Stanford Team creates biological "transistor" inside a living cell.
March 29, 2013 7:36 PM Subscribe
Have created "logic gates" they call “Boolean Integrase Logic,” or “BIL gates” for short. Original article in Science.
This is same team that created DNA storage and what they are calling a "biological Internet" before.
This means soon we'll be able to send cat pictures ON ACTUAL CATS.
posted by JHarris at 8:12 PM on March 29, 2013 [7 favorites]
posted by JHarris at 8:12 PM on March 29, 2013 [7 favorites]
If only we could create some kind of molecular computer that could solve protein docking simulations.
posted by benzenedream at 8:18 PM on March 29, 2013 [11 favorites]
posted by benzenedream at 8:18 PM on March 29, 2013 [11 favorites]
Hey, that's the guy who gave the talk I posted in in 2008. Video still available on YouTube.
His big thing then, and apparently now, was taking synthetic biology beyond genetic sequences to the realm of process modules, much as a computer program might call a function or procedure.
posted by Kid Charlemagne at 8:35 PM on March 29, 2013
His big thing then, and apparently now, was taking synthetic biology beyond genetic sequences to the realm of process modules, much as a computer program might call a function or procedure.
posted by Kid Charlemagne at 8:35 PM on March 29, 2013
"BIL gates"
In a teeny, squooshy cell? That you might call "micro-soft"?
posted by Greg_Ace at 8:37 PM on March 29, 2013 [13 favorites]
In a teeny, squooshy cell? That you might call "micro-soft"?
posted by Greg_Ace at 8:37 PM on March 29, 2013 [13 favorites]
Funny as it is, in the links there are no references to BIL gates. Seems like a NOT gate. This one does have a reference.
posted by twoleftfeet at 9:16 PM on March 29, 2013
posted by twoleftfeet at 9:16 PM on March 29, 2013
Great, soon my computer is going to sleep all day and only act nice when I need to feed it, just like my cat.
posted by jason_steakums at 9:38 PM on March 29, 2013
posted by jason_steakums at 9:38 PM on March 29, 2013
And instead of a BSOD, it will literally shit itself.
posted by jason_steakums at 9:40 PM on March 29, 2013 [4 favorites]
posted by jason_steakums at 9:40 PM on March 29, 2013 [4 favorites]
If the Internet becomes a Biological Internet, porn will only get better.
posted by twoleftfeet at 10:05 PM on March 29, 2013 [1 favorite]
posted by twoleftfeet at 10:05 PM on March 29, 2013 [1 favorite]
Let's assume that you have some non-deterministic polynomial (NP) time algorithm that's interesting. That basically means you have a way that you can check the answer to a problem quickly. There's a lot of these that are very interesting where we know how to check an answer, but we don't know how to come up with the answer. For example, the first step cryptography over the internet relies on the fact that it's difficult to factor a large number, but it being easy to check the answer once you have it. This is the essence of the P = NP question in computer science, and the answer to it (almost certainly P != NP), is a holy grail. Only a fool would publicly announce that they are on a quest to find the answer, and up until now only fools have announced that they've found it.
Now, with biological computers, we can convert any problem in NP into an actual, physical computer that solves the problem in P amount of time, i.e. all NP problems become tractable. An NP problem represents a potential solution as a string of 1s and 0s. Use C for 1 and G for 0. Program the NP checking algorithm with BILs in a cell, and during each cell cycle, perhaps right before the start of S phase (where the DNA is copied before division), have it run the NP algorithm using a specific section of the genome as the input string. Cells already have all sorts of signals to time this properly so it's easy to hook into. Start with two cells, one with a C and one with a G as the string to check. Now, if the string to check fails, you haven't found the answer to your problem, go through with S-phase, but when you copy the string to check, add one more letter to the end; one copy gets a C, one copy gets a G. This way, as the cells go through normal division, the entire space of possible strings is explored. If one cell finally finds the answer to your problem, start expressing GFP to make the cell glow bright green, and start expressing streptavidin on the membrane, so that you can easily pull out the cell. You can then sequence the cell and get the solution to your problem.
