BTC MINER uses MASSIVE HASHING POWER to CURE CANCER!

[iCEBREAKER]

Great, im glad other people have been thinking about this too. the bitcoin community should start working with stanford on an open source project to make it all happen.

"Bitcoin miners save the world! All crypto currency prices soar as all venues start accepting these coins in their never ending quest to be the ultimate currency"

Let's get real and forget all this altruism and donation BS.

If you really want people to use their resources for finding cures, make it profitable.

Check out BOINC.  It's what SETI@Home evolved into.

http://en.wikipedia.org/wiki/Berkeley_Open_Infrastructure_for_Network_Computing

https://boinc.berkeley.edu/

I propose a BOINCoin, redeemable (at market rates) for computer power, to monetize and incentivize an infrastructure for distributed computing.

Kind of like what Coinlab wanted to do, but building on successful existing tools instead of re-inventing the wheel, engine, and automatic transmission.   

[cygnusxi]

Folding is, or at least, should be amenable to a proof-of-work function. I'm studying molecular biology, so I can give more background.

Basically, folding is PoW in that:
1. It's difficult to find a configuration that is "better" then the best so far. By "best", I am referring to the minimum free energy, or a similar score system, where the atoms in the protein are in a most energetically favorable state. That is analogous to "difficulty" - If difficulty increases, then the free energy criteria required to satisfy the "folded" criteria is smaller. This problem is difficult because of the vast configuration space of possible positions for each of the protein residues (something on the order of 3^300), otherwise known as Levinthal's paradox.
2. It's relatively simple to compute the free energy of the system. (More difficult than a hash, but much simpler than trying to find the minimum free energy structure).
3. If someone manages to "cheat" (find a structure with a good MFE in substantially less time than the rest), well that's called a scientific breakthrough. And that someone gets rewarded with a lot of coins. Probably the most direct incentivizing of R&D possible.

Check out http://en.wikipedia.org/wiki/Foldit some time. Interesting concept.

Wow. This is exactly the kind of stuff that proves that if bitcoin devs and scientists shared some info you can really make it happen.

Love the idea people keep mentioning of a block chain that has normal mining blocks, and then "Folding blocks". the "normal mining blocks" keep the coins secure. since it would be a large project unlikely to be hit by 51% attack. Would it be possible to decrease difficulty of the normal mining blocks to almost nothing, so that the transaction system stays fast. that way the mining part of the coin would not steal computing power from the folding process.

the mining blocks would have no need to go up in difficulty as the folding blocks / projects vary in size. the size of the folding block / project completed could determine the reward in CureCoins.

Making the mining blocks pay out nothing would stop people from trying to bypass the folding blocks. so that solves another problem someone brought up.

[cygnusxi]

Great, im glad other people have been thinking about this too. the bitcoin community should start working with stanford on an open source project to make it all happen.

"Bitcoin miners save the world! All crypto currency prices soar as all venues start accepting these coins in their never ending quest to be the ultimate currency"

Let's get real and forget all this altruism and donation BS.

If you really want people to use their resources for finding cures, make it profitable.

Check out BOINC.  It's what SETI@Home evolved into.

http://en.wikipedia.org/wiki/Berkeley_Open_Infrastructure_for_Network_Computing

https://boinc.berkeley.edu/

I propose a BOINCoin, redeemable (at market rates) for computer power, to monetize and incentivize an infrastructure for distributed computing.

Kind of like what Coinlab wanted to do, but building on successful existing tools instead of re-inventing the wheel, engine, and automatic transmission.   

great stuff there too! maybe all of these projects from Folding, Boinc, Seti and the rest of them should all make their own coin and compete for the bitcoin worlds hashing power.

they could make "BionCoin","FoldingCoin","SetiCoin" etc

[FiatKiller]

Not quite as altruistic as trying to cure cancer, but there is this also: www.mersenne.org

Great as a stress test!

[jimhsu]

"Some" amount of centralization will be required for the initial phase of this project, but this is what I see:

Folding blocks: (centralized portion in underline)
- A consensus of N proteins that need structures is arrived at (by the scientific community in collaboration with the coin authors). This times the block reward is the total # of coins in the pool.
Let's say 10000 proteins * 100 reward = 1M coins. We may also want a % to go towards some sort of bounty (below)
- Proteins are introduced at some fixed rate, and the order is determined by SHA256 of the previous block (folding or mining). This to prevent precomputation of folding work, and also prevents too much early hoarding.
- Difficulty retargets per last folding block.
- Once all proteins have been "solved", clients can attempt to solve existing blocks with the "current" difficulty at half block reward. This would be some sort of ratio between the "already solved difficulty" that would determine the new difficulty target. The reward can halve sequentially for each additional attempt, until it goes to zero.
- By default, clients will automatically attempt to solve proteins with the lowest difficulty.

