How about another of those non-awkward Dawkins Twitter questions? Although this one actually is kind of awkward, in a non-offensive way. I don’t quite know what it means.
Does evolution rely upon digital genetics? Could there be an analogue genetics? What features of life have to be true all over the universe?
I don’t understand what ‘digital genetics’ is…it sounds very Mendelian, I think. There is a reasonable but overly simplistic perspective on genetics, a very old school way of thinking, that mutations are binary — you’ve either got a trait or you don’t. We also use a linear sequence of nucleotides that get translated into a linear sequence of amino acids, and it all looks very much like a digital computation…and can be modeled reasonably well with a computer. So in a very narrow sense, I suppose you could argue that earthly genetics is ‘digital’.
But it’s also analog. It’s all chemistry, so reaction kinetics and concentrations and equilibria drive everything. I also think it’s a flaw of that old-school classical genetics that subtle quantitative variations are difficult to detect — people were always looking for mutations with distinctive phenotypic effects. Look at the Drosophila eye mutations, for instance: white was discovered first. Wild type flies have brick red eyes, while the mutant has obvious white eyes. A single gene can flick the color between red and white.
But here’s a qualifier: it’s not simply binary, a choice between white and wildtype. FlyBase tells me there are 1060 published, classical alleles of white. The gene behind it is an ATP-dependent transporter protein — so it’s an enzyme, with analog chemical properties.
Further, when I think of all the other things that can modulate genes, it’s hard to retain an image of them being digital. Codon usage in the gene can have quantitative effects on rates of translation. Position effects — what genes or other DNA are in the neighborhood of the gene in question — can modify the regulation of the gene. The number of copies of a gene can have effects — there are lots of experiments in gene dosage in flies, and we know that humans have copy number variants.
We also know that many genes generate phenotypic effects with concentration gradients, not digital at all. In flies, bicoid is the canonical morphogen — its RNA product is concentrated at the head end of the embryo, and its protein diffuses away to form a gradient from high at the head to low at the posterior, and this gradient is read by other genes, like orthodenticle, hunchback, and knirps, to determine whether they should be switched on or off. We can push around boundaries of anatomical features in the larvae by changing the concentration of bicoid.
To make it all fuzzier still, that gradient is also a product of other cooperating genes, like stauffen and valois, that localize bicoid.
So I guess my answer to that question is that we already know what an analog genetics would look like — we just have to look at ourselves.
I’m guessing he means that instead of being limited to A, C, G or T in any given location, what would happen if you had bases that were 47.2% A and 56.8% G? Some kind of infinitely divisible scheme?
Or, you know, with the percentages adding to 100%.
I wonder if he may have been referring to the 4-base code nature of DNA. Could life exist without an organized central repository of information, and instead remain a series of loosely conglomerated self-replicating molecules working in concert?
Forgive my ignorance, but it looks to me like you are talking about analog phenotypes, not analog genetics.
Yeah, Kauffman’s theoretical ‘autocatalytic set’ of self-propagating molecules, which probably is what the very earliest life was like.
You can’t really separate phenotype and genotype that cleanly. My point is that the transmission of genetic characters is dependent on analog properties of the cell.
Yeah, you guys are onto it – Dawkins claimed in one of his books that DNA is a base 4 digital code because you can’t have anything but A, T, C, or G, just like in computers you can’t have anything but 0 or 1, with nothing in between. If I recall correctly his point was that it is the only way replication fidelity is so high, in the same way copying an analog sound recording develops hiss with each generation but copying a digital recording does not.
Does a digital computer somehow become an analog computer, just because it runs a very large and complex program, that produces a wide range of quasi-continuous outputs? A string of bits stored in an electronic circuit may get operated on a widely variable number of times, but they are still bits.
Then the question isn’t digital vs. analog — it’s a matter of fidelity. And again, we also have varying degrees of fidelity of replication in life right here on earth, so the question is one of where we can draw a line on a quantitative metric and say, fidelity below this level is intolerable, above this line it is permissive.
I also think comparisons of the genome with a digital computer are forced and inaccurate.
Again I think Dawkins was trying to draw that line, or state that the line was drawn by selection long ago (origin of life timeframe) when there was probably much greater variation in copying fidelity (and fecundity) by replicator molecules. DNA or maybe RNA originally won out over clonkier systems. Now that everything works on DNA the range of variation in fidelity is far less than it used to be.
Yes the analogy with computers only works so far, but I think it can be enlightening at a basic level.
