Slashdot Mirror
Convergence of Biology and Computers?
Pankaj Arora asks: "This summer I am working on both Bioinformatics and Molecular Biology research projects at the Mayo Clinic Rochester. Being an MIS major with a heavy CS background, I've been learning about biochemistry performing polymerase chain reactions (PCRs) and RNA retranslation among other things. I've learned biology works a lot like computers; binary has 1s and 0s, DNA has nucleotides: A, T, C, and G. Binary has 8 bits to a byte, DNA has 3 nucleotides to a codon. Computers and biology seem to have a natural fit; information is encoded and represented 'digitally' in a sense. I was wondering what people thought about the future of biology-based and genetics-based computing due to the immense efficiencies that lie in nature. This has been discussed to an extent here, but there were some specific aspects that I feel are quite important and were not discussed thoroughly, thus I have a few questions to pose to the Slashdot community."
"The aspects I would like discussed are as follows:
- In the long run, will biology rewrite computing or will modern day technology concepts and theory be integrated into biology? If both are true, which will have the greater effect? I understand long run is ambiguous in this question, but Iâ(TM)m interested in all thoughts using any applicable definition.
- Tied to the first question: How will the nature of computing, and how we perceive it, change due to biology integration? More to the point, how much of the theory we learn today may change?
- What will be the biggest issue determining the success of the adoption of biology-integrated computing? Will it be technology factors or will it be societal factors (e.g., rebellion by the Right Wing), or something else? What things must hold true to make the idea succeed?
- And perhaps the hottest issue of all: Is there anything inherently wrong with pursuing this avenue? What may be some of the consequences?
Take a look at some of the work by Richard Feynmann and Freeman Dyson - the two of them discuss(ed) biology-based computation at great length, and although they were not completely encumbered by modern methods and capabilities, their insights into the theory are pretty valuable. In addition, check out Douglas Hofstadter - I believe that _Metamagical Themas_ had an article or two about this.
-David Barak
Need Geek Rock? Try The Franchise!
I don't remember the artcile, or the location of the reference [ http://www.nature.com/nsu/000113/000113-10.html thanks google]...
Well anyways, the travelling salesman problem was solved using specially crafted DNA sequences.
The first things that come to mind is, "What time frames are you speaking in for this technology?" and "What application are you talking about?" Each of these are very important.
If you are talking raw number crunching, it might end up having some problems with competition with rival technologies. The High Productivity Computational Systems Effort @ DAPRA is intended to bridge the gap between current supercomputers and quantum computers in capability. If the realistic xpectations for quantum computers are realized, and not the hype, then it might end up making the biological tech a case of an 'also ran' much like gallium arsenide seems to have become. Unless there is something that biotech processors do better than the traditional architectures and the projected quantums, then it might remain a lab curiousity.
On the other hand, if you mean something else, like revolutionary computer-human interfaces, or AI work, or something I'm not thinking of, then we might see something generated from this indeed.
If you could be more specific about what you have intended this technology applied to...
Do you know why the road less traveled by is littered with the bones of the unwary?
Bzzt! Thanks for playing.
Mutation is the grist for the mill of natural selection. Were it not for mutation, Earth would be a swamp of highly advanced algae right now. What you want is a balance between mutation and error correction - enough correction so that organisms can survive and breed, but enough mutation so that you can have variation that will allow adaptation to new niches.
there are random mutations every time a cell divides, it's called evolution.
Rice University Department of Biochemistry and Cell Biology- "Engineering the freaks of tomorrow"
An excellent book discussing some of the isomorphisms between computers and biology is Godel, Escher, Bach: An Eternal Golden Braid by Douglas Hofstadter. I can't recommend it highly enough.
I don't know half of you half as well as I should like, and I like less than half of you half as well as you deserve. BB
We didn't understand the evolved FPGA pattern implementing an XOR either, although we think of it as a bit pattern.
By the way, DNA isn't recombined with 100% fidelity. Mostly, things work ok, but things do mutate once in a while, just as they do when you make analogue recordings. This leads me to think your "digital tape recorder" analogy isn't a very good model - "DNA bits" can and do flip.
First, there is a difference between bioinforamatics and DNA computing. Bioinformatics is the application of computer algorithms and statistical techniques to figure out how a biological system works. DNA computing is more of an engineering project, since you are addapting DNA to do your computational bidding (e.g. a DNA based microprocessor)
I my self am in the field of bioinformatics/molecular biology with my primary interest being in RNA regulation and regulatory elements. I am trying to find and figure out how RNA regulation works in model systems.