Bounty system: (decentralized)
- Once all proteins in the initial 10000 set have been solved, scientists that want a new structure to be solved can post it as work with any arbitrary block reward / difficulty. These coins come from them personally (hence the above % for the bounty)
- Clients will automatically compete for these block rewards
- Once these proteins are solved, they too go into the "pool" that can be solved at sequentially higher difficulty. they don't go into the pool, unless the researcher puts up additional block rewards (this could happen automatically).

Mining blocks:
- Only for transaction processing
- No block reward
- Miners get transaction fees, of course.
- Difficulty is decoupled from folding difficulty (PPC style).

[jimhsu]

Folding is, or at least, should be amenable to a proof-of-work function. I'm studying molecular biology, so I can give more background.

Basically, folding is PoW in that:
1. It's difficult to find a configuration that is "better" then the best so far. By "best", I am referring to the minimum free energy, or a similar score system, where the atoms in the protein are in a most energetically favorable state. That is analogous to "difficulty" - If difficulty increases, then the free energy criteria required to satisfy the "folded" criteria is smaller. This problem is difficult because of the vast configuration space of possible positions for each of the protein residues (something on the order of 3^300), otherwise known as Levinthal's paradox.
2. It's relatively simple to compute the free energy of the system. (More difficult than a hash, but much simpler than trying to find the minimum free energy structure).
3. If someone manages to "cheat" (find a structure with a good MFE in substantially less time than the rest), well that's called a scientific breakthrough. And that someone gets rewarded with a lot of coins. Probably the most direct incentivizing of R&D possible.

Check out http://en.wikipedia.org/wiki/Foldit some time. Interesting concept.

For the technical stuff:

1. Structures to be solved will be determined by consensus, and are distributed as PubMed ID's / NCBI reference sequences (this is an example: http://www.ncbi.nlm.nih.gov/protein/YP_005795070.1 - the bottommost part with a lot of random letters is the actual sequence; the other stuff is identifying information that is useful generally and also for certain domain-based/homology-based folding algorithms (too complicated to discuss here).)
2. Clients download this, begin folding work (using any algorithm). Folding is a vast topic and the # of algorithms is just as vast.
3. Structures that are below a certain minimum free energy (need to determine the exact equation for this beforehand) will be returned as PDBs (example: http://pdb.org/pdb/explore/explore.do?structureId=4A8N - this is a crystal structure, but a computationally generated model looks similar).
4. Other clients verify - if this structure indeed is below a certain MFE, block reward is generated.

[FrenchFriesKetchup]

Folding is, or at least, should be amenable to a proof-of-work function. I'm studying molecular biology, so I can give more background.

Basically, folding is PoW in that:
1. It's difficult to find a configuration that is "better" then the best so far. By "best", I am referring to the minimum free energy, or a similar score system, where the atoms in the protein are in a most energetically favorable state. That is analogous to "difficulty" - If difficulty increases, then the free energy criteria required to satisfy the "folded" criteria is smaller. This problem is difficult because of the vast configuration space of possible positions for each of the protein residues (something on the order of 3^300), otherwise known as Levinthal's paradox.
2. It's relatively simple to compute the free energy of the system. (More difficult than a hash, but much simpler than trying to find the minimum free energy structure).
3. If someone manages to "cheat" (find a structure with a good MFE in substantially less time than the rest), well that's called a scientific breakthrough. And that someone gets rewarded with a lot of coins. Probably the most direct incentivizing of R&D possible.

Check out http://en.wikipedia.org/wiki/Foldit some time. Interesting concept.

1 & 2 sounds good, but the current folding projects contain some thousand atoms (or proteins, don't remember right now) place them in a 3d space ... we are talking about some mb here i would say way to much for a blockchain! bitcoin exists since 2009 and needs something like 8 gb currently, on average a bitcoin block is not more than 10k or so (arverage! currently more like 150)

3 would be great (but unlikely)

what about an attacker who simply takes all available atoms and throws them into a space search with heuristics? he may come up with a solution that has low energy but is unrealistic due to some folding rules. or is it really that "simple"?

[FrenchFriesKetchup]

http://www.ncbi.nlm.nih.gov/protein/YP_005795070.1 - the bottommost part with a lot of random letters is the actual sequence; the other stuff is identifying information that is useful generally and also for certain domain-based/homology-based folding algorithms (too complicated to discuss here).)

What does it mean?

You simply take this string and start folding? Is there no other protein or something?

[jimhsu]

Folding is, or at least, should be amenable to a proof-of-work function. I'm studying molecular biology, so I can give more background.

