> intelligence, height, schizophrenia, etc - are necessarily massively polygenic, because one side of them is better for fitness than the other.
I'm curious which side of height you think is better. Being extremely tall or short is clearly bad for fitness (see the health issues of giants and dwarves). Being close to average seems optimal.
You say that with height one side is better for fitness than the other. Not really, it depends on the environment. Tall people need more food, but are better at fighting. In environment with little food and little violence, shortness is likely selected for. In an opposite environment tallness is likely selected for. (This is simplified, height affects other things than just food and fighting.)
Scott! Have you heard of the microbiome? It's a big fucking soup of highly variable biochemistry that is only very loosely under genetic control. And interestingly enough, all the schizophrenia risk genes worth talking about are MHC genes! Does that not tell you something?
Like, let's talk about the actual biology of the disease rather than speculating from a theoretical/statistical perspective!
Why is the abundance of Ruminococcus gnavus 10,000x higher in the guts of schizophrenics than in healthy controls? Why is that figure nearly 1,000,000x in treatment-resistant schizophrenics?
Something I usually find missing in these discussions is noticing that effects, in general, are not additive. E.g. "each of which individually has a small effect, adding up to a large total effect". Genes are not a D&D like system where each contribution gives "+1" to whatever. There are complex interactions. I think of the extreme case as breaking systems on a plane: you remove one circuit controlling an engine, and nothing happens. You remove two, still good. Remove three, suddenly the engine malfunctions, and you go straight from "it's all ok" to "plummet and crash". Obviously it's not like all biological systems are made like this (although we have plenty of redundancy inside too). The more general version is a system with different components that still sort of complement each others functionality. E.g. a city accesible by train, road and a harbor may not function quite the same if roads are closed, but it's definitely not even 1/3rd as bad as "all roads, trains and harbor being closed".
I find especially relevant for answer 3 to the "why keep small bad effect genes", because it makes a lot of sense that you have a bunch of mutations which may cause a positive outcome (close the roads, less pollution!) while causing a bad effect that is much-less-than-additive compared to all of them activating (city is inaccesible, everybody starves). This naturally makes the "bad" effects of single genes much, much harder to see. Obviously the city example is exaggerated, and you'd see the effects of closing the roads - but when it's extended to dozens or hundreds of factors, one shouldn't even expect them cause effects on the order of 1%, as the naive "additive" decomposition would suggest.
There's much to be said about this, but one thing particularly jumping out was your comment "most random mutations are deleterious."
I have to disagree: silent or neutral mutations occur all the time. Indeed, evolution works by random mutation; if most were deleterious, life wouldn't have progressed past the most rudimentary organisms.
The elephant in the room is religiosity. Like most psychological traits that's like 50% heritable. I'm unaware of any genetic analysis of it, but I'd easily bet 10:1 it is also polygenic and significantly correlated with the Schizotypy/Schizophrenia genes. Meaning if you screen or engineer to prevent Schizophrenia, you're also screening or engineering against religiosity. In the upcoming huge culture war about genetic interventions in human reproduction, that's predictably a factor that's not going to calm things down.
> Schizophrenia is bad for fitness, so if it were genetic, evolution would have eliminated those genes.
I think what is overlooked is that phenotype dispersion (is that the right term?) is good for the species, even when it is bad for the individual. If the same set of genes creates descendants all over the spectrum of a certain trait, then you end up with gay uncles who do not procreate but are still useful. Or with a spectrum of intelligence where some do more intellectual labor and others more menial. Dispersion in gene expression for any set of genes would definitely shaft the unfortunates who end up with, say, severe schizophrenia or autism, but the species as a whole would do better because of the diversity.
"Studies seem to mostly support (1), for example this study of ancient hominid genomes finds that schizophrenia genes are getting less common over time"
Have you read Julian Jaynes' The Origin of Consciousness in the Breakdown of the Bicameral Mind?
Explanation 3 for those genes hanging around actually seems quite plausible to me. In aging biology, one of the key theories about why we age is ‘antagonistic pleiotropy’, where pleiotropy is the technical term for genes having multiple functions in different tissues/times of life/contexts. The idea in this case is that a gene that causes an animal to grow up and reproduce a bit more quickly and efficiently will be passed on even if it goes on to cause late-life deterioration, because by the time you’ve made it to late life you’ve already passed on your genes (hopefully more quickly and efficiently thanks to the antagonistically pleiotropic gene variants you carry) so you’ve achieved your purpose, evolutionarily speaking.
It’s ‘harder’ for evolution to get rid of a gene that also has a small advantage than one that is merely bad for aging (or schizophrenia) and has no counterbalancing advantage, so actually the 3 explanation seems more likely than 1, though both exist on a continuum to some extent.
I'm not sure I buy this model. If there were a million independent genes that each increased your risk of schizophrenia by an absolute 1-in-a-million chance, and everyone had an independent 50/50 chance of having each gene, then it would indeed be hard to filter that out through evolution, but there also wouldn't be such a thing as being genetically prone to schizophrenia, because every single person on earth would have a chance of getting schizophrenia between 49.5% and 50.5% no matter how lucky they got with their genes.
The problem is that many small independent effects don't necessarily add up to a large difference in effects between people - a large difference between the worst possible and best possible outcomes, yes, but not necessarily between the 1st percentile and 99th percentile.
The fact that some people *can* be predicted to get schizophrenia at a much higher rate just on the basis of their genes means there's something for evolution to easily select on - if those people stop reproducing, the total amount of schizophrenia genes in the population will materially go down (because the ones evolution removed had, by assumption, like twice as many of those genes as normal). This assumes a model where these things are basically additive, but that seems roughly right to me.
>The scare-mongering here has to be false - that is, it can’t be bad to choose an embryo at the 50th percentile of schizophrenia risk rather than the 99.9th, because half of people are at the 50th percentile of schizophrenia risk and nothing bad happens to them
I think the worry is that if we either learn to genetically engineer or do embryo selection for long enough, we could pick people with 0% schizophrenia genes (which wouldn't happen naturally) and only figure it out 20 years later when we have a new generation of weirdly uncreative adults. I don't take this fear totally seriously, but it does imply we should be pretty careful with gene selection if/when we do get it to avoid intense selection pressure on everyone doing it at once
Amateur question: when you select embryos based on the presence or absence of a specific gene, is there a risk of inadvertently selecting for unrelated traits due to gene correlation? Is there such a thing as gene correlation? For instance, in choosing an embryo with a low risk for schizophrenia, might we unintentionally favour one with a higher risk for heart attacks? This would not be because the same genes cause both conditions, but rather because embryos with genes reducing schizophrenia risk might also possess genes that increase the risk of heart attacks.
I struggle with how to think about traits that are related to how people treat you. Physical attractiveness is clearly highly heritable. Does that mean that random people smiling at my oldest child when she was a baby is a heritable trait? Surely a lifetime of people being nice to you for no reason has a big effect on someone, and if you did a gene-based study you would almost certainly find 'people are nice to you' is highly heritable, but it's a clearly environmental factor. How is this controlled for, if at all?
Regarding hypothesis 2: across several cohorts in different countries, having a higher polygenic risk score for schizophrenia is positively correlated with having an artistic profession and with measures of creativity. https://www.nature.com/articles/nn.4040
In Theory 1, a gradual decline in prevalence may have been hampered and slowed in past times by schizophrenic "scary bosses", whose symptoms of occasional sudden morose suspicion and paranoia, possibly leading to unpredictable violence, may have helped them maintain dominance through fear. And (male) chiefs in ancient times tended to monopolize women and have lots of children.
Arguably this applies more to hypothesis 2, but with not so positive creativity in the form of menacing cunning.
I think no. 1 is the correct answer. The mathsy way of expressing this is "Nearly Neutral Theory", which says that just knowing the "selection coefficient", which is a number for how deleterious the mutation is, is not enough. You also need to know the effective population size, because individuals don't evolve - populations do. In species with high population sizes (e.g. bacteria) slightly deleterious mutations are eliminated more quickly. The lower the effective population size e.g. humans, the more likely that "nearly neutral" mutations are invisible to selection. If an allele isn't eliminated, the only other option is that it eventually becomes fixed i.e. the mutated version becomes the new normal, even though it was a slight downgrade.
(2) other newish financial/social/economic pressures to delay childbearing into one's 30s
I speculate that if (1) is correct, these factors should quickly and massively decrease the incidence of incapacitating early-onset schizophrenia, particularly in males, who often express it by age 20 or so.
By "quickly", I mean within a few generations? So, would be an interesting and somewhat controlled experiment but also a long one.
I think the case for 2 is pretty strong, especially with the framework of something like the Diametric model which argues for the partial integration of both schizophrenic and autistic traits along a spectrum. If this model is correct it would explain the ubiquity of both autistic and schizophrenic traits across populations.
But we started with schizophrenia being evenly distributed through the human race, so there are no low-schizophrenia groups to be found. Or might there be very small (family-sized? village-sized?) groups with high or low schizophrenia rate?
I don't think this explanation is true for intelligence and height.
Selection is proportional to the additive heritability on the absolute scale, which makes your explanation true for schizophrenia: If there is a lot of liability-scale heritability, but the heritability is due to many variants of little effect and the prevalence is low, that translates to very little absolute-scale heritability, and it is true that evolution would have a hard time removing it.
But for height and intelligence, the tiny effects happen on the absolute scale, which means that if they have a strong relationship to fitness, evolution would be very quick at changing them. Like I guess it wouldn't exactly lead to eliminating the variants, but it would move you into a region with diminishing returns to them, making the evolutionary aspect less relevant.
I'm a postdoc working in bioinformatics, with a focus on cancer & polygenic scores but with a background in (mathematical) evolutionary theory. In my experience there is a surprising amount of misinformation - or just lack of knowledge, depending on how you see it - in medicine and genomics because even many experts are so used to monogenic risks since it was all we could feasibly find for a long time that they forgot that polygenicity is most probably the norm. It's arguably the core of the modern synthesis that happened in the early 20th century. At that point, mendelian inheritance had been proven, but we could see in many traits such as height that there was instead a continuous variation. This was explained by Fisher in 1918 by large numbers of small-effect loci, see "The Correlation between Relatives on the Supposition of Mendelian Inheritance" (he didn't use the word polygenic, but it's the same concept). So we're currently mostly just retreading ground that has been covered a literal century ago. There is a lot of followup papers on this and closely related topics, such as mutational burden, muller's ratchet, etc. that all show how negative fitness small effect mutations - which are the most common mutations to begin with - can stay in a population for a long time and even become fixed.
"Evolution hasn’t had time to remove all of them yet. Because a gene that increases schizophrenia risk 0.001% barely changes fitness at all, it takes evolution forever to get rid of it. And by that time, maybe some new mildly-deleterious mutations have cropped up that need to be selected out."
This does not make sense as a story. It is not harder for evolution to remove 100 genes of small effect than it is for it to remove 1 gene of large effect; the returns to selection is controlled ONLY by (narrow) heritability and NOT by the concentration across genes. Once you know the heritability, you know how easy it is for evolution to increase/decrease the trait; further knowledge of whether it's one gene of large effect or many of small effect does not tell you anything else about response to selection!
It is, of course, possible for some unfit genes to remain in the population due to the fact that they are constantly reintroduced via mutation. I'm not saying purely-bad genes are impossible or anything. I'm just saying that "many genes of small effect" does not have any explanatory power for why the genes were not selected out.
"Most random mutations are deleterious" is so oversimplified high-school biology. Many random mutations that we know about are deleterious, because that's how we know about them. Most random mutations are completely neutral in their effect, being either silent mutations (where there is no change to amino acid sequence of resulting proteins) or in non-coding regions. For the ones that do have an effect, we pay attention to bad things but not to improvements*, so we are more likely to be unaware of the beneficial* effects of random mutations.
*And all of this chatter about "more fit," "advantageous," "improvement," "beneficial" is dependent on the environment. What is advantageous in one environment may be an extinction-level trait in another. There is no such thing as an "evolutionary mistake"--there's just a trait that may not have been selected for yet.
> this study of ancient hominid genomes finds that schizophrenia genes are getting less common over time
Wild speculation: this helps explain the explosion of ancient religions as opposed to the relative dearth of new gods. Far more people heard voices in ancient times, attributed them to the divine, and boom.
I recall in an earlier post someone did bring up schizophrenia genes of larger effect. In the face of selection that's possible with rare de novo mutations.