Now the magic is that the space of potential answers is explored in linear time. P sort of a little bit equals NP. You've bought an exponential increase in time performance, by paying with an exponential increase in energy consumption. The computer is self-replicating, and grows larger and larger until it's solved the problem or has consumed all food in sight.
I'm sure I'm not the only one who's thought of this. It's only a matter of time until somebody does this in practice. Up until now, yeast has formed the backbone of civilization through beer, wine, and bread, but now, yeast could be either our downfall or our breakthrough to a higher level of consciousness. Which will it be, yeast?
posted by Llama-Lime at 12:12 AM on March 30, 2013 [6 favorites]
Now, with biological computers, we can convert any problem in NP into an actual, physical computer that solves the problem in P amount of time, i.e. all NP problems become tractable. An NP problem represents a potential solution as a string of 1s and 0s. Use C for 1 and G for 0. Program the NP checking algorithm with BILs in a cell, and during each cell cycle, perhaps right before the start of S phase (where the DNA is copied before division), have it run the NP algorithm using a specific section of the genome as the input string. Cells already have all sorts of signals to time this properly so it's easy to hook into. Start with two cells, one with a C and one with a G as the string to check. Now, if the string to check fails, you haven't found the answer to your problem, go through with S-phase, but when you copy the string to check, add one more letter to the end; one copy gets a C, one copy gets a G. This way, as the cells go through normal division, the entire space of possible strings is explored. If one cell finally finds the answer to your problem, start expressing GFP to make the cell glow bright green, and start expressing streptavidin on the membrane, so that you can easily pull out the cell. You can then sequence the cell and get the solution to your problem.
Now the magic is that the space of potential answers is explored in linear time. P sort of a little bit equals NP. You've bought an exponential increase in time performance, by paying with an exponential increase in energy consumption. The computer is self-replicating, and grows larger and larger until it's solved the problem or has consumed all food in sight.
I'm sure I'm not the only one who's thought of this. It's only a matter of time until somebody does this in practice. Up until now, yeast has formed the backbone of civilization through beer, wine, and bread, but now, yeast could be either our downfall or our breakthrough to a higher level of consciousness. Which will it be, yeast?
posted by Llama-Lime at 12:12 AM on March 30, 2013 [6 favorites]
Interim Apple Chief Under Fire After Unveiling Grotesque New MacBook
"Honestly, I felt like the next logical step would be a laptop that feels like an extension of your body. The design may not be perfect, but I'm hoping over time maybe people will learn to love it, just as it will learn to love them."posted by XMLicious at 12:45 AM on March 30, 2013 [1 favorite]
You've bought an exponential increase in time performance, by paying with an exponential increase in energy consumption. The computer is self-replicating, and grows larger and larger until it's solved the problem or has consumed all food in sight.
Exponential increases ain't nothing to fuck with. Even granting that you can build such a computer--which is granting quite a bit, actually, as b1tr0t mentions--with one cell per string until you find the solution, you'll need a lot of cells.
Suppose you want to solve TSP (the decision-problem variation) for a measly 20 cities, for example, and suppose we have some perfect encoding of solutions to bit strings. There are 20! possible solutions = 2.4e18. The human body only has something like 1e13 or 1e14 cells, so already your computer could get as big as 24,000-240,000 people. For 30 cities you're up to 2.7e32, which is in the range of the number of bacterial cells on the planet. And even if it finds the solution, now you need to find that single glowing cell somewhere in your continent-sized biomass.
posted by equalpants at 12:52 AM on March 30, 2013 [5 favorites]
Exponential increases ain't nothing to fuck with. Even granting that you can build such a computer--which is granting quite a bit, actually, as b1tr0t mentions--with one cell per string until you find the solution, you'll need a lot of cells.
