Is Traumatic Brain Injury "Caused By Genes"?
the psychiatric genetics control group is out of control
It’s certain at this point that genes play a role in most aspects of someone’s psychology, including the parts that overlap with the DSM. Things like personality are in large part genetic, and your mental health susceptibility loads on your personality — some people are more likely than others to be diagnosed with any mental illness, and some are more likely than others, in that subset, to be diagnosed with particular mental illnesses. Through polygenic studies (which look at the interactions of many genes with individually small effects), we can talk about people having “more” or “fewer” genes linked to e.g. schizophrenia. We know people with particular genetic arrangements (“high polygenic risk”) are more likely than the baseline1 to develop schizophrenia, and their counterparts (“low polygenic risk”) are less likely.
Does this mean, per Scott Alexander’s recent article series, that schizophrenia is “mostly genetic” or “caused by genes”?
We know polygenic risk plays a meaningful role in schizophrenia. We know the monozygotic2 twins of people diagnosed with schizophrenia are more likely to develop it, but not massively so — concordance is seemingly in the ballpark of 20% to 33%, implying both genetic and non-genetic contributions. (For comparison, MZ twin concordance for autism is >90%, with variance loading more on severity than “autism y/n”.) We know that copy-number variants — duplications or deletions of several genes in one region of a chromosome — are more common in people diagnosed with SZ than the general population, but this is disproportionately true for the subset of people with very low IQs, and even the “SZ-associated CNVs” are much more associated with early developmental problems (e.g. intellectual disability, physical abnormalities) than they are with SZ.
So, overall, genetics look substantial-but-not-exclusive. Still, those polygenic risk scores are powerful. As Scott Alexander says, you could call schizophrenia a “genetic disorder” on the strength of them alone. Would this be in any way a reasonable thing to do?
“Genome-wide association study of traumatic brain injury in U.S. military veterans enrolled in the VA million veteran program” by Merritt et al. (2023) is a beautiful study — the control group for psychiatric genetics.
Traumatic brain injury is the most hard-environmental thing there is. No one spontaneously develops TBI. Anyone who suffers a sufficiently severe head trauma will have a TBI; you could take fifty people and hit them over the head at high speed, and their “polygenic risk score for TBI” wouldn’t save any of them. Lots of things cause TBI, and lots of factors influence their severity and long-term outcome (including many genetic ones). TBI prevalence goes up when new environmental risks are introduced (e.g. cars) and down when they’re mitigated (e.g. not letting people drive drunk without seatbelts).
Merritt et al. (2023) calculated polygenic risk scores for TBI, based on comparing people with and without TBIs and seeing how they differed. From a technical perspective, this study was the gold standard. Its sample was huge (~111k TBI patients, ~193k controls), easily compared to one another (all military veterans, basically all of who suffered their injuries during service), and had tons of data available (biobanks are great). They replicated their findings in FinnGen, another huge biobank covering a nontrivial percentage of the Finnish population.
They found 15 genes strongly linked to TBI. Just like any polygenic study of mental health, these genes were overwhelmingly those expressed in the brain, especially the cortex (and also, uh, testicles). They had fairly typical correlations for mental health-related genes, including neuroticism, multiple mental illnesses, and a negative correlation with intelligence:
As the authors concluded: “This first well-powered GWAS of TBI identified 15 loci including genes relevant to TBI biology, and showed that TBI is a heritable trait with comparable genetic architecture and high genetic correlation with psychiatric traits.”
None of that is wrong! They concluded exactly that, by the standards and assumptions used for any other psychiatric genetics study. It’s just that there may be some minor problems with those standards and assumptions.
Because traumatic brain injury is obviously environmental, people don’t study things like “concordance rate of TBI in families” too often. Some studies, though, do this by accident.