Now for your questions...
>In the long run, will biology rewrite computing or will modern day technology concepts and theory be integrated into biology?
Both will happen...
>If both are true, which will have the greater effect?
I don't know about biology rewriting comuting. First, yes DNA encodes information 'like' binary 1's and 0's, but we are still figuring out the system works. We know how to find some genes by just looking at the sequences, but we still have the problems with predicting genes in a sequence (e.g. gene splicing, post transciptional events, etc.
I think it would be more sane to use the modern day technological concepts and theory, but with an emphasis on parallel computing.
>I understand long run is ambiguous in this question, but Iâ(TM)m interested in all thoughts using any applicable definition.
Tied to the first question: How will the nature of computing, and how we perceive it, change due to biology integration?
Well we can have those clean computers powered by photosynthesis... ok, all kidding aside, it change computing for those tasks DNA would excel at: A DNA computer is a type of non-deterministic computer. We have to overcome some of the problems imposed by DNA... its a chemical that is in an aqueous environment that tends to mutate over time; also the DNA computers I have seen work in a test tube, and you have to sequence it to get a result. That should hopefully change in time.
>More to the point, how much of the theory we learn today may change?
In biology - a sh*t load most likely; like I said above, we are still trying to understand biological systems and how they interact with each other, including DNA and how it codes for life.
>What will be the biggest issue determining the success of the adoption of biology-integrated computing?
Get it out of the test tube first... place it on a chip, like a microprocessor. Also the energy source... I don't want to share my doritos with my desktop...
>Will it be technology factors or will it be societal factors (e.g., rebellion by the Right Wing), or something else?
Don't like the right wing, eh? Well as a card carrying member of the vast right wing conspiracy, you have just as much to worry about from the left... those environuts who think we are tampering with nature (like we haven't been doing that for the last 10000+ years (e.g agriculture). Both extremes muzzel science... get used to it.
If we start to integrate computers into our selves... yeah I think society will have some issues to face about what it means to be human. (I'll go with David Hume with this gem "I'm human because my parents were human")
>What things must hold true to make the idea succeed?
1. Perfect DNA computing
2.
3. Profit -- of course!
Ok, seriously -- there need to be interest in the scientific community, we need to figure out how DNA works in living beings... how it encodes all its data (and how about that junk DNA?). We need to get it on a chip (not a microarray chip... some times called DNA chips). And there needs to be a profit motive.
>And perhaps the hottest issue of all: Is there anything inherently wrong with pursuing this avenue? What may be some of the consequences?
Hell no! But if you are interested in DNA computing, the bioinformatic
Accentuate the positive, don't waste your mod points on the negative.
Phi = (1+sqrt(5))/2, about 1.618 This number appears all over the place in nature, and, most interestingly, in the structure of DNA: One rung of the DNA ladder has two golden mean pentagrams, two hexagons, and a golden mean rectangle in the middle, more or less. Also, the helix of the DNA molecule advances by a vertical increment of 1.618 per turn. How's that?
It's junk DNA not because we don't know what it does, but because it's never accessed at all.
The equivalent in computer science would be if you plotted every possible route through a program and some code is still never conceivably executed, that would be the equivalent of "junk DNA". Even if you went into the machine language code and replaced it with random values, the program would still never crash because it never executes.
In the computer world we tend to call that "dead code".
Thus, we do know that the "junk" is truly junk. The debate on its usefulness centers around the other physical implications of the existance of such DNA, and where it might have come from, but "computationally" (in biological terms "is it ever used to produce a protein?") it is indeed junk.
Please consult any elementary (but up to date... the understanding of junk DNA has progressed a lot in the last decade) textbook on genetics.
Simple Guide to DNA Computers
How Stuff Works - DNA Computers
No ground breaking crypto solving or Beowulfs yet but some solid calculations going on.
Bleh!
