In what ways has DNA been studied to see if there a "programmable" aspect to it?

Has nature produced anything resembling a Turing machine within the cell, perhaps using the "junk DNA" as its code? I expect nature's way would probably be very round-about and not compact.

NOTE: I am not asking about building DNA computers, as this question had recently been contorted to become.

  • 5
    $\begingroup$ What evolutionary disadvantage would it give if it did not encode for anything? (not saying it doesn't, I am just not a fan of the "Nature must have thought of xyz in all these years" questions) $\endgroup$ – nico Dec 25 '11 at 8:04
  • 2
    $\begingroup$ The whole cell is definitely capable enough to exhibit all features of Turing-complete computer, so there is no reason why to use ncDNA to this task, especially when it works well as a separator and variability reservoir. $\endgroup$ – user59 Dec 25 '11 at 18:35
  • $\begingroup$ @nico If it didn’t code or regulate, it would be garbage which the cells have to drag around with them. More expensive to replicate and maintain, more material cost (just because it’s longer and requires more material to form), requires more space in the cell. There are evolutionary disadvantages to having a large, partly non-functional genome. $\endgroup$ – Konrad Rudolph Dec 29 '11 at 9:26
  • $\begingroup$ @Konrad: one could argue, however, that our body is all but efficient, and has plenty of redundant mechanisms. That said, it is known that certain mutations on ncDNA can have visible effects so, as I was saying, it may have a function, and it probably does, but from there to saying it encodes for some sort of Turing machine... $\endgroup$ – nico Dec 29 '11 at 9:45
  • 1
    $\begingroup$ @nico Redundancy is required for a fail-safe system, it provides a direct evolutionary advantage. And while our bodies aren’t particularly efficient in many aspects that cannot be controlled by evolution (laryngeal nerve …), most isolated systems under evolutionary control have been highly optimised. For instance, a eukaryotic cell’s energy turnover is orders of magnitude more efficient than any engine or generator ever created by humans (intelligently designed). $\endgroup$ – Konrad Rudolph Dec 29 '11 at 10:05

Perhaps this question is whether the regions between genes sometimes known as 'junk DNA' has any function.

In the human genome, out of ~5 billion bases there are something like 20-30,000 genes which take up perhaps 10s of millions of base pairs, depending on how you count it. 1% of all human DNA is the common figure.

It is sometimes asked as if biologists commonly think of it as having no use, but in fact this is a researched topic, and few feel that it has no evolutionary or biological function at all.

Some of the most common uses for intergenic DNA in eukaryotes (bacteria are an entirely different topic with very different responses.

  • Transcriptional Regulation

Outside the coding sequences of the gene, there can be an extensive set of binding sities for proteins which regulate the gene. In this paper in Figure 1 the celebrated regulatory sequence of ENDO16 is can be seen in Figure 1. As you can see for almoat 2000 bp there are numerous binding sites for many sorts of promotors and inhibitory factors, as well as factors which may splice the gene in various ways.

As I recall, ENDO16 only turns on for a brief period in the development of the sea urchin and so its very tightly controlled, which means that its got a lot of regulatory elements upstream of it, controlling transcription. Its one of the most exhaustively studied genes ever and the believe they have most of it. Other human genes ive seen perusing medical literature have seen 20kb are necessary to reproduce the regulation of a gene. Still all this might at best only triple the amount of DNA actively involved.

  • Centromeres and Telomeres The physiology of the chromosomes have large regions as Deniz mentions are necessary for cell reproduction and for development. In animals (like humans) the regions are devoid of transcribed genes and might be 10-15% of the genome length (I'm eyeballing this from the UCSC genome browser on chr21 - in some organisms like yeast the centromere can be just a few hundred base pairs. So anyway we are getting somewhere now!

  • Non translated genes. There are lots of pieces of DNA that might be copied into RNA and then do not function as templates for proteins. some folks say there is a lot of this stuff. The typical view is that there are a few thousand of these sorts of beasts known, and humans are currently thought to have about 1500 of them. A small tweak in the number of genes, but they are there nonetheless.

  • Chromatin binding and organization sites Although Centromeres are places where chromatin binds, the several families of proteins which bind DNA and wrap them into organized coiled coils thought to be like spools of telephone wire to form the Nucleosome which makes the chromosomes look like the little stick men you see in text books. Chromatin can spool up just about any sort of DNA but seems to have a preference for regions which are at the ends of genes. They can be modified by enzymes (acylated, methylated) to modulate their affinity for some classes of DNA sequence. This is a hot topic of research. The ability for RNA polymerase to find transcribe a gene is not good if its wound onto chromatin and though its not a precise binding like a transcription factor, chromatin binding and regulation must be affected greatly by changes in the distance between genes and the DNA sequences which surround a gene for thousands of base pairs, which is one of the main differences between species.

Of all biological systems (at least that I know of) this one accounts for the most bulk DNA sequence and is probably as related to the difference between different species as transcription factors and almost certainly is an older system of gene regulation, if you think about it.