Basically, folding is PoW in that:
1. It's difficult to find a configuration that is "better" then the best so far. By "best", I am referring to the minimum free energy, or a similar score system, where the atoms in the protein are in a most energetically favorable state. That is analogous to "difficulty" - If difficulty increases, then the free energy criteria required to satisfy the "folded" criteria is smaller. This problem is difficult because of the vast configuration space of possible positions for each of the protein residues (something on the order of 3^300), otherwise known as Levinthal's paradox.
2. It's relatively simple to compute the free energy of the system. (More difficult than a hash, but much simpler than trying to find the minimum free energy structure).
3. If someone manages to "cheat" (find a structure with a good MFE in substantially less time than the rest), well that's called a scientific breakthrough. And that someone gets rewarded with a lot of coins. Probably the most direct incentivizing of R&D possible.

Check out http://en.wikipedia.org/wiki/Foldit some time. Interesting concept.

1 & 2 sounds good, but the current folding projects contain some thousand atoms (or proteins, don't remember right now) place them in a 3d space ... we are talking about some mb here i would say way to much for a blockchain! bitcoin exists since 2009 and needs something like 8 gb currently, on average a bitcoin block is not more than 10k or so (arverage! currently more like 150)

3 would be great (but unlikely)

what about an attacker who simply takes all available atoms and throws them into a space search with heuristics? he may come up with a solution that has low energy but is unrealistic due to some folding rules. or is it really that "simple"?

A PDB (structure) is ~500kB. DOn't know if that's too big or small, but for 10000, that is 5GB, less w/ compressing (compression ratio ~ 5:1 with 7-zip). Of course for additional structures beyond the initial set, this will increase further, don't know how much.

No, folding is not that simple. Side chain residues only fit in certain ways; otherwise the sterics will blow the thing apart. MFE also takes into account bond lengths, charge, torsion and dihedral angles, hydrophobicity, all the way to to things like implicit solvation models (which folding@home uses a simplified version of).

That said, you mention a valid concern that proteins may adopt "low energy" configs that are unrealistic. This mainly applies for very dynamic proteins, proteins that form multi-subunit complexes, transmembrane proteins, etc which all require special "rules" to fold properly, but this is kinda beyond my expertise here. To alleviate that, careful selection of the initial set of proteins that are stable, have reasonable dynamics, etc. will be required (by people smarter than me). The bounty system that I proposed can help if scientists can specify specific criteria, that can be checked easily (e.g. must have 2 alpha helices, Ser273 must not be in hydrophobic core, etc). This is complicated though, so I don't expect to see this in the first iteration of the client.

http://www.ncbi.nlm.nih.gov/protein/YP_005795070.1 - the bottommost part with a lot of random letters is the actual sequence; the other stuff is identifying information that is useful generally and also for certain domain-based/homology-based folding algorithms (too complicated to discuss here).)

What does it mean?

You simply take this string and start folding? Is there no other protein or something?

That string describes the amino acid residues on the protein (i.e. L = leucine, E = glutamic acid, etc). Any mediocre biologist has these memorized. See http://bioinformatics.istge.it/bcd/Curric/PrwAli/_18604_tabular513.gif . A protein is made up of a string of these, terminated by a NH3 (N-terminal side) and a COOH (C-terminal side). Other codes can be used to denote post-translation modifications (beyond the scope of this forum).

As for a real project, I'm far too busy (medical school + graduate school) to seriously program anything, but I'm willing to look at code / give advice / refer to people smarter than me for further advice.

[jimhsu]

For those w/ a computing background, this is one of the original papers from the Pande group (the guys behind Folding@home):

https://mega.co.nz/#!SY8DQJiJ!SZS_HC_-DQBiQJxx1N3dPF-SVw0wYLPFZx8yp_g_JKw

They go into some of the actual algorithms/equs to implement folding@home on GPUs.

[cygnusxi]

@ jimhsu   You're explanation of the folding process has been awesome. This sounds like something that might end up being a lot easier that previously thought... ( as argued by some people on this thread)

1 You convinced me folding and proof of folding is not something people can easily cheat

2 It was previously mentioned that if someone figured out a way to "Solve" the work quicker, then that would actually be a giant breakthrough in the terms of finding cures faster, and that person with the brains to figure out how to solve this proteins faster is a genius, and should probably get a noble peace prize for sharing this much faster way to solve cures.

3 So there really is no cheating. if you solve faster your a hero.

[Stoneysilence]

My questions would be, would this new coin be ASIC capable or Scrypt based so GPU's only can use it?  Ideally for research purposes and speed ASIC would be ideal.  But then that would limit the number of people able to mine it.  And wouldn't it be profitable for a drug company to get into Mining with a ton of ASICs to push out all the independent small miners?  And they would reap the benefit of the coins so they are getting paid and furthering the research?

Can you imagine the amount of TH/s a billion dollar pharmaceutical company could do if they bought out one of the small startup ASIC miners and ran an army of machines?

