How much of us is just... random?
January 19, 2020 8:03 AM   Subscribe

There are genes, of course. And environment. And gene-environment interactions. But inside of us there are also molecules bouncing around randomly, sloppy transcription, retrotransposons escaping their jails, and not nearly enough DNA to specify all of our 100 trillion (or 1 quadrillion?) synapses. Nature versus nurture misses the importance of noise.

Let's take a peek at the building of connections in the brain during development to find some of the limits of brain determinacy (PDF) even in isogenic (virtually genetically identical, typically via multigenerational inbreeding or cloning) animals:
The growth of neurites (axons and dendrites) is not random; they seek particular targets. But stochasticity is fundamental to the target-seeking, which involves semi-random events and subsequent selection.

The advancing end of the axon (or the dendrite) is called its growth cone. This is a highly active structure from which finger-like actin-filled extensions called filopodia are constantly growing out and often retracting as they test out different possible pathways on a trial-and-error basis. Just as a traveller can respond to global cues such as the direction of the sun or a compass bearing, and to local cues such as ‘turn left after the house with red shutters,’ so do axons.

It is generally accepted that the above-described processes include a random element. The initial outgrowth of individual filopodia to sample their environment is thought to be essentially random, the growth being directed by the selective stabilization of filopodia that receive positive signals. Even this stabilization has a stochastic component.
Because there's not enough DNA to specify all the synapses (connections between axons and dendrites) our brains need, semi-random noise is used to create a profusion of them during periods of exuberant connection (PDF). After that, wandering axons which have ended up in the wrong neighbourhood (and the neurons they're connected to) are pruned, restoring order to the brain.

It is difficult to prove which brain differences in humans are the result of environment and which are the result of internal noise, since even identical twins have different prenatal experiences. But in less complex animals,
where environmental factors can be better controlled, neuroanatomical differences still occur. For example, in isogenic grasshoppers, there is variability in the positions of neurons and in the branching patterns of their dendrites - as much, in fact, as in heterogenic clutches.
Does this noise-generated variability matter, or is it rendered moot by "noise-tolerant strategies of self-organization... that render moderate noise simply irrelevant to the global result? ... The importance of these noise phenomena for brain functioning is difficult to evaluate."

It's not only brains which are affected by internal noise.

The nematode worm C. elegans, a favourite of researchers, develops dependably. Every hermaphrodite creates exactly 1090 cells, always in the same order; exactly 131 of those cells die, leaving the adult with 959 cells. The hermaphrodites can be self-mated to create isogenic lineages. They can be raised in virtually identical laboratory conditions, with liquid culture to ensure there are no nutritional differences and precise temperature control. Even so - even with genes and environment virtually identical - some of them live three times as long as others. In isogenic mice, some females produces three times as many eggs as others. These, alongside many other examples, led Finch and Kirkwood to argue in Chance, Development, and Aging that "intrinsic chance" - internal noise - plays as large a role in some biological processes as genes or environment.

Internal noise can be - and often is - well-controlled via canalization (the ability to create the same phenotype regardless of variability in genes or environment) or tightly controlled reaction norms (the way that the genes change the phenotype in response to different environments). Bilateral animals, for example, are triumphs of symmetry, symmetry which would be destroyed if internal noise played too large a role. The cotton aphid, though, since it mostly floats on the wind instead of flying properly, experiences relaxed selection for wing symmetry. As a result, it has four times as much wing asymmetry as other insects, with half of that asymmetry explained by internal noise during development.

Where does the noise come from?

In The causes of evolvability and their evolution (PDF), Payne and Wagner review some of the non-genetic sources of internal noise and discuss its evolutionary importance.

Gene expression noise: Sometimes only a few molecules in a cell control the expression of a gene, so its output jumps up and down semi-randomly as their numbers change. Sometimes conflicting molecules fight over whether to turn a gene on or off. A recent study in the plant Arabidopsis thaliana found that small, short genes with large numbers of transcription factors controlling their expression were the noisiest.