Fullerton et al. (2019) aimed to study the personality development of young people who experienced head injuries in childhood or adolescence, by comparing head-injured twins (both monozygotic and dizygotic) to both their normal co-twins and normal controls. They recruited 1,241 twins, with demographics reflective of children in the area they drew from (Los Angeles), and followed them up for a decade. Around half the twins were monozygotic, half dizygotic.
One small problem: the twin concordance rate for head injury was 72%.
To be clear, both monozygotic and dizygotic twins were about as likely to be concordant or discordant for head injury. You don’t see this in things with a large “genetics-only” component. You would probably see a higher MZ concordance if you actively studied it, though — you do so for other things with an obvious huge environmental role, like PTSD. (Fun extra credit assignment: compare the concordance rate of PTSD in monozygotic twins (30-40%) to that for schizophrenia.) This is also a crazy high number, and would presumably not show up in the general population or in an exclusively severe TBI sample.3 But wow, it’s sure something, isn’t it?
I have little doubt that if we studied TBI how we study other disorders, with complex family pedigrees, adoption studies, etc., we would find a meaningful “genetic component”. We would find TBIs are more common in people whose parents, children, siblings, or extended relatives have had one. We would find higher MZ than DZ concordance, because MZ twins have more similar personalities. We would find, even, that the adopted-out children of people with TBIs are more likely to have TBIs than people with no family history. Take a look at those “TBI genes” — they’re genes for impulsivity and drinking too much. As the joke goes, the most common last words are “hold my beer and watch this”. Personality is genetic, and if you’re the kind of guy who likes drinking and taking risks, you probably have a fair few of those in your family tree. You’re also more likely to have a TBI than a teetotaler with an obsessive safety complex.4
What else would we find? Probably a link between TBI and CNVs.
As the chart shows, one of the big “genetic correlates” of TBI is ADHD polygenic risk score. This also shows up in real-world studies — people with an ADHD diagnosis are more likely to have a TBI than the general population. Accordingly, “genes more common in people diagnosed with ADHD” and “genes more common in people with TBIs” should overlap.
CNVs are more common in people diagnosed with ADHD. This is true even for people without the more severe problems common in the CNV population, such as intellectual disability, though I haven’t seen a study that breaks down risk-by-IQ in the >70 range. The CNVs more common in this population are the same ones more common in people diagnosed with ASD or SZ. As the first of these studies concluded, “[o]ur findings […] suggest that ADHD is not purely a social construct”. Well, sure, by Merritt rules.
What is ADHD?
But “ability to concentrate” is a normally distributed trait, like IQ. We draw a line at some point on the far left of the bell curve and tell the people on the far side that they’ve “got” “the disease” of “ADHD”. This isn’t just me saying this. It’s the neurostructural literature, the the genetics literature, a bunch of other studies, and the the Consensus Conference On ADHD. This doesn’t mean ADHD is “just laziness” or “isn’t biological” – of course it’s biological! Height is biological! But that doesn’t mean the world is divided into two natural categories of “healthy people” and “people who have Height Deficiency Syndrome“. Attention is the same way. Some people really do have poor concentration, they suffer a lot from it, and it’s not their fault. They just don’t form a discrete population.
—“Adderall Risks: Much More Than You Wanted To Know”
ADHD is the diagnosis people receive if they have some sufficient combination of high impulsivity/low concentration/weirdness that isn’t obviously a dfferent childhood diagnosis/being male/being poor. None of this is to say that no one with this diagnosis experiences grief from the characteristics that brought them there. Concentrating is great, not being able to concentrate sucks.5 But if you’re talking about ADHD, you need to be able to acknowledge this — nothing about the ADHD literature or the real-world context of diagnosis makes sense if you don’t. People with ADHD are at high risk of TBIs because they have traits common in people at high risk of TBIs (high impulsivity, low concentration) and disproportionately share demographic factors (being male, being poor).
Impulsivity and concentration are modulated by genetics, so you can talk about someone’s “polygenic risk score for ADHD”, and many people with high ADHD PRS will be diagnosed with ADHD. Many people with high ADHD PRS will have TBIs, because these same traits increase TBI risk. People who are high-impulsivity and low-concentration because of a CNV are more likely to have TBIs, in the same way they’d be if they were those things without a CNV.
