Largest DNA-Based Computational Circuit Created

← Back to Stories (view on slashdot.org)

Largest DNA-Based Computational Circuit Created

Posted by timothy on Thursday June 2, 2011 @04:27PM from the artificial-one-that-is dept.

angry tapir writes "Researchers from the California Institute of Technology have built [abstract] what they claim is the world's largest computational circuit based on DNA (deoxyribonucleic acid), using a technology that they said could easily scale to even greater complexity." Adds reader cwmike: "The researchers formed 130 different synthetic DNA strands that can be used to compose logic circuits. From this source material, they created one 74-molecule, four-bit circuit that can compute the square root of any number up to 15 and round down the resulting answer to the nearest integer. In their setup, the multi-layered strands of DNA are fashioned (see video) into biochemical logic gates that can perform the basic Boolean AND, OR and NOR operations executed by today's transistor-based computer processors."

40 of 57 comments (clear)

Min score:

Reason:

Sort:

In their dreams! by NoNonAlphaCharsHere · 2011-06-02 16:29 · Score: 4, Funny

I've been running a MUCH larger DNA-based computational circuit for YEARS now! Course, sometimes it fails me completely.
1. Re:In their dreams! by Fluffeh · 2011-06-02 16:37 · Score: 4, Informative
  
  Well, yes, while DNA based computing may sound like something from Science fiction, it looks like this is a small step:
  
  The researchers formed 130 different synthetic DNA strands that can be used to compose logic circuits. From this source material, they created one 74-molecule, four-bit circuit that can compute the square root of any number up to 15 and round down the resulting answer to the nearest integer.
  
  then
  
  Reif also pointed out a few downsides. One is the speed of calculation. The execution of a single gate can take anywhere from 30 to 60 minutes. Executing a four-bit square root could take up to 10 hours.
  
  So don't be expecting any DNA based mass computing revolution anytime soon.
  
  --
  Moved to http://soylentnews.org/. You are invited to join us too!
2. Re:In their dreams! by SanityInAnarchy · 2011-06-02 17:12 · Score: 4, Interesting
  
  If you're talking about your brain, that's not what this is. This is using actual DNA to perform computations.
  Oh, by the way, last I checked, it's slow, and this is no exception:
  
  Reif also pointed out a few downsides. One is the speed of calculation. The execution of a single gate can take anywhere from 30 to 60 minutes. Executing a four-bit square root could take up to 10 hours.
  This isn't like quantum computing -- maybe they can make it faster, but I really don't see this having any inherent advantages over old-school tech like CMOS anytime soon.
  What makes this interesting is the potential to do calculations inside living systems, or to actually interface our code with otherwise strictly biological processes. These "circuits" are just solutions of custom-designed DNA, and each "gate" takes small single-strands of DNA as input, and produce them as output, whether as a "wire" to another gate, or as the final output to be measured to check if the circuit is working. Now imagine putting that in a cell. (Oh, and this is why formal methods matter -- if someone's going to be putting code in your body, it's not enough to debug it, you want that shit proven correct.)
  Disclaimer: While I did take a class (COM S 433 at ISU) which attempted to examine this stuff, this was covered at the very end of the semester, and no one (including the instructors) really had a good idea how these things actually work. I know enough to be dangerous, but there's a good chance I'm wrong about pretty much anything I say here. Read the papers yourself -- it's fascinating stuff.
  
  --
  Don't thank God, thank a doctor!
3. Re:In their dreams! by Pepebuho · 2011-06-02 17:43 · Score: 1
  
  On the other hand, imagine the parallelizing opportunities with thousands and thousands of such gates!
4. Re:In their dreams! by Fluffeh · 2011-06-02 18:41 · Score: 2
  
  I am not saying that this isn't cool, but even if you had thousands and thousands of these gates, their individual processing power seems miniscule, and given the very lengthy times required to derive a solution to a problem, parallel operations would be limited.
  While I think it is very "nifty" for them to have done this, it doesn't appear (and I am happy to be corrected here) to have any direct applications, nor does it appear to be a stepping stone to anything in anything but the very distant future. Again, while I think it is nifty that they did this, and I hope that it leads to some wonderful breakthrough in the future, it seems that they are simply mucking about, making a set of items that are totally adapted for something else - do parlour tricks.
  If anything, I am more impressed with how far genetics has come rather than the fact that they got a bunch of DNA to do some arbitary maths puzzle.
  