Errors in protein synthesis: Our cells keep tight control over DNA synthesis, only allowing one mistake per billion base pairs copied. They are positively sloppy, though, when creating proteins from that DNA (sometimes sloppy in surprisingly useful ways) with "nucleotide misincorporation during transcription, tRNA misacylation during translation and kinetic trapping during protein folding. Translation is particularly prone to error, with rates of mistranslation exceeding those of DNA point mutations by several orders of magnitude."

Epigenetic modifications: Sections of DNA can be closed tight or specially marked to quiet their expression. Most epigenetic changes are programmed events in multicellular organisms (like us) which make one kind of cell different from another kind. Some epigenetic changes appear to occur randomly, throwing a monkey wrench into the engine of nature versus nurture and helping to create some of the differences between identical twins.
Each mechanism can create phenotypic variation in a population of genetically identical individuals. Such variation can, for example, provide a competitive advantage to subpopulations with adaptive phenotypes in fluctuating environments. These phenotypes may themselves be heritable, eventually made permanent by mutation or epigenetic modification, or they may simply ‘buy time’ for a population to adapt in other ways to an environmental challenge.
There may be more exotic sources of internal noise.

One research group says that they found 80 to 800 (presumably random) copy-and-paste retrotransposon duplications per neuron, suggesting that mobile DNA elements may have a role to play in the generation of diversity and complexity in the brain; another group says it's less than 0.6 duplications per neuron, hardly worth getting excited about.

Finally, we are only beginning to understand interactions between our brains and our immune system - but that might be an FPP for a different day.
posted by clawsoon (31 comments total) 91 users marked this as a favorite
 
Given that we are evolved to live in hunter-gatherer clans of 60 to 120 members, and occupy niches within the clan, it should be obvious that we're evolved to host stochastic processes within our own bodies to decide which of us occupy which niches.
posted by ocschwar at 8:18 AM on January 19


Nature versus Nurture versus Noisure.

(A butterfly in the Amazon made me write that)
posted by dances_with_sneetches at 8:44 AM on January 19 [8 favorites]


Stochasticity is key to my research in biology, and I write a few words about its importance in almost every article I publish.

My advisor had a pet peeve against calling it ‘noise’. He felt that framing was based on engineering and physics traditions, where noise is ancillary and problematic, something to be eradicated if possible. Basically it has negative connotations to many people, in casual language too.

Stochasticity doesn’t have that baggage, if only because it’s a much lesser known term, and the more we can teach folks at all levels that it is a central and key process of life, the better.

I’m not criticizing the use of ‘noise’ here in particular, and I’m not nearly as bugged by it as he is, but it think it’s an interesting example of how subtle word choice can change how people view things.
posted by SaltySalticid at 9:05 AM on January 19 [48 favorites]


There’s also some problems with ‘random’, insofar as that can fairly be applied to deterministic chaotic systems, which is part of why we started to use ‘stochastic’ in the first place but that’s a whole ‘nother rant :)
posted by SaltySalticid at 9:25 AM on January 19 [4 favorites]


> it should be obvious that we're evolved to host stochastic processes within our own bodies to decide which of us occupy which niches.

Please understand that I am asking this entirely sincerely and in good faith: Given the lengthy framing of this post in order to characterize the question of noise as a subtle and complicated topic, what about the above statement is "obvious?"
posted by Sokka shot first at 9:25 AM on January 19 [14 favorites]


I'm reminded of regeneration techniques in electrical engineering which keep the signal processing system right on the edge of run-away noise. To make them more sensitive and efficient. More used in the early days when they were struggling with anything to get them to work. Easier ways nowadays.
posted by aleph at 9:54 AM on January 19


Wow, I'm reminded of my college days in the early nineties and studying what was known about human brain development at that time in relationship to notions of computer 'agents' and evolutionary algorithms in CS. The internal metaphor I developed to think about it was a sort of 'action-reward' jungle that the developing brain was exploring based on some unknown blend of instinct and chance and using to reinforce or prune decision paths. It was also pretty speculative and conceptual.