Hmmm. Genetics doesnt always work in the ways we expect.
By analogy: when we were scampering around on all fours, a few of us had various genes for smaller nimbler front legs and feet. Some were able to walk around a bit on just our back legs. Only for short bursts. And at some serious costs in terms of backache, risks of falling over, loss of speed, etc.
And yet... this apparently rubbish set of aberrations came together and paved the way for hands ... thumbs... and a species that is able to dominate the planet.
So i suspect that individual genes that facilitate schizophrenia are often positive in ones and twos, but they cause a problem when they all crop up in one individual.
> So many of the traits we’re most interested in - intelligence, strength, schizophrenia, etc - are 𝘯𝘦𝘤𝘦𝘴𝘴𝘢𝘳𝘪𝘭𝘺 massively polygenic, because one side of them is better for fitness than the other. If they were monogenic, evolution would have already selected for the good side, and there would be no remaining genetic variance.
This is wrong. You can easily make the case for schizophrenia, but for strength and intelligence we want to draw the 𝗼𝗽𝗽𝗼𝘀𝗶𝘁𝗲 conclusion - it is definitely not the case that one side has historically been better for fitness than the other, because the amount of variation within the population is great enough that selective pressure would have had no difficulty producing an obvious response.
(Compare schizophrenia, where variation within the population is extremely low, which is what you'd expect from selective pressure against it.)
If you're having trouble with the idea that more strength could be worse, Steven Pinker put the point pretty well (though he was talking about intelligence) when he said, in my paraphrase from memory, "People present the evolution of humanity as being driven by selection for larger brain size, with increased intelligence as a happy side effect. But this is absurd. Metabolically, the brain is a pig. Any selection on brain size alone would surely have favored the pinhead."
It's a shame to cite Liu et al. 2019 without synthesizing one of the more interesting hypotheses in that paper. Granted, the paper itself is dismissive of it. It says:
“Although it has been reported that fertility among relatives of patients with schizophrenia is increased, a large cohort study and meta-analysis identified that this increase was too small to counterbalance the reduced fitness of affected patients (Bundy et al., 2011; Power et al., 2013). In fact, MacCabe et al. (2009) showed that patients with schizophrenia had fewer grandchildren than in the general population, demonstrating that the reduced reproductivity persists into subsequent generations.”
So the authors of this paper note that non-descendant relatives of schizophrenia patients have increased fertility but don’t find this fact interesting enough to incorporate into their overall interpretation.
Yet there is a way of interpreting that information that strikes me as painting a considerably more compelling picture: the decline in schizophrenia-dispositive SNPs does not presage the eradication of schizophrenia, but is part of an equilibrium-seeking adjustment to the group-level predominance of such SNPs.
This pattern is observed for left-handedness. It is advantageous to be right-handed in a right-handed world: it aids cooperation; tools are made for you. But in conjunction with other traits, it is also advantageous to be left-handed, specifically because left-handedness is rare. The word “sinister” comes from a root meaning “left” for this reason: lefties were looked down upon, but also hated, because their left-handedness gave them particular advantages—to wit, the element of surprise in combat—as a result of which left-handedness was never eradicated despite its obvious cost.
Could schizophrenia-dispositive SNPs not have a similar role? I’ll put my cards on the table and say I believe many of them do—specifically, that they contribute to shamanic and prophetic tendencies which have a positive kin-selective effect. I'd be remiss not to cite my friend Drew Schorno as helping me form this view: https://arcove.substack.com/p/null-call
Switching to a metaphor I find easier to work with: schizophrenia-dispositive SNPs, or at least some subset of them, have their effect because they raise the effective temperature of cognition: long-range connections, seemingly meaningless “accidents”, fail to be ruled out immediately as they are in a neurotypical mind. The extra “heat” means that people with a lot of schizophrenia-dispositive SNPs think a lot more wild, nonsense thoughts. This is usually not good for the individual but may be very valuable for their community, because every once in a while their wild thinking leads them to notice something very important like the hidden ill intentions of a neighboring tribe or signs of an impending drought.
Schizophrenia itself is a failure mode: the cognitive temperature is so high that the community can’t keep the lines of communication open (although—note that pre-industrial societies were a lot less likely to medicalize and ostracize people with schizophrenia-like symptoms). But schizophrenia-adjacent “disorders”, “personalities”, whatever, do appear to serve a social purpose.
This is consistent with the results that Liu et al. cite, that relatives of schizophrenia patients have some level of increased fertility while the patients themselves do not. At the same time, the net effect on the kin group is negative, which is consistent with the view that the prevalence of schizophrenia-dispositive SNPs will continue to trend downwards. I suggest that the end-point of this will not be eradication of those SNPs but a new equilibrium.
"This seems less like the sort of thing that happens naturally, and more like the sort of thing you would claim if you wanted to make your theory untestable."
Can you point to where Torrey explicitly writes this in his article?
Wouldn’t possibility #3 show up on genetic correlations between schizophrenia and other health states? From what I’ve seen reported in research studies, most genetic correlations of schizophrenia are with other diseases, including other forms of psychopathology, but also things like cardiovascular disorders and immune disorders. So #3 seems unlikely. Some studies have found positive correlations with educational attainment, but recent studies have shown that this is explained by the genetic overlap between bipolar and SZ, and that unique SZ genes had a negative effect on educational attainment.
A (very likely?) possibility that can have an outsized effect on our evolutionary understanding is that schizophrenia is like weak ankles or some such -- it's a latent thing which isn't usually a problem until it gets nurtured with a sledgehammer or similar. In the past the other genetic effects would have been excellent for fitness except in rare cases, but in a world where weed (or, I dunno, television) is freely available it becomes vastly less adaptive.
First and foremost, I would like to thank KW, for this article. KW has way with opening our minds to conversations and greater positive awareness on different topics.
As this article presents the genetic material of this matter. What about the randomness of the schizophrenic people we know, ie: one of my non-biological grandmothers sister, due to her insidious and behavioral thought patterns that altered her, was diagnosed clinically, was institutionalized, yes, we still have a medical facility for the diagnosed disease, “Greystone”. But speaking with my mother who is not related, stated, my aunt and uncles are also diagnosed with the same irrational disease. Interesting, right?
What I know as schizophrenia, has a stigma this is DID and MPD, disassociated identity disorder, and multiple personality disorder. This is how we speak of schizophrenia. In all of my studies I have realized, it has the possibility of genetic traits, (genetics saying it has to do with the 6th chromosome and 22, now, not in the 50’s)
My question to this is, how long has this truly been genetically studied? The environment in which we were reared, has evolved us or mutated us into something we are not.
One great example of this is “Sybil”. She grew up in the 50’s, was the product of a very abusive childhood, in part, due to her religious upbringing. (7th Day Adventist). In the 50’s when her story came to being, as she had 16 MPD’s, which grew a wide spread fascination by the medical community, psychiatrists and psychologists lobbied to have MPD or schizophrenia, to be put in their “bible; DSM or Diagnostic and Statistical Manual”, this rare diagnosis led to a common diagnosis. “There were less than 200 cases in Western civilization” in or around the 50’s. “But after the book and film in the 1970’s, her case sparked hundreds of thousands of prognosis’s, and by the late 80’s there were 40 thousand diagnosed case in the US alone.”
Does this make a case for genetics and environment? I think so.
For example, in 1986 (date coincides to the diagnostic psych manual for MPD), Billy Mulligan, he had over 24 different personalities. He was diagnosed young, said he was a psychopath, chronic liar, and a murderer. He was seen by several psychiatrists put on Thorazine, a common drug for MPD. He stated he was becoming worse due to the publicity, society, and medication.
This subject has always fascinated me. I see in everyday situations, like a “narcissistic/pathological liars”, people who make up lies, and believe their lies and grandiosity, has become part of the mutating brain from all of the trauma. Right? Slightly altered personality is considered either schizophrenic or MPD.
Ok, there are two examples of cognitive behavior. This leads to, if not monitored, Schizophrenia/MPD/DID. These different behaviors change the brain, people actually perceive their beliefs and morphs them into their own monsters or demons, your preference.
Ted Bundy, Charles Manson, even though these guys are murderers, they heightened their own perception/personality (MPD) to believe what they did was ok, and fought for it. True or not? I do not believe we inherit a gene for their delusional beliefs/thoughts through our society.
Hey, nice series. I ran your article through Tutor GPT and wrote about it on my Stack! It's complicated stuff, I appreciate the insight. We are working on making diagnostics for polygenic diseases so this is super fun to think about. As an aside, have people run a subset of these through a markov model to simplify the gene-only interactions? then you look at the other end of it and look at a PCA and see which of these genes are migrating over time to see what we're selecting for . if its really .1% one way or another every generation over 50 generations we'd just see the probability clould jiggle for a thousand years
"Because a gene that increases schizophrenia risk 0.001% barely changes fitness at all, it takes evolution forever to get rid of it. And by that time, maybe some new mildly-deleterious mutations have cropped up that need to be selected out."
I don't understand the second part of this. Evolutionary pressures don't operate on one mutation at a time, do they?
>The scare-mongering here has to be false - that is, it can’t be bad to choose an embryo at the 50th percentile of schizophrenia risk rather than the 99.9th, because half of people are at the 50th percentile of schizophrenia risk and nothing bad happens to them. Schizophrenia genes can be at best fitness-neutral;
I think that you're conflating two different notions of 'bad' here.
It might be desirable to select an embryo with traits that are not maximally-reproductively-fit in the evolutionary environment, but are useful in modern humans.
And given that we're talking about variance in mental traits here, that even seem like a pretty likely scenario to me.
I'm rather sure it's a combination of 1 and 3. Assuming that mutations occur almost randomly, the ones that have an advantage as well as a disadvantage will tend to be selected against (or for) more slowly. And there are a lot more ways to break something than the improve it, if it's any good at all.
I.e., the average mutation that is detrimental will be selected against more strongly if it doesn't have any advantages. And there's no reason to believe that the advantages will be related to the disadvantages, except through a VERY tortuous chain of interactions. (And, of course, almost all mutations are either neutral or detrimental. Neutral drift theory enters here importantly, but whether the "neutral mutation" is absolutely neutral is very difficult to demonstrate. E.g. some yield the exact same proteins, but make the RNA slightly less stable. Just try to prove whether that's advantageous or not.)
Additionally, some mutations are advantageous (or neutral) in some environments, and not in others. (E.g. being tall and thin is disadvantageous in really cold climates, but advantageous when the temperatures are higher. [But not too much higher. I think the advantage disappears when the average highs are over about 95F, though that's a wild guess.])
It is not at all surprising that complex traits are polygenic, for a different reason than the evolutionary one described here. We have only 20,000 genes; roughly speaking, 20,000 "ingredients". How do you get from this the staggering complexity of form and function we observe? Through combinations of genes acting together. You can wire together 3, for example, to form a time-keeping oscillator (something I had my class act out a few days ago). Just as it is silly to ask which one is the oscillator gene, it is silly to expect one, or even a few "schizophrenia genes." Even height has ~1000 determinants. The idea of one gene = one trait is incorrect and sadly widespread.
Low confidence speculation here, but I have often gotten the impression that schizophrenia is sort of the opposite of autism. I wonder if the low schizophrenia risk people have high autism risk and vice versa, with "healthy normal" being a balance between the two.
The problem of polygenicity is not exclusive to schizophrenia. Indeed most intractably difficult modern diseases are thought to be polygenic, and for the same reasons. I studied asthma in grad school, where half my lab was arguably genetics driven.
Rates of asthma have been dramatically increasing for decades (including severe asthma requiring hospitalization, so you know it's not just over diagnosis). If that's the case, shouldn't we be focusing on some environmental factor?
Enter "Danger Theory" and the Hygiene Hypothesis. The immune profile of allergies, asthma, some kinds of IBD, and other auto-inflammatory and auto-immune diseases often looks like an inappropriate activation of the immune response to a parasitic infection (an inherently polygenic process). What's that all about?
The thinking goes that humans throughout history have been exposed to constant parasite challenge, but that this has gone down dramatically in the modern era. Different people have a different 'threshold' for immune activation of a parasite, with some people allowing more infestations and others activating against the slightest threat. Why doesn't everyone activate at high levels? Well, activating the immune system is inherently destructive, so activating inappropriately can have negative side effects. Those effects are worth it when you get the activation right, but harmful when you're wrong.