Indeed, they might be even more likely. People with CNVs have lower IQs on average, even in the mildest subset of the CNV population — people with no neurodevelopmental disorders of any kind who have survived to middle and old age in reasonable health. This synergizes nastily with risk-taking. With lower premorbid intelligence, it’s harder to know if a risk is too risky:
Is traumatic brain injury “a genetic disorder”? Well, no. That’s insane.
We know this because we know exactly what causes TBI. But it’s possible, in the wild world of hypotheticals, to imagine the counterfactual. Say we somehow didn’t know, that there was some blinding factor causing us to not realize the obvious cause-effect.
We would study TBI, and we’d find a large genetic component, using every yardstick we use to measure large genetic components. We’d find it ran in families. We’d find particular genes that modified risk, and create polygenic risk scores for TBI, and predict people’s TBI likelihood based on these scores. We’d find rare CNVs that correlate with everything neurodevelopmental also correlate with TBI. We’d find strong associations between TBI and many other disorders, both before and after onset (these are known). We’d find the traditional negligible role of shared environment. There would clearly be a large environmental contribution — some people would get TBIs no matter their “genetic risk” — but it’d be less obvious, and besides, non-shared environment is probably in utero or random chance or something.
That is to say, we’d think of TBI as highly genetic in the same sense we do other mental disorders.
What would we come up with to mitigate it? Probably something like polygenic embryo screening.
From where we’re standing, we know what this would do. It would select against risk-taking, and reduce TBI rates only inasmuch as it reduces risk-taking on the population level. This is…not a good thing. Risk-taking is a beneficial characteristic for a society; people who are willing to take risks are more likely to get the high-reward outcome of high-risk/high-reward maneuvers, leading to great cultural triumphs (technological advancements that might not work, great art that might not be financially viable, radical discoveries that were previously fringe, etc). Even on a smaller scale, these traits can be personally beneficial (e.g. bets that pay off) and lead to subjective enjoyment (e.g. extreme sports). A society of people unwilling to take risks is stagnant and repressive.
Because we know more than our counterfactual counterparts, we can take steps to reduce TBI that don’t involve killing all the people who move society forward. Seatbelt laws and helmet laws signficantly reduce severe TBI rates. Exceptionally risk-taking people might be expected to ignore them, at least some of the time, but they consistently result in large jumps in how many people comply at any given time; how “normal” it is to ride in a car without a seatbelt or on a bike without a helmet is socially mediated, such that an equally risk-taking person will make different choices in different environments.
“There is no moral to this story.”
What is schizophrenia? “It’s a mental disorder where you become out of touch with reality —” I didn’t ask you to quote a patient pamphlet at me, I asked what schizophrenia is. There is a Thing-That-Schizophrenia-Is. What is it?
Here’s an interesting parallel: what is autism? “Autism is a spectrum,” then stammering, seems to be the common-enough response these days. There are a lot of disparate autisms, diagnostically — but they seem to combine to something coherently enough, given that uber-high twin concordance. When people are looking for the unifying Autism of this spectrum, they come to a few constant traits; paracosmic orientation (prioritization of the inner world above the outer world), significant environmental sensitivity (e.g. sensory problems), repetitive and “stereotyped” behaviours (stimming, special interests, etc), unusual developmental patterns (e.g. hyperlexia, speech/language delay without other developmental delays), and any number of other common factors. Not everyone will have every one of these, but you can spot similarities — patterns running in families, shared traits between people with very different “functioning levels”, a subclinical population who cluster with diagnosable people in practical terms.
What is the schizophrenia equivalent of this?