Makes sense. A loose collection probably wouldn’t copy successful versions of itself as accurately as genetic life would, and would be put out of competition pretty quickly. About the only way that could continue is if you were using a different chemical basis for life which didn’t support a DNA analog (i.e. hydrocarbon or silicon-based rather than carbohydrate based.) Of course, at that point, it’s speculation Jenga.
PZ @ #9
Agreed. No computer operates with an amount of nonsense junk code that is even comparable to what is seen in eukaryotes. Further, computers run what is coded and nothing more, they are not subject to modifications of the manner in which the code is read (or not read) by external conditions; that is, there is no such thing as computer epigenetics.
A better distinction might be “discrete” and “continuous,” so as to avoid all of the baggage that accompanies “analog” and “digital.” Dawkins’s point has always been that for selection to act, it has to act on something with sufficient copy fidelity, the mutations of which are pointwise and relatively conservative. If genetic information were stored in a continuous medium there wouldn’t be genes. Imagine if somehow something like the concentration gradients were the source of genetic information rather than an effect. According to Dawkins, selection would not produce very much variety in such a system, both because of fidelity issues and because the locus of selection would be too broad to preserve any discrete part of it for more than a couple of generations. Future organisms couldn’t build on what already worked.
“Then the question isn’t digital vs. analog — it’s a matter of fidelity.”
Actually, from a coding perspective, it is digital vs analog. The coding involves discrete values A,C,G, & T, rather than a continuous spectrum of possible values ranging from one end to another.
Fidelity is probably tangent to the point. You also have varying degrees of fidelity in a digital computer. Though fidelity is generally very high, there are reasons why things like ECC (error checking and correcting) exist; erors still happen in the digital world.
If the coding is digital, can the output really be analog, or would the output only appear to be analog due to the very large number of possible outcomes?
Consider that the image you see on your computer screen when viewing a photo is all digitally encoded with a finite number of pixels, each with discrete value for color and brightness. There is an insanely large (seemingly infinite), but discrete and finite number of images that can be possibly displayed on any computer screen.
Note to any who think digital = binary: Not so, digital only requires discrete values; even though binary is the most commonly encountered form of digital. Digital clocks weren’t originally called digital because they used digital electronics to keep time. They were digital because they displayed the time in discrete values unlike analog (hands) clocks. In a digital clock, it’s either 12:00 or 12:01; the minute hand is never somewhere between the 12 and the 1.
Until you unpack the term “genetics” into its component processes, discussion will be difficult.
Replication, transcription, and translation are digital processes with discrete values and low error rates. Folding, translocation, transport, regulation of those processes, and expression of characters are analog with continuous values.
Chances are that the original questioner wanted to know if analog substitutes for replication, transcription, and translation could work.
Yes, the output can be analog, depending on what you count as “input and output.” Imagine a computer that has been set up to control the heating element of an oven. The signal sent to the element may only have discrete values, but the way the element responds at the moment depends on a lot more than what signal the computer is sending it. Likewise, the instantaneous temperature of the oven at any given time after the element has received its signal depends on a lot more than just the heating element. One free to call the “everything else” in the story part of the “input,” or conversely free to deny that the oven temperature counts as “output.” The scope of reference counts for a lot in these discussions.
karlwithakay @14:
What is the continuous spectrum in this case? What do the continuous values correspond to chemically?
Accepting the premise that genetics is all digitally encoded, in the quaternary format (A.C.T.G); the question is, “How could this also be done in an analog (eqv: “analog voltage”) system?” AND “what features of life would be common ANYWHERE in the universe?” {Omniscient Dawkins, you.}
.
Avoiding the second, unreasonable question; Are there any “analog” features of genetics that could be affected by evolution; as in increasing/decreasing (marginally) its expression? This is a different question than the answer presenting the analog effect of multiple genes existing in the organism. This question asks if there is a single feature of the genome that says (explicitly) how much of an effect it will have. EG Is there a place in the genome that is essentially a 1.3 gallon bucket vs a 2.7 gallon bucket, with the volume of the bucket determining the colorshade of the fur.
And evolution could tweek the volume; one has 1.4 gal, one has 1.27 gal, etc.
@17:
We don’t know what the continuous spectrum would be because it’s hypothetical. Again, scope matters — even a concentration gradient isn’t continuous on the scale where you can (in principle) count individual molecules.
@12
I haven’t programmed professionally, but my programs are littered with comments, lines of code that have been commented out (some of which never worked; some of which used to work when the adjacent code was different; some just being alternative implementations), and pre-processor instructions. Beings I’m under no compulsion to clean up my code, it only has to be “good enough”. If the compiler wasn’t around to remove all of this junk for me prior to program creation, it would all be there in the final program. That wouldn’t deter the program from working correctly, but it would add unnecessary memory requirements.