Until two hundred years ago, having an immune system on high alert to parasites would have been evolutionarily advantageous (most of the time). High threshold people would be sick a lot more, since they'd allow more parasites in, while low threshold people would be out working. In today's society the dynamic is exactly the opposite, since most parasite activations are going to be wrong and therefore harmful.
If baseline immune activation is more of a rheostat than an on-off switch, it makes much more sense for this to be controlled at a population level by variable or polygenic mechanisms, than by a single mutation that 'causes' some specific baseline level of immune alertness to parasites.
Multiple studies have demonstrated genetic components to many of these immune-related diseases, but there's reason to suspect that the changing human environment/culture/society has shifted the definition of what is adaptive versus maladaptive. What was selected for yesterday may be selected against tomorrow, and that is true across millions of years of evolution. In other words, environment determines whether a gene confers fitness or not.
What is natural selection, if not long-run genetic responses to environmental stimuli? Without environmental inputs, evolution has nothing to direct it. Therefore we can expect that dramatic changes to an organism's environment will result in some adaptations becoming deleterious - especially at a population level.
What holds true for immunology almost certainly holds true for psychiatry as well. Human interactions have shifted dramatically over the past few hundred years, to the point where the human brain is operating in an environment significantly different from the training data.
(Alternately, maybe this is all just cope to avoid having to explain that we don't understand the underlying mechanisms? Even so, I think the polygenic case still holds for most human diseases I've looked at.)
I think 1% of people are at 50th percentile risk, not 50%.
Your previous argument showed that dropping the extremal 1% barely affected the next generation, but collapsing the distribution probably will have significant long-term effects (hopefully positive!).
Scott is right and Torrey is wrong in regard to these two objections. Evolution can fail to remove bad effects even if they are monogenic, or low polygenic, if they also have some fitness advantage. Two examples:
- Sickle cell anemia is monogenic but persisted due to its advantage regarding malaria.
- Hyperlipidemia with elevated ApoB drives cardiovascular disease. It is polygenic, but not greatly. It persisted because it was adaptive for energy use during period of food scarcity, such as the Ice Age.
Intuitive arguments about fitness often go wrong because we are naive about the many varied things that contribute to fitness.
Isn't there a negative correlation between height and longevity? That is, shorter people live longer on average and have more time to contribute to the well being of their offspring. That alone would exert powerful selective pressure towards moderation in height across generations.
"The scare-mongering here has to be false - that is, it can’t be bad to choose an embryo at the 50th percentile of schizophrenia risk rather than the 99.9th, because half of people are at the 50th percentile of schizophrenia risk and nothing bad happens to them."
This is normatively true, not best-of-all-possible-worlds true. Neurotypes have the prevalences they do because, approximately, that was about as common as worked best in ancestral environments. A trait being the midpoint or most common doesn't mean it's the best; it would be unreasonable to say people at the 50th percentile of intelligence probably don't have any problems that aren't also faced by people at the 99.9th percentile of intelligence. Humans are "about as smart as made for the best tradeoffs", which is less smart than ideal.
I don't think the neurotype carved out as NT is the best of all possible worlds, or that it should be "the highest-prevalence by a juggernaut huge margin". Of course, it has various benefits (e.g. acting on plans rather than theorizing about them forever, big-picture thought, caregiving, etc). A world with no primary care physicians would be a disaster. A world with no MBAs...uh...okay, we'll circle back to that one. I don't want a world where all the smart people are primary care physicians or MBAs, though.
Here's an idea for a possible fourth explanation, but with the caveat that I'm not a biologist and I'm open to arguments if this is implausible. I've thought about this every time someone brings up the "but homosexuality must have some evolutionary advantage or it wouldn't exist, something something gay uncles" line.
Suppose you have the job of manufacturing parts to a certain tolerance. The more exact your production line, the more expensive the process - basically once you've calibrated the machines to get the average part about right, the smaller you want the SD of your output parts' measurements to be, the more it'll cost you. You can trade off between two ends of a scale. At one extreme, make your process good enough that every single part that rolls off the line is within tolerance (with the most expensive machine ever). At the other extreme, make parts where those in some range like half a SD from the mean are within tolerance, with the cheapest machine you can bodge together, measure each part that comes out and trash (or recycle if you can) those that aren't. Either way, every part that makes it to the customer meets the spec. Of course, the optimal point is a balance between the cost of the materials (including how easy they are to melt down and start over) and the cost of getting the assembly line to a certain precision; if you're cutting natural diamonds then you're at the opposite end of the spectrum to if you're 3D printing plastic.
Maybe evolution is doing the same? If the optimal height of humans in some environment is 160cm, and an evolved mechanism for making everyone exactly that high even if some grow up better nourished than others would have more of a fitness or energy cost than making everyone grow "close enough", I'd expect evolution to take the latter path.
The old "group selection" post (Studies on Slack) mentioned that in on a planet that gets hit by a solar flare once an eon, species will not evolve a radiation shield if that costs slightly more in energy for each individual during all the years when there's not a need for it. Similarly, if schizoprenia / homosexuality / autism / pick your favourite condition is a net negative from a reproductive fitness perspective, even if a biological mechanism is available to filter it out and evolution would have the time to do so, I'd expect the filter not to exist if the cost it imposes on the individuals (such as extra energy expenditure) is in some sense higher than the reproductive fitness cost of not evolving the filter.
Evolution has, after all, managed to create frogs and fish that have thousands of children and 999 of them generally get eaten or otherwise die before sexual maturity, which is the "cheap machine" end of the scale (which I believe is called r/k selection in biology textbooks).
Is it plausible that schizoprenia exists because the evolutionalry cost of it not existing would be higher than otherwise? Without group selection? Or have I missed something obvious here?
The idea of "schizophrenia genes" is counterintuitive and not particularly useful. And, since schizophrenia is a mental dysfunction it is natural to look for an association with other brain mediated attributes such as intelligence or creativity.
But perhaps that is like looking for your lost key under the street light when you actually dropped it across the street.
There is a lot of data that shows an association between schizophrenia and every aspect of the immune system. A highly dysregulated immune response is a frequent comorbidity in schizophrenics and a promising area of research.
Further understandings on the interplay of the different arms of the immune system, the immune cell types implicated and their origin, the role of immune responses to common viruses, the gut brain immune axis, and aspects of inflammation and autoimmunity, are bound to shed a great deal of light on the causes of schizophrenia.
> The clearest way to resolve these questions would be to genetically engineer someone to zero schizophrenia risk and see what happened (this is beyond current technology)
Just checking, but this is also beyond current ethical boundaries, right? Right???
Studies show that the polygenic risk score for schizophrenia is not associated with creativity, indicating limited contribution of the accumulated effect of these risk variants to creativity.
However, neural inflammation and certain viral infections are strongly correlated with the development of schizophrenia, as is infection with the parasite Toxoplasmosis.
Perhaps the correlation between schizophrenia and creativity is due, in part, to shared vulnerability factors in brain architecture and function, including neural hyper-connectivity, novelty salience, cognitive disinhibition, and emotional lability.
Thus, individuals with both a genetic predisposition to specific differences in brain wiring and function, AND also having an abnormal immune response to infection by certain viruses (Epstein barr, CMV, retroviruses, etc) or parasites (Toxoplasmosis), are statistically far more likely to develop schizophrenia.
It's like rolling snake eyes in a game of chance. A creative brain is only a risk when combined with a dysregulated immune response to specific but commonly encountered pathogens.
No doubt there are other pathways to the development of schizophrenia and it's all more complex than what we currently know. Future research will undoubtedly reveal far more about it.
Viral infections correlated with the development of schizophrenia:
I'm not clear why we're rejecting #2. Possibility #1 seems like the default hypothesis, and #2 sounds reasonable to me, in a way like "the dose makes the poison". #3 is similar, but requires that the gene influence at least two unrelated systems, so it seems less likely to happen than #2 which just requires a small effect on one system. So just on intuition I'd guess a mixture of #1 and #2 would be most likely. But this is really far from anything I know well.
There are some crucial things I don't get about how polygenic risk works. Can somebody here explain a few basic things? So there are thousands of genes that cause schizophrenia, sprinkled all over the genome. So let's say there are 5000 genes that cause schizophrenia, each of which raises risk by a tiny amount, on average .02%. So does someone have to get all 5000 of them to be schizophrenic? But the chances of someone getting all 5000 of them is minuscule! Is it more that you have to get, say, at least 2000 of them. So what are the chances of someone getting 2000 out of the 5000? If these genes are scattered all over the genome, then the chance of getting any one of them is not correlated with the chance of getting any of the others. (I think. Is that right?). So then it should be possible to calculate the odds of someone getting,say, 2000 particular genes. If you take any one of these genes, what fraction of the population has it? Does one assume it's 50%? If you know how likely someone is to get any one of the genes, and approx how many genes there are that increase risk of schizophrenia, it should be possible to figure out what fraction of the population will be schizophrenic. And the result should be approx. 1%, since that what fraction of the population is. I'm sure I'm not the only one who'd like some help with understanding these basics.
Seems plausible to me that some of the genes that raise the risk of schizophrenia are genes that cause the person to have a harder life. You can think of them as genes that affect the individual's environment. For instance, physically attractive kids are better liked by peers and seen as smarter by teachers: They literally get more smiles. Just now looked up research on schizophrenia & physical attractiveness. There were studies that found photos of schizophrenics were rated as less attractive by judges than those of non-schizophrenic peers, but I think that result might be accounted for by schizophrenics being more likely to have poor grooming and and other signs of self-neglect, and a less pleasant facial expression. But then found one where the high school photos of people who later became schizophrenic were rated as less attractive. It's herer:
A. Farina et al (see record 1978-23202-001) investigated the relation between mental illness and physical attractiveness and found that female psychiatric inpatients were less attractive than normal controls. The current study extended this investigation in 2 ways. First, 28 psychiatric inpatients were compared to 3 separate control groups of 53 low, middle, and high socioeconomic status Ss. Mental patients were judged significantly less attractive than either middle- or high-income controls but were not significantly different from low-income controls. Second, to examine physical attractiveness prior to hospitalization, attractiveness ratings of the patients' high school yearbook pictures were compared with ratings of the adjacent same-sex photographs. Patients' photographs were judged significantly less attractive than their peers' even in high school. Findings suggest that being physically unattractive may predispose an individual to a number of negative social outcomes, one of which is mental illness.
((Napoleon, T., Chassin, L., & Young, R. D. (1980). A replication and extension of "Physical attractiveness and mental illness." Journal of Abnormal Psychology, 89(2), 250–253. https://doi.org/10.1037/0021-843X.89.2.250)
I realize this is somewhat off topic, but I recall from my high school biology class that researchers had isolated a gene for manic-depression. In fact two different studies, one focusing on Iceland, and the other on Pennsylvania Amish, had located two different genes, on two different chromosomes. Is this still considered true, or has it been discarded in the intervening years?
If still considered true, it would indicate that at least some mental disorders have single-gene causes.
"why are there still even these genes of very small effect?"
Explanation 1) "Evolution hasn’t had time to remove all of them yet" seems to assume that it's harder for evolution to remove a lot of small cumulative genes than one single gene. I don't believe this to be the case.
To get an intuition, I quickly wrote a small python script which evolves a small population with 1024 binary genes under two different conditions:
a) a 50% probability of death if gene 666 is True
b) a 50% probability of death if the 666th fourier component of the genome is >0.5
(I didn't implement recessive or dominant traits)
Both factors seem to have approximately the same half-life, even though one is a single gene and the other is a bunch of genes with small effect. I'm not posting the script right now as there is a high probability that I've made some embarrassing error, since I wrote it in 10 minutes and didn't do much testing, but the result seems plausible. Unless there is something special about the 666'th fourier component, I'd guess this to be true of any linear transform component.
Dunno as an excellent Genetics graduate and one who has been told that he is exceptionally creative (humble brag), and one who initially self-diagnosed and then was formally and objectively diagnosed with schizophrenia, I really, _really_, hope that there is some exceptional if rare boost to having this condition.
The big risks I see of embryo selection are Goodhart's law and out of sample extrapolation. It's possible for gene A to be correlated with B in an unselected population while still having it be a bad idea to exercise extreme selection pressure on A, either because the correlation breaks down under selection or because their are non-linear effects when taken to extremes or because there are other unknown side effects.
Those aren't issues for any plausible near-term IVF selection, but could cause problems in some of the more extreme proposals (selecting from thousands of embryos or even straight out genetic engineering.)