Hey, this sounds kind of like autism, doesn’t it? It also sounds kind of like…
Schizotypal personality disorder — the full cluster of as many of these traits as possible — is fairly common in people who later develop schizophrenia, and in people who have psychotic experiences that don’t fit full SZ criteria. It’s also common in families with “a history of schizophrenia”, which is how the concept was identified. One is reminded of Leo Kanner dutifully writing up the stories of the eleven autistic children he met — and, in the process, revealing half their immediate and extended relatives were autistic too. Its prevalence in the general population is unclear, but a large epidemiological study from the 2000s suggests around 4%.6
4% is very high! People push back when I suggest autism is that common, and, observationally, autistic people are more common than schizotypal people.7 If this is even in the ballpark of accuracy, there must be many more schizotypal than schizophrenic people. Let’s go up another level: there are subclinically-autistic people who have a lot in common with clinically-autistic people. What about subclinically-schizotypal people?
Measuring “schizotypy”, as a trait, is an area of interest for many psychologists. Schizotypal traits are, well, the ones in the STPD criteria. Some of them can be frustrating; I once felt vividly called out by a statement that “schizotypal people say they want close friends, but their asocial behaviour is inconsistent wth this”. Many are just based. People who score high on measures of “positive schizotypy” (the subset of schizotypy expressed in that axis, and by the “odd beliefs and magical thinking” cluster) are creative, insightful, have weird experiences that they like, and are more able and willing to break from convention. The world could do with more people like this. It emphatically could not do with fewer.
It’s fairly common for people to be high on schizotypal traits. The average score on scales like the one pictured above is not zero. On the Unusual Experiences scale of the short O-LIFE (pictured), the mean score in the norming sample was slightly above three; participants were older (average age 47), and younger people consistently score higher. The Schizotypal Personality Questionnaire, the most famous scale, had an average score of 26-27 (out of 72) in its norming sample. Unfortunately, its norming sample were Californian undergrads. 24ish seems typical for more normie groups — but if you take the largest possible gap there, and the smallest possible SD anyone’s found, you get about 0.27 standard deviations. The traditional “high schizotypy” cutoff on the SPQ is top 10% (40ish, depending on norms), and people in this range are very schizotypal. They also comprise, by definition, one in ten people.
Let’s recap. Schizotypy is common; most people are a little schizotypal, somewhere between 1 in 10 and 1 in 25 are very schizotypal. This is much higher than the proportion of people who are schizophrenic. Schizotypy has a whole bunch of upsides, and people like being schizotypal, much as they tend to like any sort of weird they are. The upsides of being schizotypal are greatly socially beneficial, and are things we as a species could do with more of rather than fewer. Schizotypy is a developmental thing; it’s a set of personality traits, which emerge early in life and are mostly genetic.
What is “genetic risk for schizophrenia”?
For some years now, people have been trying to identify schizophrenia before it develops. They’ve come up with a number of great screening tests to identify this population, studded with questions such as:
“Do you have strong feelings or beliefs that are very important to you,
about such things as religion, philosophy, or politics?”
“Do you daydream a lot or find yourself preoccupied with stories, fantasies,
or ideas?”
“Do other people tell you that your ideas or beliefs are unusual or bizarre?”
“Have you ever found yourself feeling mistrustful or suspicious of other people?”
“Do you feel you have special gifts or talents?”
“Do you usually prefer to be alone?”
“Do you think others ever say that your interests are unusual or that you are
eccentric?”
“Do people ever say your ideas are unusual or that the way you think is
strange or illogical?”
“Do you ever generally just feel unhappy for any length of time?”
I can’t think of any possible problems with prescribing people antipsychotics if they answer yes to these questions!
The method to this madness is that these are, in fact, all good gauges of someone’s schizotypy (and by extension “psychosis risk”). That doesn’t mean they’re bad things. Being unhappy is kind of lame, but a reasonable environmental response to, for instance, being told you’re schizophrenic because you’re a smart introvert who’s very into philosophy. The rest are all neutral to strong positive. I imagine if you’re reading this, you’re likely to answer yes to more of them than the average person.