You've asked some very broad questions which delve into both technical and social issues. I'm not much of a social theorist, but I do know something about computing and biotechnology. I'm a postdoc in a lab that studies genomics and biological regulatory networks using computational methods. There are two basic approaches to merge bio and computing: 1) You try to improve computing by using ideas or techniques from bio, and 2) You try to do something interesting in bio by using ideas from computing. Examples of (1) trying to improve computing by using bio would be such things as DNA computing or doing massive combinatorial searches in chemical solutions. In DNA computing, you use various enzymes or chemical agents to modify a DNA string. Think of it as a turing machine acting on a strip, except the strip is now a piece of DNA. Since the DNA strip is modified over the procedure, the "state function" is partially encoded in the data strip, not just internally in the chemical agent. The great advantage of DNA as a computing medium is that there are methods for selectively replicating DNA based on its "state". So you can run your chemical procedure over many different possible DNA sequences simultaneously and then only replicate the particular sequence with the desired state, which gives your answer. At the moment, DNA computing is most useful for search problems. For example, several years ago, the traveling salesman problem was tackled in a DNA system. There is a lot of research now into new operations that can be performed on DNA strings (e.g. ways of doing multiplication or addition using various enzymes and data encodings) to broaden the types of problems that can be tackled. Anyway, this is one way people are using bio to improve computing, broadly defined. In a lot of ways, this isn't really bio anymore. Scientists discovered DNA and enzymes in cells, but now we're just using them as materials for computation. People also use similar search techniques with non-biological molecules. Some similar search and amplification procedures are used to make synthetic organic compounds in drug discovery. DNA, however, is particular useful because it's a long molecule so a lot of operations can be performed on it. As far as how DNA will affect computing in the long run, I don't know. We're still very far from making a dna computer that can achieve anything like what silicon-based systems can. But there could be big technological advances eventually. I don't know of any ways that bio systems have affected our ideas about computing at a software level -- except to perhaps funnel more interest towards massive parallelism. Again, I don't want to imply pessimism about what could be invented. As for (2) how computing could affect biology, this is much less concrete. I'll interpret this to mean that one is trying to program biological systems to do something. i.e. if we give a well-defined instruction set, can we get a cell, organ, or organism to yield a particular output? This to me is just the basic problem of science -- trying to understand how stuff works. We'll be able to "program" cells, organs, or organisms if we understand them as well as we now understand the chemical properties of DNA, or even better, as well as we understand silicon-based semiconductors.
maybe you should look at this- http://mitsloan.mit.edu/news/archives/2003/50K-03. html
Binary isn't THAT efficient if you want to store information in a small space. Quaternary systems (like DNA) are more efficient space-wise.
(Simplifying wildly) DNA stores 3 base pair 'words' called codons. Each codon either codes for an amino acid (each amino acid is coded by more than one codon) - such as the sequence ATG which codes for the acid methionine; or represents a 'start gene' or 'stop gene' switch.
With three letter sequences for a codon and four possible letters for each position you end up with 64 possible codons (there are just 20 amino acids); to store the same amount in binary you would need six bases.
So DNA is actually very efficient at what it does.
Best wishes,
Mike.
I remember learning about RNA translation and PCRs in 9th grade biology. I thought the link between dna and computers was interesting so I did a search and found a Howstuffworks.com article entitled "how dna computers will work" http://computer.howstuffworks.com/dna-computer.htm
There are 3.4 Angstroms vertically between nitrogenous bases in DNA, one complete turn is 34 Angstoms, where do you get 1.618 from ?
It was Leonard Adleman (of RSA fame) who first proposed the idea of using DNA to perform simple computations in a 1994 paper entitled "Molecular computation of solutions to combinatorial problems" (you can find it here.
Adleman's DNA computer computed the answer to the Hamiltonian Path problem for a small graph. The Hamiltonian Path problem is solvable on a conventional computer, however it is an "NP-Complete" problem, which means that all known deterministic algorithims have a running time which is exponential with respect to the problem size.
Adleman's solution was to encode random paths through the graph in billions of DNA strands, then use custom engineered enzymes to eliminate those strands that were not a Hamiltonian path. Essenially, he simulated a non-deterministic machine through massive parallelism.
While this is increadibly clever, and very interesting, it isn't necissarily practical; at least, not for NP-complete problems. Adleman acheived linear execution time for an NP-complete problem, but he did so at the expense of requiring an exponential number of DNA strands with respect to his problem size. A small graph with only a few hundred nodes would require more strands of DNA than there are atoms in the universe.
This is not to say that DNA computers are of purely academic interest; Adleman's computer was merely a "proof of concept". I'm sure there exist problems in P which would benefit immensely from massively parallel computing. It's just a question of finding problems which are cost effective to implement.
Perhaps many of these "distributed" computing efforts that are underway now would better be served by a DNA computer.
I'm not questioning where he got the 1.618 from, I'm questioning where he got the information that there was a distance of 1.618 vertically between nucleotides, because the distance is 3.4, not 1.618.
What he is saying is that a single cycle of the DNA helix would fit perfectly in a cylinder 34 angstroms tall and 21 angstroms in diameter. 34/21 ~= Phi. This wasn't clear to me either but Google led me here.