About the image for this piece: It's a little known fact that lions suck the schizophrenogenic genes out of their prey. For a lion they're like Reese's peanut butter cups.
Chris Masterjohn has been thinking along the same lines, but came to a different conclusion. According to his logic (link: https://open.substack.com/pub/chrismasterjohnphd/p/unlocking-performance-and-longevity) most SNPs analyzed are just noise and there are only a few important ones, but scientists need a mechanistic (ie biochemical) model of how they interact, not brute force statistics…
Quote: Common polymorphisms by definition on average produce the average, because they are defined by their deviation from normal. Rare disease genes on average produce a rare disease, because they are defined by their pathogenicity. Natural selection minimizes the presence of harmful genes, which makes the rarity of a gene correlate to its severity. Thus, the biggest problems have to be the few rare problems, not the many common variations.
The individuals who display a disease phenotype because they are heterozygous for more than one severe defect in related pathways have been termed ‘synergistic heterozygosity’.
I have a friend with early-onset schizophrenia who considers it "a superpower that you have to work hard and learn to deal with." In other words, I think he'd take issue with a blanket statement about it decreasing fitness. Having seen his reflexes and pattern-detection in action I would find it hard to disagree; he's survived situations that would have killed someone slower on the react or less inclined to observe everything and notice connections between things. In other words, less paranoid and ready to act accordingly. Compared to him, most people are just not paying attention. Perhaps the increases in fitness for the people who are able to deal with it outweigh the decreases in fitness for those who can't. After all, we don't necessarily know who all is schizophrenic. It's not only the people who go get medicated about it. And antipsychotics decrease fitness in and of themselves, so there's a selection bias there.
None of that is to say that anything you said about the geneticity of it is wrong. His family has quite the history of it.
(The odd thing about talking about human genetic evolution is that it's probably over. Assuming we *don't* kill ourselves and *don't* halt all interesting tech advancement, we will probably mostly move away from this form the way have we mostly moved away from caves, adobe houses, and horse carts. In the interim, knowing about evolution will help us to develop patches and also to understand how our brains work. But the following is about natural human DNA evolution as if it were a going concern.)
Genomes evolve to evolve, and without going into the ways they manage that, but antropomorphizing: the overall genome and organism design gets selected toward *some balance* of dependable to "speculative". If we wanted to let human evolution continue by its own processes, it would *not* make sense to fix a lot of genes that contribute to schizophrenia, because then the results of lots of combinations of those and other genes would not get tested. In other words, if we assume our genome is at about the right place on the risk/reward curve, adjusting for less risk for individuals in the very next generation would mean fewer beneficial advancements on many-generation time scales. It's probably even harder to identify the meta-evolutionary mechanisms than the functional effects of current gene prevalences on odds of schizophrenia, but evolvability-tuning can be seen in simple situations and it makes sense to assume it happens in general.
I say this to add another spin on what function all those potentially harmful genes might serve. For instance, in the last couple millions to thousands of years, it may have served our genes well to invest in a lot of brain-tech startups. I believe (sorry forget source!) that humans have a particularly high ratio of brain-specific to other genes.
Another spin is that different *distributions* of characteristics in populations not only effect the population-level success but also affect the context in which individual genes can be beneficial or detrimental. Sheer diversity can be beneficial if the body, brain and social context can accommodate it.
There is also the possibility of non linear effects, where you only get a bad effect if you have, say, A, B, and C, and where any one or pair of them does nothing. That would further make it difficult for evolution to get rid of these, as in most people the effect is zero
The TL:DR proposal he has for why genes that lead to schizophrenia got selected for is that they can make people more prone to magical thinking and intuitive/creative ways of perceiving things that a lot of cultures historically in people playing the roll of shaman or spiritual leader. In a lot of these cultures these people weren’t expected to be celibate and were highly valued so the traits got passed on.
If you’re trying to survive in the wilderness and believe in spirits of the dead and forces of nature, you value having someone who thinks they senses things you can’t and who makes strange connections. We know that people related to schizophrenic people tend to be into conspiracy theories or occult beliefs or any kind of interest where you make connection between things other people normally wouldn’t. Someone obsessed with making connections that seem spurious but interesting feels like a summary to a lot of ancient mystical thinking imo. It’s not surprising people prone to that would be valued & I also would worry that we might be losing something of value trying to engineer those traits out of existence.
Quick note, you can't assume that because the people with the lowest skitzophrenia risk don't have other problems it follows that there is no such effect of these genes.
1) Its possible that those individuals have other compensating genes. Maybe one genetic heritage has some other set of beneficial genes that give the same benefit with another downside and other groups have the skitzophrenia promoting genes (you'd want to see if some mixed individuals do much better in average).
2) Its possible that the benefits aren't addative. Skitzophrenia is rare so what if each gene that increases risk offers a small benefit to those who have it but you don't get much extra benefit (any?) for having multiple genes that increase risk. This differs from your 3.
3) What if the 'benefit' offered by these genes is in utero or in pregnancy-- say they reduce head size. If they increase the chance of being born at all or not dying looking at those who didn't die tells you nothing.
4) Other really complex interactions between genes could be taking place.
Excellent discussion! Just to add to point (1), I think it’s important to remember that selective pressures act via *reproductive fitness. To my knowledge, the average age of onset in schizophrenia is near late-adolescence to early-adulthood. Considering that the average age of first-time mothers was 21 as recently as the 1970s, it seems totally reasonable that we might be overestimating the negative selection effects of schizophrenia, simply because most people throughout history likely began having children before they saw the full effects of their condition. I see this as similar to how many of us carry increased risk factors for cancer, heart disease, etc. that simply don’t have any effect until we’re past our main reproductive years. Clearly, there is some selective pressure acting here, or else there would not be a decline in schizophrenia over time, but my point is that it’s probably weaker than we would expect for other genetic/congenital conditions which act earlier in life.
I have skin in this game. We recently went through IVF and we had our embryos tested for polygenic disease risk. Our "best" embryo had significantly reduced risk for a wide range of diseases, except schizophrenia, where it was 85% percentile. We had other embryos with low schizophrenia risk, but higher risks of more common diseases. As you might imagine, we thought about this very carefully.
Ultimately we chose to implant the "best" embryo. Partly our reasoning is that 85% percentile of schizophrenia is still <2% absolute risk. But another part of our thinking was that there's some evidence of antagonistic pleiotropy, i.e. that higher schizophrenia polygenic risk score correlates with desirable traits.
"Higher educational attainment (EA) is negatively associated with schizophrenia (SZ). However, recent studies found a positive genetic correlation between EA and SZ. We find strong genetic dependence between EA and SZ that cannot be explained by chance, linkage disequilibrium, or assortative mating. Instead, several genes seem to have pleiotropic effects on EA and SZ, but without a clear pattern of sign concordance. Our results reveal that current SZ diagnoses aggregate over at least two disease subtypes: one part resembles high intelligence and bipolar disorder (BIP), while the other part is a cognitive disorder that is independent of BIP."
So I do think selecting for the lowest possible schizophrenia polygenic risk score is potentially a bad idea, unless you also have PGS scores for lots of other traits to ensure you're not inadvertently selecting against them.
The embryo we picked is now a happy little baby crawling around as I type this. Hopefully we made the right decision!
Conceptualizing a creativity gene is in unnecessary.
People live longer
X cost of raising children increasing
X easily available birth control
= children born later in life
Rate of SNP mutations are proportional to age of father at conception (between 20 & 40 years paternal mutation rate triples, not the maternal rate). This is perhaps ordinary entropic change of the paternal genome which was revealed by non-genetic changes in longevity due to global health and nutrition at population levels.
If you were to graph US average age of father at conception and US rate of schizophrenia and autism you should be able to see the effect of aging on accumulated genetic defects, completely independently of a “schizophrenic gene” which would be selected against strongly of course.
Unsurprisingly the rates of autism and schizophrenia have been rising in recent decades. Likewise Schizophrenia is more prevalent in males than females, understandably if it were possibly Y-linked, paternal influenced. Finally schizophrenia should be growing fastest in world regions where lifespan is showing the most rapid improvement, and paternal age of conception. I would hypothesizes it is most pronounced in central and subsaharan Africa.
Genes which favor non reproduction are easy to maintain equilibrium in a population if they enhance familial reproductive success even at the cost of individual offspring. This has long been discussed and modeled. [Dawkins, E.O. Wilson] A gene group which enhances longevity will enhance schizophrenia and autism without any biochemical relationship to the specific source of the root issue.
Consider disruption of NMDA receptor interaction with multiple modulatory transmitters - failure to attenuate sensory prediction error; Schizophrenia can be induced chemically by disrupting this system; schizophrenics are strikingly resistant to illusions normally created by imprecise perception; schizophrenics cannot involuntarily predict and track motion with their eyes. First-degree relatives also have eye tracking dysfunction but not full-blown schizophrenia. It only takes a few more SNP’s induced by age to cross the line. Epigenetic aging mechanisms induce it.
I’m gay, and have no interest in having kids. This at an individual level selects against reproductive fitness as a gene or group. However the frequency of gay men in the population is relatively stable. What produces non-reproducing offspring may enhance the reproduction of siblings indirectly, without biochemical signature.
Also, “last brother being gay” effect is driven by late maternal age of conception, whereby maternal antibodies to testosterone developed during sequential prior male pregnancies (even miscarriages) alter the process of testosterone imprinting on the fetal brain starting at 9 weeks. Another case of what appears as a possible gene effect is due to epigenetic factors related to aging, health care and nutrition at the population level.
This article had many comments, apologies if any of this is a rehash.
I believe most evolutionary theory and thinking is skewed towards considering humans as individual organisms. While of course our bodies are discrete organisms, when it comes to mind, we are necessarily linked. The brain is, among other things, a social organ. Mine, at least, seems quite busy and consumed with understanding how my larger social world works, its power dynamics, and what I need to do to have social value. What if schizophrenia is polygenic with many traits that map out and model meaning structures and complex social dynamics?
We are fundamentally complex social primates. So complex, in fact, that we cannot live alone. By the time all of us turn 12 (it is extremely rare for schizophrenia to manifest before this age), our families and tribes have nursed us, fed us, taken care of us, and bonded with us at no small sacrifice. Homo sapiens as I understand us would not lightly stop taking care of a kid that hears voices, often until past the age of reproduction.
H. sapiens evolved as tribes. We keep our elders around, for their wisdom and knowledge is a strong fitness advantage - for the whole tribe. We take care of each other, to a degree that is not rational.
Now I'm going to stereotype. Would love to hear others' data points.
The people I know who struggle with schizophrenia seem (anecdotally) to be above average when it comes to sensitivity and spirituality. I wonder if there are linkages between schizophrenia and mystics or shamans. As I observe mind and try to locate "self," I readily get lost in a sea of connections, social conditioning, culture, and my group's foundational myths and narratives - to a degree that might make me less functional, but also possibly mystical in my inner modeling of complex networks of human meaning, purpose, attention, and care.
I would posit that schizophrenia is polygenic and related to a host of genes that assist in the mapping and modeling of complex human social and meaning structures, and would require extremely fine tuning to select out without resulting in asociality or sociopathy.
Or maybe I have just watched *A Beautiful Mind* too many times?🤷♀️
> intelligence, height, schizophrenia, etc - are necessarily massively polygenic, because one side of them is better for fitness than the other.
I'm curious which side of height you think is better. Being extremely tall or short is clearly bad for fitness (see the health issues of giants and dwarves). Being close to average seems optimal.
You say that with height one side is better for fitness than the other. Not really, it depends on the environment. Tall people need more food, but are better at fighting. In environment with little food and little violence, shortness is likely selected for. In an opposite environment tallness is likely selected for. (This is simplified, height affects other things than just food and fighting.)
Scott! Have you heard of the microbiome? It's a big fucking soup of highly variable biochemistry that is only very loosely under genetic control. And interestingly enough, all the schizophrenia risk genes worth talking about are MHC genes! Does that not tell you something?
Like, let's talk about the actual biology of the disease rather than speculating from a theoretical/statistical perspective!
Why is the abundance of Ruminococcus gnavus 10,000x higher in the guts of schizophrenics than in healthy controls? Why is that figure nearly 1,000,000x in treatment-resistant schizophrenics?