  --
  Moved to http://soylentnews.org/. You are invited to join us too!
5. Re:In their dreams! by buchner.johannes · 2011-06-02 18:53 · Score: 2
  
  ... last I checked, it's slow, and this is no exception: ... This isn't like quantum computing
  There shall be Quantum-DNA computing then! I hear DNA is prone to hybridisation.
  
  --
  NB: The message above might reflect my opinion right now, but not necessarily tomorrow or next year.
6. Re:In their dreams! by IDK · 2011-06-02 23:02 · Score: 1
  
  Interesting, but what if we see this as a storage device instead? Isn't the purpose of DNA to be robust, which is perfect for longtime storage, and a DNA-molecule should be very small thus enabling very high storage densities, exactly what our cells are using it for!
7. Re:In their dreams! by gilleain · 2011-06-03 00:10 · Score: 1
  
  These "circuits" are just solutions of custom-designed DNA, and each "gate" takes small single-strands of DNA as input, and produce them as output, whether as a "wire" to another gate, or as the final output to be measured to check if the circuit is working. Now imagine putting that in a cell. (Oh, and this is why formal methods matter -- if someone's going to be putting code in your body, it's not enough to debug it, you want that shit proven correct.)
  I can imagine putting this in a cell. I just can't imagine why - or how.
  The problem with most bottom-up molecular design like this is trying to replicate in miniature the macro-level machines we make. Biological design is necessarily messy; it has to work under a huge range of noisy conditions, and with components of varying quality. It's not like the well-ordered, quiet, dry environment that you find in a computer.
  For example : what would happen if the DNA computer starts to get transcribed? Or broken down by anti-viral defences? Or just clogs up the normal processes, like the cytoskeleton, or cellular repair? Further, how would the output of the computation be transmitted to the outside world? Perhaps by interfacing with ion channels in the membrane, or triggering synthesis of coloured chemicals?
  So that's the how : as for the why - I could see that it would be any easy way to grow biological cell-based processors. Then if you could link them up, you get massive parallel computing. Again, though, there is the problem of coordination across membranes and control of cell division and cell spatial patterning.
8. Re:In their dreams! by Nefarious+Wheel · 2011-06-03 00:49 · Score: 1
  
  I want a DNA computer that can decide I need new teeth regrown, then show my body how to do it.
  
  --
  Do not mock my vision of impractical footwear
9. Re:In their dreams! by Nefarious+Wheel · 2011-06-03 00:52 · Score: 1
  
  Bound to be a decent horror movie in this plot, though. Imagine a cultured brain that didn't stop growing.
  Hmm. "The Brain from Planet Arous". On the shelf...
  
  --
  Do not mock my vision of impractical footwear
10. Re:In their dreams! by SanityInAnarchy · 2011-06-03 16:38 · Score: 1
  
  I can imagine putting this in a cell. I just can't imagine why - or how.
  I can imagine a few 'why's. Right now, in the barely-related field of DNA Origami, people are building machines out of DNA which target cancer cells -- I've heard it described as a "box" filled with chemotherapy poison, opens when it detects a cancer cell. The results are pretty jaw-dropping, if maybe a bit preliminary -- something like a 98% kill rate on the cancer cells without touching the healthy cells.
  I'm being deliberately vague, because I'm already out of my depth on the comp sci side, and the bio side is just that much farther beyond me. I understood about half of what you said, so I'll defer to you if you don't see it being feasible to put this in a cell.
  But here's the problem from the comp sci side: These just don't seem likely to be useful for computing alone, unless I'm right about it being useful to do this with DNA somehow (inputs and outputs are strands of DNA, etc). But this:
  
  I could see that it would be any easy way to grow biological cell-based processors. Then if you could link them up, you get massive parallel computing.
  Well, let's speculate on this... Let's be generous and assume we can get one usable circuit per core -- keep in mind that while these interactions are happening in parallel, you also need a reasonable concentration of each component if you want them to actually interact, so the circuit itself isn't necessarily parallel. Let's also be generous and assume a minute (not 10 hours) to perform an operation, since square root is a fairly complex operation.
  A quick Google suggests something like 100 trillion cells in the human body. So, we have 100 trillion operations per minute, or about 1.7 trillion operations per second, or 1.7 terahertz. Even if we're generous and call it 1.7 teraflops, that's 1.7 teraflops in a structure the size of a human body.
  And Intel got 1 teraflops on a 40W chip, four years ago. While the power requirements of the human body are somewhat more difficult to estimate, another random Google puts that at something like 80W. And just look at the size of the wafer in that video, versus the human standing next to it -- the wafer is barely bigger than his thumb.
  So, I think I'm giving this optimistic numbers, and while it's not coming back as "You're crazy, that'll never work!" (other than the 1 minute per op instead of 10 hours), it still looks like CMOS 4 years ago beats DNA years from now.
  