Consider the implications of the following:
These are personality traits, which are mostly genetic, very similar between MZ twins, etc
These are “related to schizophrenia”
Schizophrenia has low MZ twin concordance, i.e. a large non-genetic component
If you identify “schizophrenia genes” as “the genes that cause schizophrenia” and select against them, such as through anti-SZ embryo screening, what happens?
You select against people who daydream, who get deep into ideas, who enjoy the unusual, who are confident in their gifts and talents, who break from convention, who come up with creative insights, who don’t judge their worth by social acceptance, who care about complex issues.
And maybe you reduce SZ prevalence a little. Not too much — people would still be out there getting traumatic brain injuries, after all.
“Schizophrenia genes”, like “TBI genes”, are “variants that are much more common in people diagnosed with SZ than the general population”. This doesn’t mean they’re “the genes that cause schizophrenia”, any more than TBI genes “cause TBI” without any other factors. Given twin discordance, the low proportion of schizotypal people who develop SZ, and the fact not everyone who develops SZ was ultra-schizotypal before, there are clearly other factors.
TBI itself is one — it increases the risk more in people with a family history of schizophrenia (= people who are probably more schizotypal), but not exclusively. Abusing particular drugs is another; heavy use of pure meth is really bad for you, including for people who don’t seem to have any particular schizotypal tendency, and even people who’ve abused other drugs at high doses long-term and would presumably have had any “real” inclination they have come out already.8 Birth injury (overlapping with traumatic brain injury) seems to play a role. Trauma seems to. There are probably many other complex factors. There are “neither schizotypy nor clearly environmental” factors, too — people with lower IQs seem more likely to both experience psychosis and have it become chronic if they do, which is unlikely to be caused by schizotypy.9
Schizophrenia research is in a similar position to autism research many years ago, when an unrealistically narrow impression led to large swathes of the autism spectrum being missed. Because schizophrenia onsets so late compared to autism, this hasn’t been overcome despite a decades-long head-start; developmental comparisons are too focused on things like nonspecific delays, and not on the whole profile of schizotypy. We can identify “schizophrenia genes” — trivially so, because people diagnosed with SZ have common traits (e.g. schizotypy) that we know must be genetic. But that doesn’t mean they’re “the genes that cause schizophrenia” any more than “the genes for personality traits that raise TBI risk” are “the genes that cause TBI”. Schizophrenia is not TBI, and has more uniquely-genetic contribution,10 but “schizophrenia polygenic risk scores” predict many more things than SZ alone. Many of those things are good, perhaps many of them better than the alternative.
The last of the three posts to which this is in part a reply makes a statement:
The scare-mongering here [that selecting strongly against “schizophrenia PRS” would be bad] 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 only true normatively. People at the 50th percentile of “having takes on religion/philosophy/politics”, “fantasy-proneness”, “unusual ideas”, “being confident in their gifts and talents”, and “breaking from convention” do fine for themselves. We’ve built our society around them. But I don’t want to be good enough. Average intelligence is good enough. Average health is good enough. I think we can do better. I don’t think many of us would trade brilliance for averageness if it might decrease some risks slightly. Why should we do it here? If we can be our best possible selves — both as individuals and as a society, a culture, a species — we’ll need to bring those traits along with us. They’re the only reasons any of us care about such an unusual idea as that, after all.
"How common is schizophrenia at baseline?” is a mess. Somewhere between 0.5% and 1% is accepted enough at this point; not higher, maybe lower. If you aggregate it with the most closely related psychoses (e.g. schizoaffective) it goes up, but probably not far past 1%. Note: this means only a tiny minority of people with high PRS for schizophrenia will ever be diagnosed with it. (If you have “the statistics gear in your brain”, you can work this out yourself, but it’s concerningly common for people to talk about schizophrenia PRS in ways that imply many more high-PRS people will develop it than is realistic.)