Something I usually find missing in these discussions is noticing that effects, in general, are not additive. E.g. "each of which individually has a small effect, adding up to a large total effect". Genes are not a D&D like system where each contribution gives "+1" to whatever. There are complex interactions. I think of the extreme case as breaking systems on a plane: you remove one circuit controlling an engine, and nothing happens. You remove two, still good. Remove three, suddenly the engine malfunctions, and you go straight from "it's all ok" to "plummet and crash". Obviously it's not like all biological systems are made like this (although we have plenty of redundancy inside too). The more general version is a system with different components that still sort of complement each others functionality. E.g. a city accesible by train, road and a harbor may not function quite the same if roads are closed, but it's definitely not even 1/3rd as bad as "all roads, trains and harbor being closed".
I find especially relevant for answer 3 to the "why keep small bad effect genes", because it makes a lot of sense that you have a bunch of mutations which may cause a positive outcome (close the roads, less pollution!) while causing a bad effect that is much-less-than-additive compared to all of them activating (city is inaccesible, everybody starves). This naturally makes the "bad" effects of single genes much, much harder to see. Obviously the city example is exaggerated, and you'd see the effects of closing the roads - but when it's extended to dozens or hundreds of factors, one shouldn't even expect them cause effects on the order of 1%, as the naive "additive" decomposition would suggest.
There's much to be said about this, but one thing particularly jumping out was your comment "most random mutations are deleterious."
I have to disagree: silent or neutral mutations occur all the time. Indeed, evolution works by random mutation; if most were deleterious, life wouldn't have progressed past the most rudimentary organisms.
The case for 2 would be mostly the positive effects of Schizotypy. Some are known, it's not just creativity, nice overview paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4373632/
The elephant in the room is religiosity. Like most psychological traits that's like 50% heritable. I'm unaware of any genetic analysis of it, but I'd easily bet 10:1 it is also polygenic and significantly correlated with the Schizotypy/Schizophrenia genes. Meaning if you screen or engineer to prevent Schizophrenia, you're also screening or engineering against religiosity. In the upcoming huge culture war about genetic interventions in human reproduction, that's predictably a factor that's not going to calm things down.
> Schizophrenia is bad for fitness, so if it were genetic, evolution would have eliminated those genes.
I think what is overlooked is that phenotype dispersion (is that the right term?) is good for the species, even when it is bad for the individual. If the same set of genes creates descendants all over the spectrum of a certain trait, then you end up with gay uncles who do not procreate but are still useful. Or with a spectrum of intelligence where some do more intellectual labor and others more menial. Dispersion in gene expression for any set of genes would definitely shaft the unfortunates who end up with, say, severe schizophrenia or autism, but the species as a whole would do better because of the diversity.
There, I explained group selection.
Didn't Sasha Gusev use group differences in schizophrenia PGS to argue that they were biased by ancestry. Seems like it might be related to 1.
"Studies seem to mostly support (1), for example this study of ancient hominid genomes finds that schizophrenia genes are getting less common over time"
Have you read Julian Jaynes' The Origin of Consciousness in the Breakdown of the Bicameral Mind?
Explanation 3 for those genes hanging around actually seems quite plausible to me. In aging biology, one of the key theories about why we age is ‘antagonistic pleiotropy’, where pleiotropy is the technical term for genes having multiple functions in different tissues/times of life/contexts. The idea in this case is that a gene that causes an animal to grow up and reproduce a bit more quickly and efficiently will be passed on even if it goes on to cause late-life deterioration, because by the time you’ve made it to late life you’ve already passed on your genes (hopefully more quickly and efficiently thanks to the antagonistically pleiotropic gene variants you carry) so you’ve achieved your purpose, evolutionarily speaking.
It’s ‘harder’ for evolution to get rid of a gene that also has a small advantage than one that is merely bad for aging (or schizophrenia) and has no counterbalancing advantage, so actually the 3 explanation seems more likely than 1, though both exist on a continuum to some extent.
I'm not sure I buy this model. If there were a million independent genes that each increased your risk of schizophrenia by an absolute 1-in-a-million chance, and everyone had an independent 50/50 chance of having each gene, then it would indeed be hard to filter that out through evolution, but there also wouldn't be such a thing as being genetically prone to schizophrenia, because every single person on earth would have a chance of getting schizophrenia between 49.5% and 50.5% no matter how lucky they got with their genes.
The problem is that many small independent effects don't necessarily add up to a large difference in effects between people - a large difference between the worst possible and best possible outcomes, yes, but not necessarily between the 1st percentile and 99th percentile.
The fact that some people *can* be predicted to get schizophrenia at a much higher rate just on the basis of their genes means there's something for evolution to easily select on - if those people stop reproducing, the total amount of schizophrenia genes in the population will materially go down (because the ones evolution removed had, by assumption, like twice as many of those genes as normal). This assumes a model where these things are basically additive, but that seems roughly right to me.
>The scare-mongering here has to be false - that is, it can’t be bad to choose an embryo at the 50th percentile of schizophrenia risk rather than the 99.9th, because half of people are at the 50th percentile of schizophrenia risk and nothing bad happens to them
I think the worry is that if we either learn to genetically engineer or do embryo selection for long enough, we could pick people with 0% schizophrenia genes (which wouldn't happen naturally) and only figure it out 20 years later when we have a new generation of weirdly uncreative adults. I don't take this fear totally seriously, but it does imply we should be pretty careful with gene selection if/when we do get it to avoid intense selection pressure on everyone doing it at once
"based on the evolutionary argument above, I doubt this one"
I think the evolutionary argument also applies to (3), not just (2).
Amateur question: when you select embryos based on the presence or absence of a specific gene, is there a risk of inadvertently selecting for unrelated traits due to gene correlation? Is there such a thing as gene correlation? For instance, in choosing an embryo with a low risk for schizophrenia, might we unintentionally favour one with a higher risk for heart attacks? This would not be because the same genes cause both conditions, but rather because embryos with genes reducing schizophrenia risk might also possess genes that increase the risk of heart attacks.
Pleitropi is very uncommon.
I struggle with how to think about traits that are related to how people treat you. Physical attractiveness is clearly highly heritable. Does that mean that random people smiling at my oldest child when she was a baby is a heritable trait? Surely a lifetime of people being nice to you for no reason has a big effect on someone, and if you did a gene-based study you would almost certainly find 'people are nice to you' is highly heritable, but it's a clearly environmental factor. How is this controlled for, if at all?
Regarding hypothesis 2: across several cohorts in different countries, having a higher polygenic risk score for schizophrenia is positively correlated with having an artistic profession and with measures of creativity. https://www.nature.com/articles/nn.4040
Free full text on ResearchGate: https://www.researchgate.net/publication/277889916_Polygenic_risk_scores_for_schizophrenia_and_bipolar_disorder_predict_creativity
In Theory 1, a gradual decline in prevalence may have been hampered and slowed in past times by schizophrenic "scary bosses", whose symptoms of occasional sudden morose suspicion and paranoia, possibly leading to unpredictable violence, may have helped them maintain dominance through fear. And (male) chiefs in ancient times tended to monopolize women and have lots of children.
Arguably this applies more to hypothesis 2, but with not so positive creativity in the form of menacing cunning.
I think no. 1 is the correct answer. The mathsy way of expressing this is "Nearly Neutral Theory", which says that just knowing the "selection coefficient", which is a number for how deleterious the mutation is, is not enough. You also need to know the effective population size, because individuals don't evolve - populations do. In species with high population sizes (e.g. bacteria) slightly deleterious mutations are eliminated more quickly. The lower the effective population size e.g. humans, the more likely that "nearly neutral" mutations are invisible to selection. If an allele isn't eliminated, the only other option is that it eventually becomes fixed i.e. the mutated version becomes the new normal, even though it was a slight downgrade.
https://en.wikipedia.org/wiki/Nearly_neutral_theory_of_molecular_evolution
Now add in, at least among certain groups:
(1) birth control and
(2) other newish financial/social/economic pressures to delay childbearing into one's 30s
I speculate that if (1) is correct, these factors should quickly and massively decrease the incidence of incapacitating early-onset schizophrenia, particularly in males, who often express it by age 20 or so.
By "quickly", I mean within a few generations? So, would be an interesting and somewhat controlled experiment but also a long one.
I think the case for 2 is pretty strong, especially with the framework of something like the Diametric model which argues for the partial integration of both schizophrenic and autistic traits along a spectrum. If this model is correct it would explain the ubiquity of both autistic and schizophrenic traits across populations.
https://doi.org/10.1038/s41380-022-01543-5
Huh, so 1 million years ago at the dawn of humanity, Schizophrenia rates were through the roof?
Did civilisation only progress recently because selection effects had finally pushed Schizophrenia rates low enough to permit functional societies?
Reminds me of that "The Bicameral Mind" book, it'd explain a lot of their analysis of history if everyone used to be way more Schizophrenic
But we started with schizophrenia being evenly distributed through the human race, so there are no low-schizophrenia groups to be found. Or might there be very small (family-sized? village-sized?) groups with high or low schizophrenia rate?
I don't think this explanation is true for intelligence and height.
Selection is proportional to the additive heritability on the absolute scale, which makes your explanation true for schizophrenia: If there is a lot of liability-scale heritability, but the heritability is due to many variants of little effect and the prevalence is low, that translates to very little absolute-scale heritability, and it is true that evolution would have a hard time removing it.
But for height and intelligence, the tiny effects happen on the absolute scale, which means that if they have a strong relationship to fitness, evolution would be very quick at changing them. Like I guess it wouldn't exactly lead to eliminating the variants, but it would move you into a region with diminishing returns to them, making the evolutionary aspect less relevant.
I'm a postdoc working in bioinformatics, with a focus on cancer & polygenic scores but with a background in (mathematical) evolutionary theory. In my experience there is a surprising amount of misinformation - or just lack of knowledge, depending on how you see it - in medicine and genomics because even many experts are so used to monogenic risks since it was all we could feasibly find for a long time that they forgot that polygenicity is most probably the norm. It's arguably the core of the modern synthesis that happened in the early 20th century. At that point, mendelian inheritance had been proven, but we could see in many traits such as height that there was instead a continuous variation. This was explained by Fisher in 1918 by large numbers of small-effect loci, see "The Correlation between Relatives on the Supposition of Mendelian Inheritance" (he didn't use the word polygenic, but it's the same concept). So we're currently mostly just retreading ground that has been covered a literal century ago. There is a lot of followup papers on this and closely related topics, such as mutational burden, muller's ratchet, etc. that all show how negative fitness small effect mutations - which are the most common mutations to begin with - can stay in a population for a long time and even become fixed.
"Evolution hasn’t had time to remove all of them yet. Because a gene that increases schizophrenia risk 0.001% barely changes fitness at all, it takes evolution forever to get rid of it. And by that time, maybe some new mildly-deleterious mutations have cropped up that need to be selected out."
This does not make sense as a story. It is not harder for evolution to remove 100 genes of small effect than it is for it to remove 1 gene of large effect; the returns to selection is controlled ONLY by (narrow) heritability and NOT by the concentration across genes. Once you know the heritability, you know how easy it is for evolution to increase/decrease the trait; further knowledge of whether it's one gene of large effect or many of small effect does not tell you anything else about response to selection!
It is, of course, possible for some unfit genes to remain in the population due to the fact that they are constantly reintroduced via mutation. I'm not saying purely-bad genes are impossible or anything. I'm just saying that "many genes of small effect" does not have any explanatory power for why the genes were not selected out.
"Most random mutations are deleterious" is so oversimplified high-school biology. Many random mutations that we know about are deleterious, because that's how we know about them. Most random mutations are completely neutral in their effect, being either silent mutations (where there is no change to amino acid sequence of resulting proteins) or in non-coding regions. For the ones that do have an effect, we pay attention to bad things but not to improvements*, so we are more likely to be unaware of the beneficial* effects of random mutations.
*And all of this chatter about "more fit," "advantageous," "improvement," "beneficial" is dependent on the environment. What is advantageous in one environment may be an extinction-level trait in another. There is no such thing as an "evolutionary mistake"--there's just a trait that may not have been selected for yet.
> this study of ancient hominid genomes finds that schizophrenia genes are getting less common over time
Wild speculation: this helps explain the explosion of ancient religions as opposed to the relative dearth of new gods. Far more people heard voices in ancient times, attributed them to the divine, and boom.
I recall in an earlier post someone did bring up schizophrenia genes of larger effect. In the face of selection that's possible with rare de novo mutations.
Hmmm. Genetics doesnt always work in the ways we expect.