  --
  Don't thank God, thank a doctor!
11. Re:In their dreams! by SanityInAnarchy · 2011-06-03 17:04 · Score: 1
  
  You might be thinking of algorithmic self-assembly... and, sort of.
  That class I mentioned spent most of our time working with an abstract tile assembly model. We had assignments to build programs and run them on one of these simulators, and otherwise spent most of the time proving stuff about them.
  Unfortunately, while it's really easy to learn how to build these, I'm not sure there's a tutorial I can point you to.
  The good news is that it is massively parallel. It's related, somewhat, to cellular automata, or Conway's Game of Life, except that it's strictly growth -- once a tile attaches, there's normally no way to detach it. But the idea is that anywhere a tile can attach, it eventually will, and fairly quickly -- while you do often have entirely sequential algorithms, your algorithm can theoretically scale with the number of locations tiles can currently attach. It's also Turing-complete, and kind of makes for a cool visualization of a Turing machine. You can also just make cool fractal patterns instead of trying to compute anything.
  The bad news is that errors happen quite often, and there's no good way to correct or avoid them (yet). There are certain shapes which are impossible to build. The parallelization isn't by any means free -- one interesting example was a Turing machine with a binary alphabet which operates simultaneously on the entire set of natural numbers (or rather, spawned a new Turing machine for each natural number), with appropriate "output" to check for whether a given machine had halted, but each new machine runs at half the speed of the previous one, otherwise they'd run out of room. While tiles are cheap, tile types are not -- it's apparently expensive to synthesize DNA, but cheap to replicate it. And to do anything that isn't just compiling a Turing machine (which produces an impossible number of tile types), while technically algorithmic, is incredibly difficult and doesn't seem to benefit in any way from any of the languages or tools we've developed for programming.
  This isn't just "parallel programming is hard." It's not just a different language or a different syntax. It's more like going back to drawing logic circuits by hand, only none of the components bear any resemblance to anything you've used before.
  So, it's not likely anyone's going to be using these to mine bitcoins or crack RSA anytime soon. On the other hand, it's at the stage right now where you can still come in with a crazy idea that no one ever thought of before, and move the entire field by yourself. I suppose that's true of any field, but most fields, everyone's already thought of the more obvious crazy ideas.
  
  --
  Don't thank God, thank a doctor!
A single logic gate operation takes 30-60 minutes by parallel_prankster · 2011-06-02 16:35 · Score: 3, Funny

Thats exactly how my DNA-based computer works too :)
Did you say 'de oxy ribo nucleic ACID'? by joocemann · 2011-06-02 16:58 · Score: 1

That sounds like dangerous stuff! Its an ACID! That could hurt people too!
I sure hope they are very safe about these circuits and have warnings all over them. Something about this sciency nerdy 'deoxy' whatever the hell you called it ACID is going to get us killed. God bless us all. ;)
1. Re:Did you say 'de oxy ribo nucleic ACID'? by RoFLKOPTr · 2011-06-02 17:15 · Score: 2
  
  That sounds like dangerous stuff! Its an ACID!
  I dropped some DNA last night and tripped some balls.
2. Re:Did you say 'de oxy ribo nucleic ACID'? by NoNonAlphaCharsHere · 2011-06-02 17:17 · Score: 2
  
  Chill. Take two tablets of acetylsalicylic acid with a glass of hydric acid and wait for the palpitations to pass.
3. Re:Did you say 'de oxy ribo nucleic ACID'? by somersault · 2011-06-02 21:11 · Score: 2
  
  My DNA tripped last night and my balls dropped.
  
  --
  which is totally what she said
"I compile to gene" by Anonymous Coward · 2011-06-02 17:09 · Score: 1

While the speed presents an inherent limitation, the fact that they made a compiler for it is seriously cool.
1. Re:"I compile to gene" by c0lo · 2011-06-02 18:39 · Score: 1
  
  While the speed presents an inherent limitation, the fact that they made a compiler for it is seriously cool.
  Suggestion: wait until they'll implement the -O3 flag in that compiler (10 hours for an execution is not so exciting).
  