  • Copy Number Variations and repetitive regions just a side note, but small and very long repeat sequences can show up in the intergenic regions as well as inside a gene boundary to account for some of the differences between individuals. they can be quite short or quite long.

Well hope this helps?

  • $\begingroup$ Just to be more accurate, about 1.5% of the human genome is coding DNA (Nature 409, 860-921 (15 February 2001)) $\endgroup$ – Gergana Vandova Jan 8 '12 at 22:59
  • 1
    $\begingroup$ wow - thanks for the bounty vote. I'd like to add that the mathematical Turing machine would involve using the DNA as memory and sequences over long distance affecting each other in a systematic way. chromatin sort of behaves like this, but the analogy is a stretch... $\endgroup$ – shigeta Jan 17 '12 at 1:47
  • $\begingroup$ just found this nice blog post with some hot links on 'Junk DNA' and how its being studied recently. phylogenomics.blogspot.com/2011/06/… $\endgroup$ – shigeta Feb 22 '12 at 17:35

Depends on what you mean by "non-coding".

There are structural elements in telomeres & centromeres -- although the DNA there does not code for proteins, it contributes to the three dimensional structure of the chromosome.

"Non-Coding" DNA can also act as a binding substrate for many proteins: transcription factors, enhancers, histone proteins; and thus control regulation indirectly through these intermediaries.

Promoter regions upstream of transcribed/translated regions are the combinatorial control switches/dials of our genome, and have tremendous regulatory importance.

Non-Coding DNA also acts as repertoires of mobile DNA elements, which enable fast evolution / "plasticity" by copying&pasting exons around (L1 transductions) or by copying into coding regions & interrupting them.

Lastly, they can act as the sandbox of evolution: Non-coding regions not genetically linked to functional regions are very likely to not suffer from purifying selection, so they can act as templates of random evolution - where the vast majority of mutations won't have any impact positive or negative. This enables exploring brand new combinations, which may then become new exons / miRNA / regulatory regions or get shuffled into other regions to enable new functionality.

  • 5
    $\begingroup$ You may wish to expand on the 'sandbox of evolution' point =) $\endgroup$ – Rory M Jan 2 '12 at 14:11
  • 1
    $\begingroup$ @Rory Good point, will do. $\endgroup$ – Deniz Jan 4 '12 at 11:11

By programmable, I suppose you mean that it contains information or can be altered in response to some input or stimulus. The answer is "no" for both. Well, sort of.

Does noncoding DNA contain information? By definition, no. There are probably many regions of the genome that appear to have no information, only later to be found to contain introns, regulatory elements such as enhancers, boundary element, MAR/SARs, targeting sites, etc. Even functional tests (such as removing the region) may not reveal anything because the effects could be minor, or only evident under special conditions. But arguably, if you remove a region and it has an effect on the organism, then it's not really a noncoding DNA, it's just you didn't see the coding before hand.

As for the latter, can it altered, the answer is again "no," or at least "apparently not." Intergenic regions (those stretches of DNA that do not contain obvious or characterized transcribed regions or their control elements) are very stable between organisms and even between species. They seem to have a mutation rate expected for having no information, and thus free to mutate slowly without being swept away. There is no evidence (as far as I know) of any region of the genome being purposefully altered, with the exception of a handful of specific genes whose regulation is controlled by DNA nicking or some such.

Perhaps I am missing your question, being a biologist and not really knowing what a "Turing Machine" is. If I misunderstood, please clarify.

  • $\begingroup$ I don't have a reference to give right now, but mutations in introns can, for instance, affect splicing. $\endgroup$ – nico Jan 2 '12 at 14:22
  • 1
    $\begingroup$ Nico, that's not necessary. I'm well-aware that mutations in introns can affect splicing (as well as expression should the gene have intronic enhancer elements). However, I was talking about mutations outside of transcribed regions. Part of the problem with the question is divining what he or she means by "noncoding DNA." $\endgroup$ – KAM Jan 2 '12 at 21:26
  • $\begingroup$ Sure, mine was just an example $\endgroup$ – nico Jan 2 '12 at 21:39
  • 1
    $\begingroup$ @Gergana Vandova, OK. I guess it's a difference of definition. If you define "coding" as open-reading-frames, then you're right. I look at "coding" as carrying useful information. Overall, though, I think it's a vague term. I've heard some describe rRNAs as coding, since they code for structural information. It's a semantic argument. $\endgroup$ – KAM Jan 9 '12 at 0:50
  • 1
    $\begingroup$ @KAM I think that this is the common definition of noncoding DNA. Tried to find a reference, but couldn't find any except wikipedia and some dictionaries. Would be nice if people don't spend time trying to infer what the asker meant.... $\endgroup$ – Gergana Vandova Jan 9 '12 at 1:30

I am very suprised nobody mentioned the field of DNA-computing. It is proofen by Leonard Adleman and Richard Lipton that you can compute with DNA molecules.

In the article of Adleman they present an experiment to solve an instance of the Traveling-Salesman-Problem. Because this problem is in NP one can say that the DNA is turing-complete.