By analogy: when we were scampering around on all fours, a few of us had various genes for smaller nimbler front legs and feet. Some were able to walk around a bit on just our back legs. Only for short bursts. And at some serious costs in terms of backache, risks of falling over, loss of speed, etc.
And yet... this apparently rubbish set of aberrations came together and paved the way for hands ... thumbs... and a species that is able to dominate the planet.
So i suspect that individual genes that facilitate schizophrenia are often positive in ones and twos, but they cause a problem when they all crop up in one individual.
> So many of the traits we’re most interested in - intelligence, strength, schizophrenia, etc - are 𝘯𝘦𝘤𝘦𝘴𝘴𝘢𝘳𝘪𝘭𝘺 massively polygenic, because one side of them is better for fitness than the other. If they were monogenic, evolution would have already selected for the good side, and there would be no remaining genetic variance.
This is wrong. You can easily make the case for schizophrenia, but for strength and intelligence we want to draw the 𝗼𝗽𝗽𝗼𝘀𝗶𝘁𝗲 conclusion - it is definitely not the case that one side has historically been better for fitness than the other, because the amount of variation within the population is great enough that selective pressure would have had no difficulty producing an obvious response.
(Compare schizophrenia, where variation within the population is extremely low, which is what you'd expect from selective pressure against it.)
If you're having trouble with the idea that more strength could be worse, Steven Pinker put the point pretty well (though he was talking about intelligence) when he said, in my paraphrase from memory, "People present the evolution of humanity as being driven by selection for larger brain size, with increased intelligence as a happy side effect. But this is absurd. Metabolically, the brain is a pig. Any selection on brain size alone would surely have favored the pinhead."
4.) Schizophrenia is mostly a spandrel. I have no papers or examples, just another option.
It's a shame to cite Liu et al. 2019 without synthesizing one of the more interesting hypotheses in that paper. Granted, the paper itself is dismissive of it. It says:
“Although it has been reported that fertility among relatives of patients with schizophrenia is increased, a large cohort study and meta-analysis identified that this increase was too small to counterbalance the reduced fitness of affected patients (Bundy et al., 2011; Power et al., 2013). In fact, MacCabe et al. (2009) showed that patients with schizophrenia had fewer grandchildren than in the general population, demonstrating that the reduced reproductivity persists into subsequent generations.”
So the authors of this paper note that non-descendant relatives of schizophrenia patients have increased fertility but don’t find this fact interesting enough to incorporate into their overall interpretation.
Yet there is a way of interpreting that information that strikes me as painting a considerably more compelling picture: the decline in schizophrenia-dispositive SNPs does not presage the eradication of schizophrenia, but is part of an equilibrium-seeking adjustment to the group-level predominance of such SNPs.
This pattern is observed for left-handedness. It is advantageous to be right-handed in a right-handed world: it aids cooperation; tools are made for you. But in conjunction with other traits, it is also advantageous to be left-handed, specifically because left-handedness is rare. The word “sinister” comes from a root meaning “left” for this reason: lefties were looked down upon, but also hated, because their left-handedness gave them particular advantages—to wit, the element of surprise in combat—as a result of which left-handedness was never eradicated despite its obvious cost.
Could schizophrenia-dispositive SNPs not have a similar role? I’ll put my cards on the table and say I believe many of them do—specifically, that they contribute to shamanic and prophetic tendencies which have a positive kin-selective effect. I'd be remiss not to cite my friend Drew Schorno as helping me form this view: https://arcove.substack.com/p/null-call
The credit also goes to you, Scott, for helping me understand how schizophrenia-dispositive traits can be marginally beneficial. This was in the context of predictive processing: https://slatestarcodex.com/2016/09/12/its-bayes-all-the-way-up/
Switching to a metaphor I find easier to work with: schizophrenia-dispositive SNPs, or at least some subset of them, have their effect because they raise the effective temperature of cognition: long-range connections, seemingly meaningless “accidents”, fail to be ruled out immediately as they are in a neurotypical mind. The extra “heat” means that people with a lot of schizophrenia-dispositive SNPs think a lot more wild, nonsense thoughts. This is usually not good for the individual but may be very valuable for their community, because every once in a while their wild thinking leads them to notice something very important like the hidden ill intentions of a neighboring tribe or signs of an impending drought.
Schizophrenia itself is a failure mode: the cognitive temperature is so high that the community can’t keep the lines of communication open (although—note that pre-industrial societies were a lot less likely to medicalize and ostracize people with schizophrenia-like symptoms). But schizophrenia-adjacent “disorders”, “personalities”, whatever, do appear to serve a social purpose.
This is consistent with the results that Liu et al. cite, that relatives of schizophrenia patients have some level of increased fertility while the patients themselves do not. At the same time, the net effect on the kin group is negative, which is consistent with the view that the prevalence of schizophrenia-dispositive SNPs will continue to trend downwards. I suggest that the end-point of this will not be eradication of those SNPs but a new equilibrium.
"This seems less like the sort of thing that happens naturally, and more like the sort of thing you would claim if you wanted to make your theory untestable."
Can you point to where Torrey explicitly writes this in his article?
Wouldn’t possibility #3 show up on genetic correlations between schizophrenia and other health states? From what I’ve seen reported in research studies, most genetic correlations of schizophrenia are with other diseases, including other forms of psychopathology, but also things like cardiovascular disorders and immune disorders. So #3 seems unlikely. Some studies have found positive correlations with educational attainment, but recent studies have shown that this is explained by the genetic overlap between bipolar and SZ, and that unique SZ genes had a negative effect on educational attainment.
A (very likely?) possibility that can have an outsized effect on our evolutionary understanding is that schizophrenia is like weak ankles or some such -- it's a latent thing which isn't usually a problem until it gets nurtured with a sledgehammer or similar. In the past the other genetic effects would have been excellent for fitness except in rare cases, but in a world where weed (or, I dunno, television) is freely available it becomes vastly less adaptive.
First and foremost, I would like to thank KW, for this article. KW has way with opening our minds to conversations and greater positive awareness on different topics.
As this article presents the genetic material of this matter. What about the randomness of the schizophrenic people we know, ie: one of my non-biological grandmothers sister, due to her insidious and behavioral thought patterns that altered her, was diagnosed clinically, was institutionalized, yes, we still have a medical facility for the diagnosed disease, “Greystone”. But speaking with my mother who is not related, stated, my aunt and uncles are also diagnosed with the same irrational disease. Interesting, right?
What I know as schizophrenia, has a stigma this is DID and MPD, disassociated identity disorder, and multiple personality disorder. This is how we speak of schizophrenia. In all of my studies I have realized, it has the possibility of genetic traits, (genetics saying it has to do with the 6th chromosome and 22, now, not in the 50’s)
My question to this is, how long has this truly been genetically studied? The environment in which we were reared, has evolved us or mutated us into something we are not.
One great example of this is “Sybil”. She grew up in the 50’s, was the product of a very abusive childhood, in part, due to her religious upbringing. (7th Day Adventist). In the 50’s when her story came to being, as she had 16 MPD’s, which grew a wide spread fascination by the medical community, psychiatrists and psychologists lobbied to have MPD or schizophrenia, to be put in their “bible; DSM or Diagnostic and Statistical Manual”, this rare diagnosis led to a common diagnosis. “There were less than 200 cases in Western civilization” in or around the 50’s. “But after the book and film in the 1970’s, her case sparked hundreds of thousands of prognosis’s, and by the late 80’s there were 40 thousand diagnosed case in the US alone.”
Does this make a case for genetics and environment? I think so.
For example, in 1986 (date coincides to the diagnostic psych manual for MPD), Billy Mulligan, he had over 24 different personalities. He was diagnosed young, said he was a psychopath, chronic liar, and a murderer. He was seen by several psychiatrists put on Thorazine, a common drug for MPD. He stated he was becoming worse due to the publicity, society, and medication.
This subject has always fascinated me. I see in everyday situations, like a “narcissistic/pathological liars”, people who make up lies, and believe their lies and grandiosity, has become part of the mutating brain from all of the trauma. Right? Slightly altered personality is considered either schizophrenic or MPD.
Ok, there are two examples of cognitive behavior. This leads to, if not monitored, Schizophrenia/MPD/DID. These different behaviors change the brain, people actually perceive their beliefs and morphs them into their own monsters or demons, your preference.
Ted Bundy, Charles Manson, even though these guys are murderers, they heightened their own perception/personality (MPD) to believe what they did was ok, and fought for it. True or not? I do not believe we inherit a gene for their delusional beliefs/thoughts through our society.
Thank you again, KW, great food for thought.
Hey, nice series. I ran your article through Tutor GPT and wrote about it on my Stack! It's complicated stuff, I appreciate the insight. We are working on making diagnostics for polygenic diseases so this is super fun to think about. As an aside, have people run a subset of these through a markov model to simplify the gene-only interactions? then you look at the other end of it and look at a PCA and see which of these genes are migrating over time to see what we're selecting for . if its really .1% one way or another every generation over 50 generations we'd just see the probability clould jiggle for a thousand years
"Because a gene that increases schizophrenia risk 0.001% barely changes fitness at all, it takes evolution forever to get rid of it. And by that time, maybe some new mildly-deleterious mutations have cropped up that need to be selected out."
I don't understand the second part of this. Evolutionary pressures don't operate on one mutation at a time, do they?
>The scare-mongering here has to be false - that is, it can’t be bad to choose an embryo at the 50th percentile of schizophrenia risk rather than the 99.9th, because half of people are at the 50th percentile of schizophrenia risk and nothing bad happens to them. Schizophrenia genes can be at best fitness-neutral;
I think that you're conflating two different notions of 'bad' here.
It might be desirable to select an embryo with traits that are not maximally-reproductively-fit in the evolutionary environment, but are useful in modern humans.
And given that we're talking about variance in mental traits here, that even seem like a pretty likely scenario to me.
I'm rather sure it's a combination of 1 and 3. Assuming that mutations occur almost randomly, the ones that have an advantage as well as a disadvantage will tend to be selected against (or for) more slowly. And there are a lot more ways to break something than the improve it, if it's any good at all.
I.e., the average mutation that is detrimental will be selected against more strongly if it doesn't have any advantages. And there's no reason to believe that the advantages will be related to the disadvantages, except through a VERY tortuous chain of interactions. (And, of course, almost all mutations are either neutral or detrimental. Neutral drift theory enters here importantly, but whether the "neutral mutation" is absolutely neutral is very difficult to demonstrate. E.g. some yield the exact same proteins, but make the RNA slightly less stable. Just try to prove whether that's advantageous or not.)
Additionally, some mutations are advantageous (or neutral) in some environments, and not in others. (E.g. being tall and thin is disadvantageous in really cold climates, but advantageous when the temperatures are higher. [But not too much higher. I think the advantage disappears when the average highs are over about 95F, though that's a wild guess.])
It is not at all surprising that complex traits are polygenic, for a different reason than the evolutionary one described here. We have only 20,000 genes; roughly speaking, 20,000 "ingredients". How do you get from this the staggering complexity of form and function we observe? Through combinations of genes acting together. You can wire together 3, for example, to form a time-keeping oscillator (something I had my class act out a few days ago). Just as it is silly to ask which one is the oscillator gene, it is silly to expect one, or even a few "schizophrenia genes." Even height has ~1000 determinants. The idea of one gene = one trait is incorrect and sadly widespread.
Low confidence speculation here, but I have often gotten the impression that schizophrenia is sort of the opposite of autism. I wonder if the low schizophrenia risk people have high autism risk and vice versa, with "healthy normal" being a balance between the two.
The problem of polygenicity is not exclusive to schizophrenia. Indeed most intractably difficult modern diseases are thought to be polygenic, and for the same reasons. I studied asthma in grad school, where half my lab was arguably genetics driven.
Rates of asthma have been dramatically increasing for decades (including severe asthma requiring hospitalization, so you know it's not just over diagnosis). If that's the case, shouldn't we be focusing on some environmental factor?
Enter "Danger Theory" and the Hygiene Hypothesis. The immune profile of allergies, asthma, some kinds of IBD, and other auto-inflammatory and auto-immune diseases often looks like an inappropriate activation of the immune response to a parasitic infection (an inherently polygenic process). What's that all about?
The thinking goes that humans throughout history have been exposed to constant parasite challenge, but that this has gone down dramatically in the modern era. Different people have a different 'threshold' for immune activation of a parasite, with some people allowing more infestations and others activating against the slightest threat. Why doesn't everyone activate at high levels? Well, activating the immune system is inherently destructive, so activating inappropriately can have negative side effects. Those effects are worth it when you get the activation right, but harmful when you're wrong.