  --
  Questions raise, answers kill. Raise questions to stay alive.
Once upon a time there was a magic book... by VortexCortex · 2011-06-02 17:18 · Score: 1

[What the hell are you talking about you liars! There is not now, nor has there ever been a magical book -- magical thinking perhaps...]
Blah blah piano too loud for me to hear you say: "In this book, 'the magic spirits', which are small ... DNA, can be represented as colored lines with arrow heads marking their" -- ffffFFFFFUUU! (I lost)
(See video link in TFS -- you rage, you loose.)
Science: it's a tough life by Walt+Dismal · 2011-06-02 18:07 · Score: 4, Funny

I keep telling girls at bars I'm a scientist who just wants to perform a complex calculation using DNA and then they throw a drink in my face for some reason.
4-bit square root? by sgunhouse · 2011-06-02 18:17 · Score: 2

Seems pretty trivial to me. They did say round down, so ...
0 --> 0
1 --> 1
2 --> 1
3 --> 1
4 --> 2 ...
8 --> 2
9 --> 3 ...
15 --> 3
For most inputs, the answer matches the highest 1 bit in the number (exceptions: 0, 1 and 9). Wouldn't be hard to make a circuit do that.
10 hours? As in, 20 operations? Seems excessive...
1. Re:4-bit square root? by maxwell+demon · 2011-06-02 19:00 · Score: 2
  
  Actually, if we name the input bits i0 (LSB) to i3 (HSB) and the output bits o0 (LSB) and o1(HSB) then the calculation to do is:
  o1 = i4 OR i3
  o0 = (i4 OR NOT i3) AND (i1 OR i2 OR i3)
  Indeed, not very complicated.
  
  --
  The Tao of math: The numbers you can count are not the real numbers.
2. Re:4-bit square root? by angel'o'sphere · 2011-06-02 21:02 · Score: 1
  
  Which implies that the input is encoded in binary ... which it isn't.
  Sometimes it is really interesting and helpfull to RTFA.
  angel'o'sphere
  
  --
  Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
3. Re:4-bit square root? by maxwell+demon · 2011-06-02 22:27 · Score: 1
  
  Which implies that the input is encoded in binary ... which it isn't.
  From the Science abstract (emphasis by me):
  
  we experimentally demonstrated several digital logic circuits, culminating in a four-bit square-root circuit
  They were encoding numbers into bits and didn't use binary?
  
  --
  The Tao of math: The numbers you can count are not the real numbers.
4. Re:4-bit square root? by angel'o'sphere · 2011-06-02 23:08 · Score: 1
  
  Well, that is how articles are written I assume.
  If you read further you see that they use "standard DNA" chemical reactions. DNA fragments (data) are docking on the DNA gates (circuits). The input is read/decoded and transformd into new DNA strands which are able to dock at the next circuit. There the same process continues. Over several steps the original input is decoded and reconstructed into a computational result which happens to be the square root of the input.
  It is not 100% clear but for me that idicates that the input and output is not coded binary. (In other words the mentioning of 4-bits is only an analogon to be able to compare it with current technology)
  angel'o'sphere
  
  --
  Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
5. Re:4-bit square root? by Muad'Dave · 2011-06-03 00:28 · Score: 1
  
  I think your logic gates are backwards. I used espresso and came up with (Inputs A-D, MSB to LSB and outputs X and Y, MSB to LSB, NOT has highest precedence):
  X = A OR B
  Y = (NOT B AND D) OR (NOT B AND C)
  I wrote out the truth table, and my version seems right.
  
  --
  Tiller's Rule: Never use a word in written form that you've only heard and never read. You will end up looking foolish.
6. Re:4-bit square root? by martyb · 2011-06-03 02:37 · Score: 1
  
  Actually, if we name the input bits i0 (LSB) to i3 (HSB) and the output bits o0 (LSB) and o1(HSB) then the calculation to do is:
  
  o1 = i4 OR i3 o0 = (i4 OR NOT i3) AND (i1 OR i2 OR i3)
  
  I like what you did there! But... i0 is not used, and i4 is not defined. If we adjust the formula by shifting the indices down by one so that we have:
  
  o1 = i3 OR i2 o0 = (i3 OR NOT i2) AND (i0 OR i1 OR i2)
  
  Then it seems to work just fine! (Please forgive the appearance; workaround to overcome /.'s formatting and lameness filter limitations.)
  