DNA doesn’t have an editing function that removes non-functioning nucleotides when sections are added, so it’s more like the original code that the programmer worked with. It’s only in the mRNA that we see the cleaned-up, short version. So genetics would be more akin to a computer system that has to compile the program every time it runs something. Still not perfect, but more accurate.
@13
I think we should stick to fidelity. The practical line between analog & digital is blurring, apart from the fact that analog and digital media have each improved dramatically over the last hundred years. So why couldn’t chemistry undergo the same selection process over a longer timeline, albeit unguided?
Dawkins has answered this before and, as others have mentioned here, he argues that a analog could not have the fidelity of a digital system.
From River Out of Eden,
I’ll let more knowledgeable people debate whether that argument works.
BTW,the question of whether reality itself is analog or digital doesn’t have the simple answer you might expect. Scientific American asked the question expecting the overwhelming response to be digital. After all, the quantum of quantum mechanics seems to imply a discrete reality. However, more argued for the analog side than they expected. Those on the analog side rightly pointed out that in quantum mechanics continuous fields are the fundamental object and that discrete quantities arise only after those fields are put under constraints (e.g, quantum numbers arise when an electron is bound to an atom; the states of a free electron are continuous).
“I haven’t programmed professionally, but my programs are littered with comments, lines of code that have been commented out (some of which never worked; some of which used to work when the adjacent code was different; some just being alternative implementations), and pre-processor instructions.”
But none of that gets to the C.P.U. at run time. It’s removed either at compile time or just before run time in the case of interpreted languages. If you decompile an .exe file you don’t get your comments back.
I’ve used the digital/analog comparison a lot when teaching. Digital genetics refers the fact that genes are discrete units and for any locus there is a finite number of alleles that are segregating.Analog genetics refers to blending inheritance.
The fundamental difference between the two is that with digital genetics children pass on the units of heredity they inherit from their parents, maybe with mutation. With analog genetics, children would pass on the average of the units they inherited.
The significance is that evolution by natural selection doesn’t work under blending inheritance because variation is destroyed too quickly.
Right. But we’d have to assume that sexual reproduction could evolve under analog genetics.
firstapproximation @21:
Continuous fields and free plane wave states are fundamental objects in our theories, because we start with (continuous) classical theories. That doesn’t mean they’re fundamental in reality.
This paragraph from David Tong’s essay says it all, really. At least the first sentence isn’t silly.
Yech.
Another Matt @24:
Continuous reproduction spectrum? Dating could get complicated.
@Rob Grigjanis #25:
True, I suppose. It’s hard enough to get our macroscale heads around what might actually be happening at the quantum level. I’m not au fait with the maths, but I expect that even continuous fields (or their representation, however useful a tool that might be) must break down at the Planck scale.
@22
And if you reverse transcribe mRNA to DNA, you don’t get your introns & promoters back.
Maybe non-functional DNA is a little more like variables in a program that have been declared and defined but then never used in functions or expressions. Some compilers could cut them out, but if they are compiled in they’ll still take up memory.
Always curious @ #20
While that would appear to be a tempting analogy, it lacks an understanding of *how* bits of the genome become nonfunctional. When you “comment out” bad code, you are doing it deliberately and selectively, and you are only doing it to bad code. Evolutionary genomic change does not work like that, the process is far and away less controlled, and involves factors beyond selection (which could be analogised to a “programmer” though that does take us uncomfortably close to ID).
Rob Grigjanis,
Considering that we don’t know exactly how reality works then we have to go on the basis of our best theories. Other theories have been proposed, some of which could be categorized as ‘discrete’, but none have been as successful as quantum field theory.
_ _ _
Rich Woods,
Well, I guess you can argue that under the modern understanding of renormalization our theories are viewed as approximations. These approximations work whether the “true” theory contains a discrete or continuous space-time, so we should remain agnostic to the question.
_ _ _
reedcartwright,
I remember Dawkins mentioning this too, but can’t remember where.
This rabbit-hole goes very deep. As PZ pointed out, even in cases of single yes/no phenotypes where a single gene codes for either one or the other trait, there is still a complex, analog molecular system under it, subject to physics, dynamics, etc.
…and under THAT is a very “digital”-seeming, quantized spacetime with a pixel pitch of the Planck length and a clock tick of the Planck Time :)
Maybe the answer to “digital or analog?” is “what scale are we talking about here?”