Until two hundred years ago, having an immune system on high alert to parasites would have been evolutionarily advantageous (most of the time). High threshold people would be sick a lot more, since they'd allow more parasites in, while low threshold people would be out working. In today's society the dynamic is exactly the opposite, since most parasite activations are going to be wrong and therefore harmful.
If baseline immune activation is more of a rheostat than an on-off switch, it makes much more sense for this to be controlled at a population level by variable or polygenic mechanisms, than by a single mutation that 'causes' some specific baseline level of immune alertness to parasites.
Multiple studies have demonstrated genetic components to many of these immune-related diseases, but there's reason to suspect that the changing human environment/culture/society has shifted the definition of what is adaptive versus maladaptive. What was selected for yesterday may be selected against tomorrow, and that is true across millions of years of evolution. In other words, environment determines whether a gene confers fitness or not.
What is natural selection, if not long-run genetic responses to environmental stimuli? Without environmental inputs, evolution has nothing to direct it. Therefore we can expect that dramatic changes to an organism's environment will result in some adaptations becoming deleterious - especially at a population level.
What holds true for immunology almost certainly holds true for psychiatry as well. Human interactions have shifted dramatically over the past few hundred years, to the point where the human brain is operating in an environment significantly different from the training data.
(Alternately, maybe this is all just cope to avoid having to explain that we don't understand the underlying mechanisms? Even so, I think the polygenic case still holds for most human diseases I've looked at.)
I think 1% of people are at 50th percentile risk, not 50%.
Your previous argument showed that dropping the extremal 1% barely affected the next generation, but collapsing the distribution probably will have significant long-term effects (hopefully positive!).
Scott is right and Torrey is wrong in regard to these two objections. Evolution can fail to remove bad effects even if they are monogenic, or low polygenic, if they also have some fitness advantage. Two examples:
- Sickle cell anemia is monogenic but persisted due to its advantage regarding malaria.
- Hyperlipidemia with elevated ApoB drives cardiovascular disease. It is polygenic, but not greatly. It persisted because it was adaptive for energy use during period of food scarcity, such as the Ice Age.
Intuitive arguments about fitness often go wrong because we are naive about the many varied things that contribute to fitness.
Isn't there a negative correlation between height and longevity? That is, shorter people live longer on average and have more time to contribute to the well being of their offspring. That alone would exert powerful selective pressure towards moderation in height across generations.
"The scare-mongering here has to be false - that is, it can’t be bad to choose an embryo at the 50th percentile of schizophrenia risk rather than the 99.9th, because half of people are at the 50th percentile of schizophrenia risk and nothing bad happens to them."
This is normatively true, not best-of-all-possible-worlds true. Neurotypes have the prevalences they do because, approximately, that was about as common as worked best in ancestral environments. A trait being the midpoint or most common doesn't mean it's the best; it would be unreasonable to say people at the 50th percentile of intelligence probably don't have any problems that aren't also faced by people at the 99.9th percentile of intelligence. Humans are "about as smart as made for the best tradeoffs", which is less smart than ideal.
I don't think the neurotype carved out as NT is the best of all possible worlds, or that it should be "the highest-prevalence by a juggernaut huge margin". Of course, it has various benefits (e.g. acting on plans rather than theorizing about them forever, big-picture thought, caregiving, etc). A world with no primary care physicians would be a disaster. A world with no MBAs...uh...okay, we'll circle back to that one. I don't want a world where all the smart people are primary care physicians or MBAs, though.
Here's an idea for a possible fourth explanation, but with the caveat that I'm not a biologist and I'm open to arguments if this is implausible. I've thought about this every time someone brings up the "but homosexuality must have some evolutionary advantage or it wouldn't exist, something something gay uncles" line.
Suppose you have the job of manufacturing parts to a certain tolerance. The more exact your production line, the more expensive the process - basically once you've calibrated the machines to get the average part about right, the smaller you want the SD of your output parts' measurements to be, the more it'll cost you. You can trade off between two ends of a scale. At one extreme, make your process good enough that every single part that rolls off the line is within tolerance (with the most expensive machine ever). At the other extreme, make parts where those in some range like half a SD from the mean are within tolerance, with the cheapest machine you can bodge together, measure each part that comes out and trash (or recycle if you can) those that aren't. Either way, every part that makes it to the customer meets the spec. Of course, the optimal point is a balance between the cost of the materials (including how easy they are to melt down and start over) and the cost of getting the assembly line to a certain precision; if you're cutting natural diamonds then you're at the opposite end of the spectrum to if you're 3D printing plastic.
Maybe evolution is doing the same? If the optimal height of humans in some environment is 160cm, and an evolved mechanism for making everyone exactly that high even if some grow up better nourished than others would have more of a fitness or energy cost than making everyone grow "close enough", I'd expect evolution to take the latter path.
The old "group selection" post (Studies on Slack) mentioned that in on a planet that gets hit by a solar flare once an eon, species will not evolve a radiation shield if that costs slightly more in energy for each individual during all the years when there's not a need for it. Similarly, if schizoprenia / homosexuality / autism / pick your favourite condition is a net negative from a reproductive fitness perspective, even if a biological mechanism is available to filter it out and evolution would have the time to do so, I'd expect the filter not to exist if the cost it imposes on the individuals (such as extra energy expenditure) is in some sense higher than the reproductive fitness cost of not evolving the filter.
Evolution has, after all, managed to create frogs and fish that have thousands of children and 999 of them generally get eaten or otherwise die before sexual maturity, which is the "cheap machine" end of the scale (which I believe is called r/k selection in biology textbooks).
Is it plausible that schizoprenia exists because the evolutionalry cost of it not existing would be higher than otherwise? Without group selection? Or have I missed something obvious here?
The idea of "schizophrenia genes" is counterintuitive and not particularly useful. And, since schizophrenia is a mental dysfunction it is natural to look for an association with other brain mediated attributes such as intelligence or creativity.
But perhaps that is like looking for your lost key under the street light when you actually dropped it across the street.
There is a lot of data that shows an association between schizophrenia and every aspect of the immune system. A highly dysregulated immune response is a frequent comorbidity in schizophrenics and a promising area of research.
Further understandings on the interplay of the different arms of the immune system, the immune cell types implicated and their origin, the role of immune responses to common viruses, the gut brain immune axis, and aspects of inflammation and autoimmunity, are bound to shed a great deal of light on the causes of schizophrenia.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9082498/#:~:text=It%20has%20been%20shown%20that%20schizophrenia%20is,from%20innate%20to%20adaptive%20immunity%20and%20from.
> The clearest way to resolve these questions would be to genetically engineer someone to zero schizophrenia risk and see what happened (this is beyond current technology)
Just checking, but this is also beyond current ethical boundaries, right? Right???
Studies show that the polygenic risk score for schizophrenia is not associated with creativity, indicating limited contribution of the accumulated effect of these risk variants to creativity.
However, neural inflammation and certain viral infections are strongly correlated with the development of schizophrenia, as is infection with the parasite Toxoplasmosis.
Perhaps the correlation between schizophrenia and creativity is due, in part, to shared vulnerability factors in brain architecture and function, including neural hyper-connectivity, novelty salience, cognitive disinhibition, and emotional lability.
Thus, individuals with both a genetic predisposition to specific differences in brain wiring and function, AND also having an abnormal immune response to infection by certain viruses (Epstein barr, CMV, retroviruses, etc) or parasites (Toxoplasmosis), are statistically far more likely to develop schizophrenia.
It's like rolling snake eyes in a game of chance. A creative brain is only a risk when combined with a dysregulated immune response to specific but commonly encountered pathogens.
No doubt there are other pathways to the development of schizophrenia and it's all more complex than what we currently know. Future research will undoubtedly reveal far more about it.
Viral infections correlated with the development of schizophrenia:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10302918/
T. gondii as a risk factor in the development of schizophrenia:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6382596/
Accumulated effect of schizophrenic and creativity risk factors:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6792478/
I'm not clear why we're rejecting #2. Possibility #1 seems like the default hypothesis, and #2 sounds reasonable to me, in a way like "the dose makes the poison". #3 is similar, but requires that the gene influence at least two unrelated systems, so it seems less likely to happen than #2 which just requires a small effect on one system. So just on intuition I'd guess a mixture of #1 and #2 would be most likely. But this is really far from anything I know well.
There are some crucial things I don't get about how polygenic risk works. Can somebody here explain a few basic things? So there are thousands of genes that cause schizophrenia, sprinkled all over the genome. So let's say there are 5000 genes that cause schizophrenia, each of which raises risk by a tiny amount, on average .02%. So does someone have to get all 5000 of them to be schizophrenic? But the chances of someone getting all 5000 of them is minuscule! Is it more that you have to get, say, at least 2000 of them. So what are the chances of someone getting 2000 out of the 5000? If these genes are scattered all over the genome, then the chance of getting any one of them is not correlated with the chance of getting any of the others. (I think. Is that right?). So then it should be possible to calculate the odds of someone getting,say, 2000 particular genes. If you take any one of these genes, what fraction of the population has it? Does one assume it's 50%? If you know how likely someone is to get any one of the genes, and approx how many genes there are that increase risk of schizophrenia, it should be possible to figure out what fraction of the population will be schizophrenic. And the result should be approx. 1%, since that what fraction of the population is. I'm sure I'm not the only one who'd like some help with understanding these basics.
Seems plausible to me that some of the genes that raise the risk of schizophrenia are genes that cause the person to have a harder life. You can think of them as genes that affect the individual's environment. For instance, physically attractive kids are better liked by peers and seen as smarter by teachers: They literally get more smiles. Just now looked up research on schizophrenia & physical attractiveness. There were studies that found photos of schizophrenics were rated as less attractive by judges than those of non-schizophrenic peers, but I think that result might be accounted for by schizophrenics being more likely to have poor grooming and and other signs of self-neglect, and a less pleasant facial expression. But then found one where the high school photos of people who later became schizophrenic were rated as less attractive. It's herer:
A. Farina et al (see record 1978-23202-001) investigated the relation between mental illness and physical attractiveness and found that female psychiatric inpatients were less attractive than normal controls. The current study extended this investigation in 2 ways. First, 28 psychiatric inpatients were compared to 3 separate control groups of 53 low, middle, and high socioeconomic status Ss. Mental patients were judged significantly less attractive than either middle- or high-income controls but were not significantly different from low-income controls. Second, to examine physical attractiveness prior to hospitalization, attractiveness ratings of the patients' high school yearbook pictures were compared with ratings of the adjacent same-sex photographs. Patients' photographs were judged significantly less attractive than their peers' even in high school. Findings suggest that being physically unattractive may predispose an individual to a number of negative social outcomes, one of which is mental illness.
((Napoleon, T., Chassin, L., & Young, R. D. (1980). A replication and extension of "Physical attractiveness and mental illness." Journal of Abnormal Psychology, 89(2), 250–253. https://doi.org/10.1037/0021-843X.89.2.250)
I realize this is somewhat off topic, but I recall from my high school biology class that researchers had isolated a gene for manic-depression. In fact two different studies, one focusing on Iceland, and the other on Pennsylvania Amish, had located two different genes, on two different chromosomes. Is this still considered true, or has it been discarded in the intervening years?
If still considered true, it would indicate that at least some mental disorders have single-gene causes.
"why are there still even these genes of very small effect?"
Explanation 1) "Evolution hasn’t had time to remove all of them yet" seems to assume that it's harder for evolution to remove a lot of small cumulative genes than one single gene. I don't believe this to be the case.
To get an intuition, I quickly wrote a small python script which evolves a small population with 1024 binary genes under two different conditions:
a) a 50% probability of death if gene 666 is True
b) a 50% probability of death if the 666th fourier component of the genome is >0.5
(I didn't implement recessive or dominant traits)
Both factors seem to have approximately the same half-life, even though one is a single gene and the other is a bunch of genes with small effect. I'm not posting the script right now as there is a high probability that I've made some embarrassing error, since I wrote it in 10 minutes and didn't do much testing, but the result seems plausible. Unless there is something special about the 666'th fourier component, I'd guess this to be true of any linear transform component.
Dunno as an excellent Genetics graduate and one who has been told that he is exceptionally creative (humble brag), and one who initially self-diagnosed and then was formally and objectively diagnosed with schizophrenia, I really, _really_, hope that there is some exceptional if rare boost to having this condition.