[jimhsu]

An ASIC for folding ... I'd love to see that. This isn't a hash function we're talking about here. Folding (at least decent algos) is quite a bit more complicated (even talking about the basic molecular dynamics calcs, much more if you're going to introduce solvation models, etc). Also, folding is very much a floating-point intensive operation, which is really tough to do on ASIC, unlike hash functions. The more fundamental problem is that most folding algos are amenable to parallelization only up to the # of atoms in your system - if your system doesn't have more than 2000 atoms, for instance, then 2000 or 4000 cores doesn't really make a difference.

That said if a pharma really wants to invest millions into this project, well that would simply serve to validate its existence. Similarly to how ASIC development is supporting BTC's existence.

And no, folding doesn't "cure cancer". It does though make the development of targeted drugs against proteins a whole lot easier (for cancer as well as other diseases). If you're going to hit something, it sure is helpful to know what you're actually hitting, and that's part of the problem that protein folding can address. The field as a whole has moved on from studying intact proteins to trying to study the effects of mutations (i.e. why does mutated B-amyloid aggregate more in Alzheimer's?), or trying to do drug-protein docking (to find better inhibitors, etc).

BTW, there are already competitions for protein folding, the most well known of which is CASP: http://predictioncenter.org/casp10/index.cgi ... every year, they give out sequences to researchers all over the world, and hold a structure in secret that has been solved (with X-ray crystallography or EM) before, but not published. Scientists then publish their predictions, and they are compared to the structures that they have for a ranking, which are then released.

[cygnusxi]

Ive been exchanging some emails with a Stanford professor. He wants to know what kind of interest people will have in this... I set up a small poll asking how much of your hashing power you would use to mine a coin that solve cures.

Please visit this poll and vote. I put it on a sub domain I set aside for a mining pool. No need to log in just click your choice.

http://miningpool.marketsmash.com/demo_3.php

It sounds like if the bitcoin community shows enough interest... it will happen. so, if you want to mine for a cure and put the energy bill to better use, you better vote or post your comments here on this thread.

Maybe one of the moderators could be kind enough to host this poll right here on the bitcointalk forum!

[CoinHoarder]

This is a wonderful idea.

I would put my full support behind a coin whose intention is to help others!

[Stoneysilence]

An ASIC for folding ... I'd love to see that. This isn't a hash function we're talking about here. Folding (at least decent algos) is quite a bit more complicated (even talking about the basic molecular dynamics calcs, much more if you're going to introduce solvation models, etc). Also, folding is very much a floating-point intensive operation, which is really tough to do on ASIC, unlike hash functions. The more fundamental problem is that most folding algos are amenable to parallelization only up to the # of atoms in your system - if your system doesn't have more than 2000 atoms, for instance, then 2000 or 4000 cores doesn't really make a difference.

Thanks, guess my ignorance was showing. :blush:

[jimhsu]

This is an initiative that definitely needs community as well as external support. It isn't as simple as tweaking a few lines of code in bitcoin-qt. And glad to see that Stanford is behind this, at least in part. As a former Cal bear though, we have our differences. Lol

[skull88]

Ive been exchanging some emails with a Stanford professor. He wants to know what kind of interest people will have in this... I set up a small poll asking how much of your hashing power you would use to mine a coin that solve cures.

Please visit this poll and vote. I put it on a sub domain I set aside for a mining pool. No need to log in just click your choice.

http://miningpool.marketsmash.com/demo_3.php

It sounds like if the bitcoin community shows enough interest... it will happen. so, if you want to mine for a cure and put the energy bill to better use, you better vote or post your comments here on this thread.

Maybe one of the moderators could be kind enough to host this poll right here on the bitcointalk forum!

I quit mining a while ago, probably pick it back up again and invest in some new mining hardware for this one.
Great to see this idea going somewhere.

[FiatKiller]

This is an initiative that definitely needs community as well as external support. It isn't as simple as tweaking a few lines of code in bitcoin-qt. And glad to see that Stanford is behind this, at least in part. As a former Cal bear though, we have our differences. Lol

It would be good press for digital  currency instead of just the silk road/hit order stuff we read now.

[iCEBREAKER]

This isn't a hash function we're talking about here. Folding (at least decent algos) is quite a bit more complicated (even talking about the basic molecular dynamics calcs, much more if you're going to introduce solvation models, etc). Also, folding is very much a floating-point intensive operation, which is really tough to do on ASIC, unlike hash functions. The more fundamental problem is that most folding algos are amenable to parallelization only up to the # of atoms in your system - if your system doesn't have more than 2000 atoms, for instance, then 2000 or 4000 cores doesn't really make a difference.

Folding and other molecular dynamic computations have great potential for replacing SHA/Scrypt in a some kind of CureCoin.  They are memory intensive, and not trivially parallelizable.