“Identical” twins are not quite identical, so it’s better to use the more technical term. I don’t think this is big enough to singlehandedly explain the relatively high discordance in SZ. It might play a role in large personality discordances (e.g. one SZ twin and one twin with very low schizotypy, one autistic twin and one severely allistic twin, one very gender-non-conforming gay twin and one very gender-conforming straight twin).
A problem through all these studies (which I’m shelving for now, because it should be its own post) is defining the borderlands of TBI and its effects. “Mild TBI” shades into “not TBI”. It’s not at all obvious how many of the long-term impacts of mild TBI are “organic” (caused by the injury itself) and how many are “functional” (not direct consequences of the injury). Adults with mild TBI have more severe long-term cognitive sequelae than adults with moderate TBI, which has no plausible organic mechanism. Because mild TBI is so much more common than severe, and people with mild TBIs are a lot more able to participate in studies, most studies of this sort are overwhelmingly mild.
Merritt et al. (2023) probably underestimate the personality (= genetic) impact on TBI, due to being restricted to military TBIs. There are clearly-enough personality factors in who gets a service-related TBI and who doesn’t, but the large share of real-world TBIs that occur under “hold my beer and watch this” circumstances, or other contexts like alcohol-fuelled assaults, are excluded. The controls are also veterans, and people who enroll in the military will have particular personality traits more or less often than the general population.
There’s a lot of additional characteristics that have been tagged onto ADHD in the past few years, mostly by people very demographically different from the populations most likely to be diagnosed (adults, often fairly well-off, much more female-skewed, lower impulsivity). This is tricky. Certainly these people are trying to describe a subjective distress, and find a reasonably acceptable language for it that the people around them can understand. If you listen to this too hard, though, you fail to notice the biggest thing the ADHD literature tells you, which is that it’s primarily a way of pathologizing young children who don’t act by middle-class norms. (This can be completely compatible with it having ‘real disorder elements’ that cause great subjective distress! More things in Heaven and Earth.)
A tricky part here is that while it’s clear enough what an “autistic person who isn’t schizotypal” looks like, a “schizotypal person who isn’t autistic” is less obvious. This is the last refuge of the imprinted brain/diametric model guys — they insist the very large overlap is all misdiagnosis, because it’s so hard to distinguish the two. In a clinical setting, actually, you can distinguish them perfectly well — the “autism signal” and the “schizotypal signal” are testable for separately, and predict different outcomes. In the real world, where “people diagnosable with autism” and “people not diagnosable with autism” blend into one another, the distinction is much harder.
Quinones ascribes the spike in meth-induced psychosis to a synthesis method that introduces too many contaminations, but Dynomight suggests it’s the opposite problem, i.e. unprecedentedly-pure meth at unprecedenedly-high doses.
I think there’s significant diagnostic bias in who’s “schizophrenic” vs “bipolar”, and apparent intelligence is part of it. “Bipolar” is a more socially acceptable diagnosis; people are openly bipolar, and the prognosis is better (definitely not from diagnostic bias, hey). “Schizophrenia” is an Outgroup Label. Someone who a given diagnostician can relate better to is more likely to be diagnosed with bipolar than schizophrenia, and intelligence is an axis in that relatability. There’s a “real” psychosis-IQ relationship in that bipolar people also have lower IQs, but things like “bipolar has an inverted J curve, schizophrenia doesn’t” are suspicious. (I think there’s also an element of “IQ is intended for neurotypical populations, and can draw weird results for weird people” that artificially deflates scores compared to an “impression” of someone’s intelligence, but this is not schizotypy-specific, nor does it explain the whole thing.)
I think it’s likely some of the “schizophrenia genes” are true schizophrenia genes, i.e. they raise the likelihood of a schizotypal person developing chronic psychosis. I don’t think we’d be able to distinguish these well from e.g. intelligence genes, which would also be more concentrated in particular directions in schizophrenic than schizotypal people. (“Animal models?” Oh yeah, you go work out from first principles why that mouse is acting weird.) I am deeply unconvinced we’d be able to narrow them down specifically in the first place — genome-wide associations work, individual genes get messy. I wonder what a hypothetical “true schizophrenia gene” would do in a simultypal person.