Suppose you want to solve TSP (the decision-problem variation) for a measly 20 cities, for example, and suppose we have some perfect encoding of solutions to bit strings. There are 20! possible solutions = 2.4e18. The human body only has something like 1e13 or 1e14 cells, so already your computer could get as big as 24,000-240,000 people. For 30 cities you're up to 2.7e32, which is in the range of the number of bacterial cells on the planet. And even if it finds the solution, now you need to find that single glowing cell somewhere in your continent-sized biomass.
posted by equalpants at 12:52 AM on March 30, 2013 [5 favorites]
which is in the range of the number of bacterial cells on the planet. And even if it finds the solution, now you need to find that single glowing cell somewhere in your continent-sized biomass.
So if I understand your calculations correctly--and please correct me if I've misunderstood because I'd hate to toss this wort needlessly--a yeastputer is going to be a 'nay' on higher level of human consciousness, and 'yay' for downfall of all civilization.
I'm not concerned about the setup time, that's like counting the design and production time of an IC CPU against its run time. Let's be realistic, afterall.
posted by Llama-Lime at 1:17 AM on March 30, 2013
So if I understand your calculations correctly--and please correct me if I've misunderstood because I'd hate to toss this wort needlessly--a yeastputer is going to be a 'nay' on higher level of human consciousness, and 'yay' for downfall of all civilization.
I'm not concerned about the setup time, that's like counting the design and production time of an IC CPU against its run time. Let's be realistic, afterall.
posted by Llama-Lime at 1:17 AM on March 30, 2013
"Hey, that's the guy who gave the talk I posted in in 2008. Video still available on YouTube."
He gave a talk at the last big phage conference in Brussels last summer that is newer and focuses more on this work specifically. Also hey I just recently wrote a blog post that might work as a good introduction to this.
posted by Blasdelb at 1:50 AM on March 30, 2013 [2 favorites]
He gave a talk at the last big phage conference in Brussels last summer that is newer and focuses more on this work specifically. Also hey I just recently wrote a blog post that might work as a good introduction to this.
posted by Blasdelb at 1:50 AM on March 30, 2013 [2 favorites]
Llama-Lime, you might be interested in this 1994 paper by Leonard Adleman (of RSA fame).
posted by Blazecock Pileon at 7:14 AM on March 30, 2013 [1 favorite]
posted by Blazecock Pileon at 7:14 AM on March 30, 2013 [1 favorite]
I love my cat, but I am not the best cat owner, as I often don't pay as much attention to him as he wants when I am on the computer. (Queue Samizdata saying "C'mon, buddy, I'M DOIN' STUFF!")
So, in the future, I will have to worry about petting my computer instead of my cat? Or the two of them beating the crap out of each other while I am out of the house in a cataclysmic battle to decide who gets all my attention?
Not so much.
Samizdata gets a bioimplant providing computing services?
Much.
posted by Samizdata at 8:34 AM on March 30, 2013 [1 favorite]
So, in the future, I will have to worry about petting my computer instead of my cat? Or the two of them beating the crap out of each other while I am out of the house in a cataclysmic battle to decide who gets all my attention?
Not so much.
Samizdata gets a bioimplant providing computing services?
Much.
posted by Samizdata at 8:34 AM on March 30, 2013 [1 favorite]
My dad heard about this lab on the radio and told me I should apply during my gap year between undergrad and grad school next year.
HAHAHAHAHA I WISH.
posted by Strass at 12:46 PM on March 30, 2013
HAHAHAHAHA I WISH.
posted by Strass at 12:46 PM on March 30, 2013
Thanks for the link, Blazecock Pileon. A nostalgic line from the paper: "The fastest super computers currently available can execute approximately 1012 operations per second," which is basically a teraflop. Today we're at ~20 petaflops, which is within an order of magnitude doubling performance every 18 months, which is the performance gain one would expect from the individual components of the supercomputer; the communication has scaled exactly on par. No big surprises.