It's amazing to see how much more progress and grounding these fields have today. Thanks for sharing!
posted by meinvt at 10:26 AM on January 19 [1 favorite]


meinvt, you might enjoy this bit from the "surprisingly useful ways" link (The evolutionary consequences of erroneous protein synthesis):
Certain picornaviruses carry, within a long polypeptide, a short sequence (~19 amino-acids) that induces eukaryotic ribosomes to skip a peptide bond. This skip-inducing 2A sequence allows these viruses to encode multiple proteins using a single, compact sequence without paying the price of encoding a protease. Such ribosome skipping, in essence a bug in the translational hardware uncovered and exploited by viruses, is now being coopted by human biological engineers.
posted by clawsoon at 10:57 AM on January 19 [6 favorites]


...obvious that we're evolved to host stochastic processes...

Isn’t there a need for some pretty sophisticated explanation about how evolutionary pressures can favour individuals whose behaviour is stochastically determined over individuals whose behaviour is genetically determined?
posted by Segundus at 11:04 AM on January 19 [1 favorite]


Well, for example, homosexuality. If it's genetically determined, there's an obvious selective pressure against it. If it's stochastically determined, you have a small percentage of the clan that will be caring for the clan's orphans instead of having children of their own, ergo selective pressure to sustain it.
posted by ocschwar at 12:07 PM on January 19


Doc said my cancer was caused by random mutations. Surprise!
posted by The Underpants Monster at 12:49 PM on January 19 [2 favorites]


Given that we are evolved to live in hunter-gatherer clans of 60 to 120 members, and occupy niches within the clan, it should be obvious that we're evolved to host stochastic processes within our own bodies to decide which of us occupy which niches.

This is ... not an accurate use of "stochastic processes"; initial assumption ("given that...") needs citation; and there's a huge logical gap between stochastic differences in individual anatomy and complex behavioral phenomena in groups of humans (which is itself an emergent property that can be subject to other, not necessarily related randomness or stochasticity).

Well, for example, homosexuality. If it's genetically determined, there's an obvious selective pressure against it.

Erm, also no. The second half of this comments posits one possible selective pressure toward homosexuality - a sort of grandmother effect. I've seen some claims that rates of homosexuality in animals increases with population pressure, as well, though I don't know if there's any valid science behind that or not. Regardless of such hypotheses, it is not at all clear or obvious that having some percentage of a population (of non-human animals who are actually subject to natural selection, let's say) be (exclusively - since many animals display bisexual behavior as well) homosexual is a selective disadvantage. So long as enough individuals reproduce to maintain the population, it may be an entirely neutral feature.
posted by eviemath at 12:56 PM on January 19 [12 favorites]


Don't call it noise... call it Rock Music!
posted by oneswellfoop at 2:32 PM on January 19 [2 favorites]


In baboon troops, the dominant male does most of the procreating. Young non-dominant males tend to leave their troop as adolescents and either find a new troop or not survive. Because of the one-to-many sexual relationship between the procreating male and the females, homosexuality would be a sensible response. I've read a couple books, not an expert.

why is it called noise and not randomness? ya, I kno, RTFA.
posted by theora55 at 2:34 PM on January 19


Noise is an engineering term for the randomness that interferes with its opposite, signal.

There's a bit of teleology involved in deciding what is signal, and therefore noise. Signal is the information you are trying to discern. But that depends on what you are attending to and hence the teleology. The purpose of a telecommunications signal, for instance, is to communicate certain information.

Stochasticity is an unwieldy word, but it isn't dependent upon any a-priori assumption about what part of what is observed is meaningful or not.

Randomness is an ok word, though no less unwieldy. But the implication is more toward the "meaningless" side rather than that meaning can be adduced from the data somehow, despite the random elements within it.

But FWIW, dictionaries rate them as equivalent. My take is that people want to avoid the connotations of "random" as being unintelligible.
posted by sjswitzer at 2:57 PM on January 19 [8 favorites]


Well, for example, homosexuality

Hoo,boy! Genetic determination of caste is a thing in social insects, fyi
posted by SaltySalticid at 3:25 PM on January 19 [2 favorites]


I debated what word to use for the central idea, since every paper seemed to have a different word/phrase for it. Internal or intrinsic? Developmental or not? Noise or chance or randomness or stochasticity?

I went with "internal noise" because it seemed like the simplest way to express both parts of the central idea, but I appreciate the comments about the subtler implications of each word choice, and I might've gone with something different if I did it again.