  _i i3_i2_i1_i0 o1 o0 s _0 _0__0__0__0 _0__0 0 _1 _0__0__0__1 _0__1 1 _2 _0__0__1__0 _0__1 1 _3 _0__0__1__1 _0__1 1 _4 _0__1__0__0 _1__0 2 _5 _0__1__0__1 _1__0 2 _6 _0__1__1__0 _1__0 2 _7 _0__1__1__1 _1__0 2 _8 _1__0__0__0 _1__0 2 _9 _1__0__0__1 _1__1 3 10 _1__0__1__0 _1__1 3 11 _1__0__1__1 _1__1 3 12 _1__1__0__0 _1__1 3 13 _1__1__0__1 _1__1 3 14 _1__1__1__0 _1__1 3 15 _1__1__1__1 _1__1 3
  
  What I'd like to know: is that the MINIMAL set of logic operations required to compute this? How could one prove it?
7. Re:4-bit square root? by munozdj · 2011-06-03 03:10 · Score: 1
  
  What I'd like to know: is that the MINIMAL set of logic operations required to compute this? How could one prove it?
  Apply Karnaugh maps or the Quine-McCluskey method to find the minimal set. It might not be unique, but it'll be minimal.
  
  --
  Democracy: Crowdsourcing a country near you
Re:Destroy it! by maxwell+demon · 2011-06-02 18:41 · Score: 2, Insightful

Just look at the word: "deoxyribonucleic acid".
First: deoxy. Removal of oxygen. Now we know oxygen is important for life. This will make us all suffocate!
Then: nucleic. It's nuclear! Not only will we suffocate, we will also be irradiated!
Then: acid. Not only will we suffocate and get irradiated, we will also get vitriolised!
You see, three dangers in one! And who knows what hidden dangers are in the ribo part!
Oh, and they will tell you that we all have it in our bodies anyway. Well, how did it get into our bodies? It's just proof that the contamination is already far too widespread!
And sometimes you will even get a biologist admit that our death is preprogrammed in the DNA in our body. Therefore, remove all the DNA, and you'll live forever! :-)

--
The Tao of math: The numbers you can count are not the real numbers.
Ooooo! by Grindalf · 2011-06-02 19:15 · Score: 1

You are meddling with forces that you cannot possibly control or understand! And erm, can you do one in Mac? :0)

--
The purpose of existence is to make money.
Re:Thanks for the clarification by catmistake · 2011-06-02 20:16 · Score: 1

But what is this "California Institute of Technology?" Is that anything like Caltech?

--
The Admin and the Engineer
Hmmmm by rossdee · 2011-06-02 20:27 · Score: 1

What is a blue whale?
Towel? by Kamiza+Ikioi · 2011-06-02 22:13 · Score: 1

I've been running a MUCH larger DNA-based computational circuit for YEARS now! Course, sometimes it fails me completely.
That's because you need to wipe it off with a towel when you are finished or the keyboard gets too sticky to operate. ;)

--
I8-D
Not ground breaking by belthize · 2011-06-03 00:19 · Score: 1

The mice have been doing this for ages.
Synaptic MultiProcessing by sgt+scrub · 2011-06-03 00:55 · Score: 1

I wonder if they can get multiple strands to work together for processing. I don't know how they would get around the body rejecting it but I could see it being useful as an implant. For one, it could send impulses at calculated times to initiate motor response. That would require a boat load of computation though.

--
Having to work for a living is the root of all evil.
Douglas Adams, prophet by ErnyCowan · 2011-06-03 01:24 · Score: 1

The Hitch Hiker's Guide to the Galaxy immediately came to mind. Fact emulates fiction .... So who are the white mice? Has anyone seen a porpoise recently?
A computer based on this by imric · 2011-06-03 01:34 · Score: 2

would give 'computer virus' new meaning.

--
Paranoia is a Survival Trait!
Re:question by MokuMokuRyoushi · 2011-06-03 01:36 · Score: 1

if you're gonna use a ./ meme, at least spell linux properly =P
If you're going to pick on the use of a /. meme, at least spell /. properly.

--
Humans are terrible replicators of Godly things.
*New* DNA-based computational ciruit discovered. by gestalt_n_pepper · 2011-06-03 03:06 · Score: 1

Called a brain. While little used by journalists or politicians, dolphins have found them quite useful.

--
Please do not read this sig. Thank you.