But what's shocking is the energy consumption. In 1994: "Existing super computers are far less energy efficient; executing at most 109 operations per joule." So what's the improvement here? are we up to 1012? 1011? Nope, the top super computers are still almost exactly 2×109 operations per Joule, 20 years later. The purely DNA-based computations in that 1994 paper are 10 billion times more energy efficient than transistor-based computations. Of course, with materials prep that's not going to look as impressive, but on a mass-produced scale the material prep disappears as there would be a lot of recycling.
But it brings up a question that I've never seen answered: what's the asymptotic performance of thermodynamical processes such as DNA base-pairing? Say you have a massive vat of DNA in buffer that you use for long term storage. The best data-retreival mechanism is almost certainly going to be DNA base-pairing for your data of interest. What concentrations of request DNA are necessary for what performance? Can this be improved by slushing your DNA vat through lots of small chambers? There's a lot of cool engineering that can be done here, and almost nobody is even considering it. It's a wide open field where somebody could have lots of fundamental techniques named for them with a minimal amount of ingenuity. That DNA storage stuff is plain-Jane commodity ordering of materials from commercial services, and they're not even bothering to get into the different coding schemes that have been invented in information theory over the past half century!
And Strass, I wouldn't be so dismissive of your chances. Drew Endy is extremely personable. I was setting up my poster at a conference a few years ago, and Drew came up and started chatting with me out of the blue. Of course, having just flown in and being jet-lagged, the name Drew Endy rung a bell, but I was not able to place his name at all with his particular research. When I got back to my room, I saw that he was keynoting. This synthetic biology stuff is not my field of research, but I'm somewhat envious, as it looks like fun, and is true engineering with a ton of exciting low-hanging fruit.
posted by Llama-Lime at 1:38 PM on March 31, 2013
But what's shocking is the energy consumption. In 1994: "Existing super computers are far less energy efficient; executing at most 109 operations per joule." So what's the improvement here? are we up to 1012? 1011? Nope, the top super computers are still almost exactly 2×109 operations per Joule, 20 years later. The purely DNA-based computations in that 1994 paper are 10 billion times more energy efficient than transistor-based computations. Of course, with materials prep that's not going to look as impressive, but on a mass-produced scale the material prep disappears as there would be a lot of recycling.
But it brings up a question that I've never seen answered: what's the asymptotic performance of thermodynamical processes such as DNA base-pairing? Say you have a massive vat of DNA in buffer that you use for long term storage. The best data-retreival mechanism is almost certainly going to be DNA base-pairing for your data of interest. What concentrations of request DNA are necessary for what performance? Can this be improved by slushing your DNA vat through lots of small chambers? There's a lot of cool engineering that can be done here, and almost nobody is even considering it. It's a wide open field where somebody could have lots of fundamental techniques named for them with a minimal amount of ingenuity. That DNA storage stuff is plain-Jane commodity ordering of materials from commercial services, and they're not even bothering to get into the different coding schemes that have been invented in information theory over the past half century!
And Strass, I wouldn't be so dismissive of your chances. Drew Endy is extremely personable. I was setting up my poster at a conference a few years ago, and Drew came up and started chatting with me out of the blue. Of course, having just flown in and being jet-lagged, the name Drew Endy rung a bell, but I was not able to place his name at all with his particular research. When I got back to my room, I saw that he was keynoting. This synthetic biology stuff is not my field of research, but I'm somewhat envious, as it looks like fun, and is true engineering with a ton of exciting low-hanging fruit.
posted by Llama-Lime at 1:38 PM on March 31, 2013
Having recently read Blood Music I'm a little creeped out if still interested...
posted by treblekicker at 3:38 PM on March 31, 2013 [1 favorite]
posted by treblekicker at 3:38 PM on March 31, 2013 [1 favorite]
« Older Pay phone time machine | Keep shouting at the devil, kid. Newer »
This thread has been archived and is closed to new comments
posted by coolxcool=rad at 7:54 PM on March 29, 2013 [3 favorites]