Rich Woods @27:
The infinities encountered in quantum field theory are a sort of break-down. That’s not a problem computation-wise at accessible energies (and maybe a lot further), but it tells us that we don’t understand really high energy (equivalently, very short distance) physics. So until something better comes along, we’re stuck with effective low energy theories, for ever increasing values of ‘low’. As far as I can see, ‘analog or digital’ is hidden from us, and the choice is down to aesthetics.
@9 PZ Myers (if I may speak freely oh my king) *
Regarding the digital computer, comparisons of anything (with the possible exceptions of paperweight, lightbulb, rabbithole, post-it and calendar) are not only forced and inaccurate at best, but are more likely lazy, gullible and pointless.
Yeah… the machine does things kinda similar to various other tools / vectors / organisms, and does some things better (for some value of better), but mutates them so much in the process that any possible comparisons become little more than sources of confusion and obfuscation.
* Parenthetical included as a viral gift to visiting adjudicators of the hive mind.
@30
I’m not arguing for a Cosmic Programmer. Maybe I’m wrong, but the analogy of DNA:computer program is NOT supposed to explain how the system got to be the way it is. Rather it’s about how the system presently works (on non-evolutionary time scales).
My point in #20 was more about how there is much more junk scattered throughout the original instructions (DNA or code) than existed at “run time” {however that is defined}. That some of that includes proper/nearly proper instructions is interesting for understanding the history of that section, but is largely incidental otherwise.
At the present time, computer programs (post-compile) are swelling in size relative to the actual amount of code necessary to perform the functions they actually execute. And why might that be? One reason seems to be the necessity to have a complex program run properly in a variety of different operating environments. Automatic updates & such are helpful, but sometimes an update to one program crashes another program. Maybe we can gain some efficiency by thinking outside the box a little further…
Ultimately, my interest here isn’t to say that they are exactly the same. But exploring the similarities will be valuable both to computational science and to biology. Are there alternative computational methods that are faster/better for modelling biological systems? How might a program rewrite itself without destabilizing? Might there be a way to generate a more compact program that will adapt to its operating system & its users’ habits? The answers might lie somewhere in, or be inspired by, the field of genetics. And geneticists certainly stand to gain if better computers can help them out.
To me ‘analog’ vs ‘digital’ refers to the mechanism of information storage. So storing data as a sequence of discrete quantities (1s and 0s) is digital, while storing data on a continuous medium (like a vinyl record or magnetic tape) is analog. DNA stores information digitally.
I can imagine life based on the vinyl record or magnetic tape analogies:
1. Vinyl information gets copied by pressing a pattern into malleable media. By analogy you might imagine some king of analog life where, for example, the organism falls into clay at the bottom of a lake and makes an impression, and decomposes away. Then sediment gets deposited in the impression. The lake dries out and the sediment dries into a new ‘copy’ of the organism. When the water returns, this organism is washed out and lands in a new spot. Now there are 2 impressions of the organism, and the cycle can repeat. And certain impressions would be more likely to lead to further impressions than others.. for example by causing the organism to ‘pop’ out when it dries, the way certain grasses ‘explode’ their seeds when they dry.
1. Perhaps there is some physical system in which magnetic ‘ribbons’ appear, perhaps inside neutron stars or something like that. One magnetic ribbon would be ‘copied’ to another by simply lining them up – the local magnetic moments would align. However, there are problems with signal loss – the final magnetization would be the average of the two strips, and you’d need a way to amplify only one of them.
,
The core of the question revolves around distinctions between discrete and continuous values. DNA can be described in a discrete representation: namely, one of a limited set of nucleotides (a discrete state), each assigned an ordinal position in a chain (also a discrete state).
In the current standard model of physics, continuous coordinates are used. Time/space coordinates, and the various scalar field value are modeled in continuous, not discrete coordinates. Perhaps that might change in future, but in the meantime, the standard model of physics is continuous, not discrete.
A plausible example of a genetic system with continuous values: one where spatial configuration of a DNA strand or protein partially or completely encodes the genome. I was under the impression that there is some evidence to support the idea that some genetic information is transmitted via spatial configuration.
Well, 1060 is a surprisingly big number to me, but, what does that mean? Is allele #1060 only different from the other 1059 because it has at least 1 B.P. difference, but it behaves exactly the same , i.e., the expression of each allele results in an identical expression in the phenotype? Or, are there actually 1060 different phenotype expressions that result in white eyes? Or is it a combination of both – maybe there are twenty different expressions and dozens of the alleles each produce the same expression?