulation= more garbage.

The issue with distributing these calculations is that the more processors = less efficient. Blue Gene is going to go to 1 processor-one atom and will build in enormous bandwidth and connectivity to overcome the bottleneck this will entail. you really think this problem has been magically "solved". you can take the multiple time-step/updating approach that others have tried (with little success) and apply that to a "NET"-based solution. get real. there IS a way to solve the protein folding problem: basic research, better understanding of physical forces, smarter algorithms, patience. folding@home is just showmanship and of the crassest calibre.

oh, and for all of you out there who still think this is a "noble", academic pursuit: folding@home is looking for venture funding to spin out and make its (academic)founders rich. a typical story into todays academic-greed-fest.

but hey, its your CPU cycles and it makes you feel good go ahead. but cut the nobler-than-thou crap. you're being had.


posted by thinlash at 5:04 PM on February 20, 2001

So you think it sounds pretty good then? :)
posted by rodii at 5:33 PM on February 20, 2001

What we should really use all this compute power for is a Perl implementation of Quake rendering to the linux ascii display driver...
posted by muppetboy at 11:37 PM on February 20, 2001