After I finished the post I discovered Kevin Mitchell, who works in developmental neurobiology and genetics and has written extensively on the subject. His book Innate says this:
But this is only half the story. Genetic variation is only one source of differences in how our brains get wired. The processes of development themselves introduce another crucial source of variation—one that is often overlooked. The genome does not encode a person. It encodes a program to make a human being. That potential can only be realized through the processes of development. Those processes of development are noisy, in engineering terms. They display significant levels of randomness, at a molecular level. This creates strong limits on how precisely the outcome can be controlled.

Thus, even if the genetic instructions are identical between two people, the outcome will still differ. Just as the faces of identical twins differ somewhat, so does the physical structure of their brains, especially at the cellular level.
He has a number of related blog posts, not all of which I've had a chance to read yet:

Nature, nurture and noise
Noisy genes and the limits of genetic determinism
Robustness and fragility in neural development

And some posts focused more on genetics and the brain which touch on noise stochasticity tangentially:

Somatic mutations make twins’ brains less identical
The genetics of emergent phenotypes
Are human brains especially fragile?
Gay genes? Yeah, but no, well kind of… but, so what?
posted by clawsoon at 4:00 PM on January 19 [9 favorites]


I am not so smart, but what a rich and fascinating discussion. Nice post!
posted by j_curiouser at 4:25 PM on January 19 [2 favorites]


I've sometimes claimed that there is nothing in the genes that encodes the shape of the liver. Sure, its topology and lobes are determined by genes but not the final shape. It simply grows into the space that's left over from the stomach and other viscera. It simply fills up the space that's left.

Clawsoon put it better above: genes are a program, not a blueprint. But further, that program is full of hacks and bugs and patches and a lot of nonsense too. And the program's execution is in an environment that is physically and chemically constrained. All of these things come together dynamically over time, and one would be hard pressed to find any meaning or purpose in the genes (indeed, you should reject teleology altogether). But nevertheless, it works.... well enough.
posted by sjswitzer at 4:29 PM on January 19 [2 favorites]


sjswitzer: It simply fills up the space that's left.

As it happens, liver size is tightly globally regulated; if you cut a chunk out, it'll grow back very close to its original size. If you put five livers in an animal, they'll grow to a fifth the size of a normal liver, giving a total size equal to what a single liver would've been.

Other organs, though - and I apologize for not having the review article I'm thinking of at hand to remind myself of which organ has been used for this in experiments - have intrinsic developmental programs. Attach five of them to an animal, and you'll get five full-sized organs.

As usual in biology, the answer is, "Why not both?"

As for shape... who knows? You may well be right.
posted by clawsoon at 4:43 PM on January 19 [5 favorites]


(Oops... it's the thymus and spleen I was thinking of. Liver size is tightly regulated, but I'm not sure if anybody has tried putting five of them in an animal.)
posted by clawsoon at 4:47 PM on January 19 [6 favorites]


What I'm suggesting is that it's regulated by the space that's available. But anyway that's just my dingbat theory.

But what I'm getting at here is that the fact that all of the pieces of our bodies fit together perfectly as an elaborate jigsaw puzzle is not due to the fact that each of these pieces has a blueprint in the genes, but rather that in the course of development physical constraints (among others) limit their growth so that no matter how well or how badly those pieces turn out, they will inevitably fit snugly in 3D space. The genes did need to encode that information, but rather their developmental programs evolved to use that to achieve the shapes they arrived at.
posted by sjswitzer at 5:05 PM on January 19 [2 favorites]


Well, for example, homosexuality. If it's genetically determined, there's an obvious selective pressure against it.

Hey there. This is a classic homophobic argument for the unnaturalness of homosexuality that is also completely incorrect. There have been studies that have found that (in the specific populations sampled) that homosexual men have more children on average that heterosexual men. In other words, there is not an obvious selective pressure against homosexuality: there is an obvious selective pressure for it, since it can increase reproductive fitness.

You assume that homosexuals don't have children, and you are wrong.