The alternative is too drab.
The big risks I see of embryo selection are Goodhart's law and out of sample extrapolation. It's possible for gene A to be correlated with B in an unselected population while still having it be a bad idea to exercise extreme selection pressure on A, either because the correlation breaks down under selection or because their are non-linear effects when taken to extremes or because there are other unknown side effects.
Those aren't issues for any plausible near-term IVF selection, but could cause problems in some of the more extreme proposals (selecting from thousands of embryos or even straight out genetic engineering.)
Man, you can't leave this topic alone, can you...
...oh wait, wrong poly. Well, it probably explains that one too.
About the image for this piece: It's a little known fact that lions suck the schizophrenogenic genes out of their prey. For a lion they're like Reese's peanut butter cups.
Chris Masterjohn has been thinking along the same lines, but came to a different conclusion. According to his logic (link: https://open.substack.com/pub/chrismasterjohnphd/p/unlocking-performance-and-longevity) most SNPs analyzed are just noise and there are only a few important ones, but scientists need a mechanistic (ie biochemical) model of how they interact, not brute force statistics…
Quote: Common polymorphisms by definition on average produce the average, because they are defined by their deviation from normal. Rare disease genes on average produce a rare disease, because they are defined by their pathogenicity. Natural selection minimizes the presence of harmful genes, which makes the rarity of a gene correlate to its severity. Thus, the biggest problems have to be the few rare problems, not the many common variations.
The individuals who display a disease phenotype because they are heterozygous for more than one severe defect in related pathways have been termed ‘synergistic heterozygosity’.
I have a friend with early-onset schizophrenia who considers it "a superpower that you have to work hard and learn to deal with." In other words, I think he'd take issue with a blanket statement about it decreasing fitness. Having seen his reflexes and pattern-detection in action I would find it hard to disagree; he's survived situations that would have killed someone slower on the react or less inclined to observe everything and notice connections between things. In other words, less paranoid and ready to act accordingly. Compared to him, most people are just not paying attention. Perhaps the increases in fitness for the people who are able to deal with it outweigh the decreases in fitness for those who can't. After all, we don't necessarily know who all is schizophrenic. It's not only the people who go get medicated about it. And antipsychotics decrease fitness in and of themselves, so there's a selection bias there.
None of that is to say that anything you said about the geneticity of it is wrong. His family has quite the history of it.
(The odd thing about talking about human genetic evolution is that it's probably over. Assuming we *don't* kill ourselves and *don't* halt all interesting tech advancement, we will probably mostly move away from this form the way have we mostly moved away from caves, adobe houses, and horse carts. In the interim, knowing about evolution will help us to develop patches and also to understand how our brains work. But the following is about natural human DNA evolution as if it were a going concern.)
Genomes evolve to evolve, and without going into the ways they manage that, but antropomorphizing: the overall genome and organism design gets selected toward *some balance* of dependable to "speculative". If we wanted to let human evolution continue by its own processes, it would *not* make sense to fix a lot of genes that contribute to schizophrenia, because then the results of lots of combinations of those and other genes would not get tested. In other words, if we assume our genome is at about the right place on the risk/reward curve, adjusting for less risk for individuals in the very next generation would mean fewer beneficial advancements on many-generation time scales. It's probably even harder to identify the meta-evolutionary mechanisms than the functional effects of current gene prevalences on odds of schizophrenia, but evolvability-tuning can be seen in simple situations and it makes sense to assume it happens in general.
I say this to add another spin on what function all those potentially harmful genes might serve. For instance, in the last couple millions to thousands of years, it may have served our genes well to invest in a lot of brain-tech startups. I believe (sorry forget source!) that humans have a particularly high ratio of brain-specific to other genes.
Another spin is that different *distributions* of characteristics in populations not only effect the population-level success but also affect the context in which individual genes can be beneficial or detrimental. Sheer diversity can be beneficial if the body, brain and social context can accommodate it.
There is also the possibility of non linear effects, where you only get a bad effect if you have, say, A, B, and C, and where any one or pair of them does nothing. That would further make it difficult for evolution to get rid of these, as in most people the effect is zero
I’d highly recommend everyone listen to Robert Sapolskys lecture on schizophrenia that covered exactly this topic.
https://m.youtube.com/watch?v=nEnklxGAmak&t=1359s&pp=ygUWU2Fwb2xza3kgc2NoaXpvcGhyZW5pYQ%3D%3D
The TL:DR proposal he has for why genes that lead to schizophrenia got selected for is that they can make people more prone to magical thinking and intuitive/creative ways of perceiving things that a lot of cultures historically in people playing the roll of shaman or spiritual leader. In a lot of these cultures these people weren’t expected to be celibate and were highly valued so the traits got passed on.
If you’re trying to survive in the wilderness and believe in spirits of the dead and forces of nature, you value having someone who thinks they senses things you can’t and who makes strange connections. We know that people related to schizophrenic people tend to be into conspiracy theories or occult beliefs or any kind of interest where you make connection between things other people normally wouldn’t. Someone obsessed with making connections that seem spurious but interesting feels like a summary to a lot of ancient mystical thinking imo. It’s not surprising people prone to that would be valued & I also would worry that we might be losing something of value trying to engineer those traits out of existence.
Quick note, you can't assume that because the people with the lowest skitzophrenia risk don't have other problems it follows that there is no such effect of these genes.
1) Its possible that those individuals have other compensating genes. Maybe one genetic heritage has some other set of beneficial genes that give the same benefit with another downside and other groups have the skitzophrenia promoting genes (you'd want to see if some mixed individuals do much better in average).
2) Its possible that the benefits aren't addative. Skitzophrenia is rare so what if each gene that increases risk offers a small benefit to those who have it but you don't get much extra benefit (any?) for having multiple genes that increase risk. This differs from your 3.
3) What if the 'benefit' offered by these genes is in utero or in pregnancy-- say they reduce head size. If they increase the chance of being born at all or not dying looking at those who didn't die tells you nothing.
4) Other really complex interactions between genes could be taking place.
Possible support for possibility #2: https://www.nature.com/articles/nn.4040
Excellent discussion! Just to add to point (1), I think it’s important to remember that selective pressures act via *reproductive fitness. To my knowledge, the average age of onset in schizophrenia is near late-adolescence to early-adulthood. Considering that the average age of first-time mothers was 21 as recently as the 1970s, it seems totally reasonable that we might be overestimating the negative selection effects of schizophrenia, simply because most people throughout history likely began having children before they saw the full effects of their condition. I see this as similar to how many of us carry increased risk factors for cancer, heart disease, etc. that simply don’t have any effect until we’re past our main reproductive years. Clearly, there is some selective pressure acting here, or else there would not be a decline in schizophrenia over time, but my point is that it’s probably weaker than we would expect for other genetic/congenital conditions which act earlier in life.
I think more intelligent people have fewer children, so not a slam dunk for a positively selected
I have skin in this game. We recently went through IVF and we had our embryos tested for polygenic disease risk. Our "best" embryo had significantly reduced risk for a wide range of diseases, except schizophrenia, where it was 85% percentile. We had other embryos with low schizophrenia risk, but higher risks of more common diseases. As you might imagine, we thought about this very carefully.
Ultimately we chose to implant the "best" embryo. Partly our reasoning is that 85% percentile of schizophrenia is still <2% absolute risk. But another part of our thinking was that there's some evidence of antagonistic pleiotropy, i.e. that higher schizophrenia polygenic risk score correlates with desirable traits.
This paper (https://www.nature.com/articles/s41467-018-05510-z) for example finds that:
"Higher educational attainment (EA) is negatively associated with schizophrenia (SZ). However, recent studies found a positive genetic correlation between EA and SZ. We find strong genetic dependence between EA and SZ that cannot be explained by chance, linkage disequilibrium, or assortative mating. Instead, several genes seem to have pleiotropic effects on EA and SZ, but without a clear pattern of sign concordance. Our results reveal that current SZ diagnoses aggregate over at least two disease subtypes: one part resembles high intelligence and bipolar disorder (BIP), while the other part is a cognitive disorder that is independent of BIP."
There is also some evidence that polygenic risk scores for schizophrenia predict creativity: (https://www.nature.com/articles/nn.4040)
So I do think selecting for the lowest possible schizophrenia polygenic risk score is potentially a bad idea, unless you also have PGS scores for lots of other traits to ensure you're not inadvertently selecting against them.
The embryo we picked is now a happy little baby crawling around as I type this. Hopefully we made the right decision!
Conceptualizing a creativity gene is in unnecessary.
People live longer
X cost of raising children increasing
X easily available birth control
= children born later in life
Rate of SNP mutations are proportional to age of father at conception (between 20 & 40 years paternal mutation rate triples, not the maternal rate). This is perhaps ordinary entropic change of the paternal genome which was revealed by non-genetic changes in longevity due to global health and nutrition at population levels.
If you were to graph US average age of father at conception and US rate of schizophrenia and autism you should be able to see the effect of aging on accumulated genetic defects, completely independently of a “schizophrenic gene” which would be selected against strongly of course.
Unsurprisingly the rates of autism and schizophrenia have been rising in recent decades. Likewise Schizophrenia is more prevalent in males than females, understandably if it were possibly Y-linked, paternal influenced. Finally schizophrenia should be growing fastest in world regions where lifespan is showing the most rapid improvement, and paternal age of conception. I would hypothesizes it is most pronounced in central and subsaharan Africa.
Genes which favor non reproduction are easy to maintain equilibrium in a population if they enhance familial reproductive success even at the cost of individual offspring. This has long been discussed and modeled. [Dawkins, E.O. Wilson] A gene group which enhances longevity will enhance schizophrenia and autism without any biochemical relationship to the specific source of the root issue.
Consider disruption of NMDA receptor interaction with multiple modulatory transmitters - failure to attenuate sensory prediction error; Schizophrenia can be induced chemically by disrupting this system; schizophrenics are strikingly resistant to illusions normally created by imprecise perception; schizophrenics cannot involuntarily predict and track motion with their eyes. First-degree relatives also have eye tracking dysfunction but not full-blown schizophrenia. It only takes a few more SNP’s induced by age to cross the line. Epigenetic aging mechanisms induce it.
I’m gay, and have no interest in having kids. This at an individual level selects against reproductive fitness as a gene or group. However the frequency of gay men in the population is relatively stable. What produces non-reproducing offspring may enhance the reproduction of siblings indirectly, without biochemical signature.
Also, “last brother being gay” effect is driven by late maternal age of conception, whereby maternal antibodies to testosterone developed during sequential prior male pregnancies (even miscarriages) alter the process of testosterone imprinting on the fetal brain starting at 9 weeks. Another case of what appears as a possible gene effect is due to epigenetic factors related to aging, health care and nutrition at the population level.
This article had many comments, apologies if any of this is a rehash.
I believe most evolutionary theory and thinking is skewed towards considering humans as individual organisms. While of course our bodies are discrete organisms, when it comes to mind, we are necessarily linked. The brain is, among other things, a social organ. Mine, at least, seems quite busy and consumed with understanding how my larger social world works, its power dynamics, and what I need to do to have social value. What if schizophrenia is polygenic with many traits that map out and model meaning structures and complex social dynamics?
We are fundamentally complex social primates. So complex, in fact, that we cannot live alone. By the time all of us turn 12 (it is extremely rare for schizophrenia to manifest before this age), our families and tribes have nursed us, fed us, taken care of us, and bonded with us at no small sacrifice. Homo sapiens as I understand us would not lightly stop taking care of a kid that hears voices, often until past the age of reproduction.
H. sapiens evolved as tribes. We keep our elders around, for their wisdom and knowledge is a strong fitness advantage - for the whole tribe. We take care of each other, to a degree that is not rational.
Now I'm going to stereotype. Would love to hear others' data points.
The people I know who struggle with schizophrenia seem (anecdotally) to be above average when it comes to sensitivity and spirituality. I wonder if there are linkages between schizophrenia and mystics or shamans. As I observe mind and try to locate "self," I readily get lost in a sea of connections, social conditioning, culture, and my group's foundational myths and narratives - to a degree that might make me less functional, but also possibly mystical in my inner modeling of complex networks of human meaning, purpose, attention, and care.
I would posit that schizophrenia is polygenic and related to a host of genes that assist in the mapping and modeling of complex human social and meaning structures, and would require extremely fine tuning to select out without resulting in asociality or sociopathy.
Or maybe I have just watched *A Beautiful Mind* too many times?🤷♀️