.I don't think this is anything too different from what I said in my post.
All genes are genes for something. Therefore, all genetic conditions will be conditions where the genes cause something that cause the disorder.
The two examples I gave were that cystic fibrosis, a classic genetic disease, is also a disease of lung mucus. And that if there were some virus where likelihood of getting it was 100% genetic, it would look like a 100% genetic condition.
I described this a bit in Argument 6B, "Genes could just be a proxy for some more satisfying cause of schizophrenia", and partly in the subsequent Open Thread, where I said that "part of my argument is that there may never be a satisfying unitary story of schizophrenia (any more than there will be a satisfying unitary story of what causes kidney disease), so instead of treating genes as the IOU for the satisfying story downstream of the genes, we should just go with the genes."
In terms of the schizotypy connection, the paper you cite finds many detrimental facets of negative schizotypy, but beneficial aspects of positive schizotypy (some of which I'm not impressed by; they seem to be things like "if they're in a cult, they're happier with the cult" and "sometimes the hallucinations say nice things to them"). But my impression is that in general, negative schizotypy is genetically correlated with schizophrenia, but positive schizotypy mostly isn't. See eg https://www.frontiersin.org/journals/psychiatry/articles/10.3389/fpsyt.2015.00143/full
"While positive schizotypy is the underlying dimension explaining psychotic features, it is not necessarily related to Krankheitswert. Negative schizotypy, however, appears closely related to schizophrenia regarding its heritability"
...and https://www.cambridge.org/core/services/aop-cambridge-core/content/view/47381AEC2205356A81390D3D214CF4B9/S0007125022001799a.pdf/relationship_between_polygenic_risk_scores_and_symptom_dimensions_of_schizophrenia_and_schizotypy_in_multiplex_families_with_schizophrenia.pdf
"These results provide genetic evidence in support of the spectrum model of schizophrenia, and support the view that negative and disorganised symptoms may have greater genetic basis than positive symptoms, making them better indices of familial liability to schizophrenia"...We further note that the association of SCZ PRS with only the negative dimension of schizotypy is in agreement with previous epidemiological findings that show familial predisposition to SCZ in the relatives of probands without a history of a psychotic episode is likely to be better indexed by the negative symptoms."
A simpler test of your thesis is whether schizophrenia risk is associated with creativity directly. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4590283/ looks at this and finds yes, but weakly. Given the polygenic scoring ability we actually have, someone in the 99.9th percentile of schizophrenia risk is about 0.2 percentage points more likely to have a creative job than someone at the median. Given hypothetical perfect scoring ability, it's 1.75 pp (these look larger as relative effects, 20% and 175% respectively). I'm not sure how much this actually reflects increased creativity compared to less suitability for other jobs and/or more interest in creative ones. I think profession in this study was something like self-rated; there are lots of schizophrenic-seeming people who think they're musicians, artists, and writers; but fewer who have convinced others.
It's implausible that any substantial portion of the population will be selected for low schizophrenia risk, because even if everyone in the population opts for polygenic selection, most people will be selecting for other things like low diabetes risk or whatever. The people most likely to select for low schizophrenia risk would be people in families with high likelihood of schizophrenia, or people who incidentally notice that one of their eggs has very high schizophrenia risk.
So the way I would think about this is that now when polygenic scores are weak, given the weak schizophrenia->creativity correlation, nothing we do to imputed schizophrenia risk can affect creativity much. Later, when polygenic scores are strong, we'll be able to identify embryos at very high risk to actually get schizophrenia, without lowering overall risk significantly on a population-wide level (and in fact we won't want to, since if the kid probably won't get schizophrenia we'd rather use that selection ability to prevent heart attacks or whatever).