If you're going to spew homophobic propaganda, you should at least have good data.
posted by medusa at 5:39 PM on January 19 [12 favorites]


The distinction between programs and blueprints is pretty fascinating to me. You see it in the distinction between imperative and declarative styles in programming. In the imperative style, there's no limit to what can be expressed (within the universe of discourse), but you run into the halting problem: in principle, it's impossible (not even just hard) to say what a given program will do without running it. In the declarative style, a language is designed specifically so that its results can be read more or less directly from its source. This is great for things like GUI builders where you want a WYSIWYG development environment. But there will be corners you can never reach with this approach.

Kolmogorov compression is the idea that the information content of a thing is the size of the smallest program that can produce it. Well, genes are programs for a weird biochemical computer. And while evolution doesn't necessarily select for the most efficient encoding of programs, it does seem to exploit the fact that "grow this far, rotate 15 degrees, then branch by 30 degrees and repeat" is a more efficient way to encode a tree than anything like a blueprint.

Similar ideas exist in technology. The MPEG-4 standard, for instance, does not specify how to encode a video! It only specifies how to decode one. The bitstream itself is something like a program to execute on the player. Just about the only way to know what an MPEG-4 stream represents visually is to run the program. And while there's a canonical way to decode a stream (as with DNA, FWIW) there's no canonical way to encode it. That is up to the ingenuity of people designing codecs, or the amount of computing power you want to put into the encoding.

All of this is to say that genes are pretty clever because they "discovered" that it's more efficient to encode a program of "how to" rather than a blueprint of "what." And it's a good thing too, because the other idea would never have worked.
posted by sjswitzer at 5:39 PM on January 19 [11 favorites]


Well, for example, homosexuality. If it's genetically determined, there's an obvious selective pressure against it. If it's stochastically determined, you have a small percentage of the clan that will be caring for the clan's orphans instead of having children of their own, ergo selective pressure to sustain it.

This sort of just so story approach when you see a trait is a really bad habit. It is a fallacy of believing that every trait is an adaptation. You don't need a story for every trait. Biology is messy; a lot of things that happen at low frequencies are just random and the selective pressure that preserves "unfavorable" phenotypes (in terms of numbers of offspring) often has nothing to do with the phenotype itself. Evolution is a amazing but it can't do miracles; it does kludge fixes for problems as they come up, which have other effects elsewhere.

Applying the kind of logic you use here is (for example) like looking for an explanation of sickle cell anemia in terms of a low percentage of sick people being beneficial for community building within a group. Or some other handwavy thing. But that's not the trait being selected for: It's malaria resistance, and sickle cell anemia is a random byproduct of that in a subpopulation.
posted by mark k at 8:58 PM on January 19 [10 favorites]


sjswitzer: The distinction between programs and blueprints is pretty fascinating to me. You see it in the distinction between imperative and declarative styles in programming.

Now I want someone to create a programming language which adds a bit of stochasticity to every function return in order to simulate life-like conditions. (Probably best if there's some spontaneous bit-flipping in memory, too, and no distinction between data and code.)

How successful would we be at programming in that language? How successful would evolutionary algorithms be?
posted by clawsoon at 4:16 AM on January 20 [2 favorites]


ConwZy's Game of StBchastic Life
posted by clawsoon at 4:41 AM on January 20 [3 favorites]


How successful would evolutionary algorithms be?

They vary, though haven't found as much application as was initially hoped for.
posted by eviemath at 10:13 AM on January 20


eviemath: They vary, though haven't found as much application as was initially hoped for.

Would you happen to know if anyone has played with them on top of a stochastic processing engine?
posted by clawsoon at 10:52 AM on January 20




"Random" - so the cartoon of the butterfly, surrounded by spring buds, on the Great Wall of China, madly flapping wings, and explaining, "Those New Yorkers are in for a hell of a winter"

In other words, the actions of the individual DO matter. And some of the "manners", "stories", "superstitions", "folklore" - are reminders that individual actions have consequences.

So maybe not noise/signal or random/deterministic - but "the stone that the builders rejected, has become the cornerstone".

Penicillin was a nuisance - until it wasn't
posted by Barbara Spitzer at 1:54 AM on January 28


« Older Corporate Music - How to Compose with no Soul   |   "Grace was like, school is so cool, I want to go... Newer »


You are not currently logged in. Log in or create a new account to post comments.