Article of Adleman

For a deeper understanding see

  • 1
    $\begingroup$ -1: a) DNA can be used to compute, but the cell does not naturally do that. b) specific sequences of DNA/RNA are requested to do nucleic acid based computing, which are not necessarily present naturally, and/or not necessarily in ncDNA. c) the apparatus needed to do that is not only DNA based, specific enzymes, added in specific sequences are needed. $\endgroup$ – nico Jan 4 '12 at 11:01
  • $\begingroup$ I dont see the sentence in the question where your restrictions are mentioned. Can you help me? $\endgroup$ – peri4n Jan 4 '12 at 11:06
  • 1
    $\begingroup$ sorry, I did not notice that someone edited the question, I was still thinking about the first version of it, so my comment does not really apply anymore. However, the system does not allow me to remove the -1 now, please do some edit to your answer so I can change my vote. $\endgroup$ – nico Jan 4 '12 at 11:16

Nature has done pretty well in the subject of formal computation. So much that we are still trying to keep its pace.

As about your question, it depends on your definition of "non-coding DNA".

In general, DNA together with the machinery in charge of its maintenance is Turing-complete in several senses. Take a look, for example, to the existence of mobile genetic elements: some of them are "subprograms" coding for reverse transcriptases which in turn are able to reproduce the original program. I must note that for doing this with a Turing-complete formalism like the lambda-calculus, well, you need to do a fairly long and complicated program: http://crpit.com/confpapers/CRPITV26Larkin.pdf . And the lambda calculus is "easier" than bare Turing-machines, meaning that you can write shorter programs than with Turing machines for doing the same thing. So, my (somehow specious) argument is that any real-world information machine capable of self-replication is with high likelihood Turing-machine equivalent.

It just happens that the essential feature of mobile genetic elements is to ensure its survival, so, that's probably the reason we haven't found a fragment of DNA able to do something as interesting as calculating square root.

If you refer to the part of DNA that can not do anything at all, well, for talking about Turing-machines and computation you need a way in which some "data" can be "interpreted" as a program. A totally inert piece of DNA does not fill that role, by definition.

  • 1
    $\begingroup$ Unfortunately, it can be shown trivially that self-replication isn’t sufficient for being Turing complete (imagine the programming language “Rep” which has a single command, “rep”, which prints “rep”; clearly this language allows to write self-replicating programs and clearly it isn’t Turing complete). $\endgroup$ – Konrad Rudolph Jan 3 '12 at 22:49
  • $\begingroup$ @Rudolph: that's what I meant by "specious". So, it is not a formal proof, and you will need to do some more work to have one. However, it would surprise me that a machine more complex than anything we have invented (including our beloved computers) would be not Turing-complete. $\endgroup$ – dsign Jan 4 '12 at 8:58
  • 1
    $\begingroup$ The problem is not that the proof is specious, it’s that the proof is impossible since the assertion is provably wrong. I agree with the more specific point that cells in particular are Turing complete – but not all self-replicating information machines are. $\endgroup$ – Konrad Rudolph Jan 4 '12 at 10:50
  • $\begingroup$ @Rudolph: Agree, my mistake. $\endgroup$ – dsign Jan 5 '12 at 18:43

Let me answer your question by splitting it into two parts:

Can DNA be used as an programmable medium (=band) for Turing machine?

The answer is YES.

Starting from the breaking paper by Shapiro et al. in Nature, followed by another great article by Parker, there are many scientific publications about how to use DNA for computing. Unfortunately, these findings are still not applicable for classic computations and DNA computers will hardly substitute the normal ones in the nearest future.

Can a junk DNA work as Turing machine?

There is a known principle in Computer Science called "Junk In, Junk Out". Same in the case of DNA -- there is a way to use junk DNA for computing, but the result of the computation will mostly be junk too: as long as we don't know what this DNA is for, and this doesn't seem to operate as Turing machine on its own, there is hardly possible to get something reasonable by running this DNA on a turin machine...

  • $\begingroup$ "Junk" here has a specific meaning. And by the way, its usually written "Garbage in, garbage out" $\endgroup$ – John Smith Jan 5 '12 at 15:31
  • 1
    $\begingroup$ Thank you. I am aware of what "junk DNA" means, just wanted to play around with the word a little bit to make my point clearer. $\endgroup$ – Alexander Galkin Jan 5 '12 at 15:41
  • $\begingroup$ My question wasn't so much using a dna strip in place of the paper for a turing machine. That's not very interesting. I am asking if nature already has computation machines in a cell, which might help explain some of the non-coding DNA $\endgroup$ – John Smith Jan 5 '12 at 15:46
  • 1
    $\begingroup$ You understandably misunderstood the question because some moron has hijacked my question and changed it into something completely different. If you look at the comments under the question and the high rated answers you see they are discussing my original question, not the one you saw. I've tried to edit it back to what I asked. $\endgroup$ – John Smith Jan 5 '12 at 16:57
  • 1
    $\begingroup$ Indeed, I noticed the mismatch between the question and the comments, but I thought that it is because the question was improved in reply to comments. Now it is much more clear, thank you for your efforts! $\endgroup$ – Alexander Galkin Jan 5 '12 at 17:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.