Wednesday, May 4, 2016

Genetics in psychiatry - hope or hype?

This blog was inspired by discussions I have been having with Allen Frances and also partly in response to some blogs he was written about the role of genetics in psychiatry. He is, I think it's fair to say, highly skeptical that genetics will be of much use in psychiatry, as he discusses here. His jaded view stems partly from the relentless hype which accompanies a lot of announcements of large-scale genetics projects or of their results, and also from a perceived lack of progress of these efforts in explaining genetic risk for psychiatric disorders. With the rapid pace of developments in psychiatric genetics, it is worth taking a beat to consider what we know and to try and separate the hope from the hype.

Psychiatric disorders are highly heritable. Twin and adoption studies have demonstrated conclusively that differences in risk for these conditions are at least partly, and in many cases largely, due to genetic differences. Indeed, psychiatric disorders are far more heritable than conditions like heart disease or cancer. That is not to say that genetics explains all the risk, but certainly enough of it to make it worthwhile understanding the underlying biological mechanisms.

Genetic risk for mental illness overlaps clinical boundaries. Having a relative with a diagnosis of say, schizophrenia, increases an individual’s statistical risk not just of that disorder, but also of bipolar disorder, autism, depression, ADHD, epilepsy, intellectual disability and many others. Moreover, specific mutations often manifest in diverse clinical diagnoses across individuals. From an etiological perspective, the diagnostic categories in the DSM thus do not represent natural kinds.

Most of the genetic risk does not lie in common variants. This should not come as a surprise. Any genetic variant that increases risk of psychiatric disorders should be strongly selected against because such disorders greatly increase mortality and reduce fecundity (number of offspring). While genome-wide association studies have identified some, indeed many, common variants associated with risk for conditions like schizophrenia, or shared risk across disorders, these collectively explain very little (~7%) of the overall genetic risk. (Even techniques such as genome-wide complex trait analysis suggest only about 25% of genetic risk for schizophrenia is tagged by common variants. That is, if you take estimates from such techniques at face value, which I do not, for reasons I discuss here). Common variants likely contribute to modifying effects of genetic background but clearly do not explain disease risk by themselves. The oft-repeated line that disorders like schizophrenia are “due to the cumulative effects of many variants of small effect” is thus not supported – at least this is not a complete description of the genetic architecture.

Many cases are caused by rare mutations. It has been known for decades that some specific genetic syndromes, such as Fragile X syndrome or velocardiofacial syndrome (now called 22q11.2 deletion syndrome), convey very high risk of psychiatric disorders, such as autism or schizophrenia. It was thought by many that these conditions were somehow exceptional and not relevant to the remaining cases of idiopathic and non-syndromic cases, which were believed, on rather flimsy grounds, to have a very different genetic architecture. It turns out that, far from being exceptional, those long-known genetic disorders are perfect exemplars of the genetic architecture of psychiatric conditions. The development of chromosomal microarray technologies and whole-exome or whole-genome sequencing has led to the discovery of dozens, indeed now hundreds, of similar rare genetic disorders that predispose to very high risk of mental illness. More and more such disorders are being recognised as additional cases are sequenced. Clinical categories like autism or schizophrenia are, from an etiological point of view, umbrella terms encompassing hundreds of rare genetic disorders.

Sequencing can reveal rare mutations. Identifying such rare conditions will become easier as more and more cases and controls are sequenced, giving us greater power to discriminate pathogenic mutations from the background of rare mutations that we all carry. These efforts have only just begun but we are already seeing them yield results. Already 20-25% of cases of autism and ~10% of cases of schizophrenia can be ascribed to a primary genetic mutation – a huge increase in diagnostic yield from just a couple years ago.

Genetic causation is complex. Describing these as rare genetic disorders is not meant to imply simple genetic causation. All the known mutations can give rise to a wide variety of effects in different individuals and are often carried by clinically unaffected people. The pathogenic effects of particular mutations can be modified by other genetic variants that any individual may carry, whether common or rare. There is nothing unusual in this scenario, however – even the most classic Mendelian disorders, such as cystic fibrosis or Huntington’s disease, are subject to modifying effects from other genes. Specifically, we can expect that more severe cases, with earlier onset, will be more likely due to single genetic mutations, often arising de novo, while less severe cases, with later onset, will include a bigger contribution from inherited mutations and more interplay between multiple mutations in any given individual. 

Non-genetic factors are also important. The variation in phenotype often observed between monozygotic twins also shows that non-genetic factors – such as stochastic processes of brain development or later experience – can have large effects on the phenotypic outcome associated with any given genotype. These observations place important limits on the ability to genetically predict disease risk in anything other than a probabilistic fashion. 

Genetic diagnoses are still useful. Even with the complexities mentioned above, it will still often be possible to identify a primary causal mutation in individual patients. (See here for a more detailed discussion of inferences of genetic causality). Such diagnoses are immediately useful in many ways – they give a definitive etiological diagnosis that complements the often more fluid diagnoses based on symptoms; they allow clinicians to group patients and define new syndromes; they empower patients and their families to help drive such efforts; they inform on genetic risk to subsequent offspring; in some cases they may already give information on likely responsiveness to treatments. Just as importantly, they give biological entry points to dissect the underlying pathogenic mechanisms and hopefully develop new treatments.

Genetics is just the first step. There is a regrettable culture of hype around much of science these days, driven by the need to promise near-term translational impact to secure funding. This has certainly been the case in the field of psychiatric genetics. It is crucial to recognise, and, in my opinion, to state publicly, that identifying pathogenic mutations is just the first step in what will be a very long journey to develop new therapeutics.

The drug discovery model that has worked so well for disorders like cancer will simply not work for most of the rare disorders causing psychiatric illness. In cancer, the pathogenic effects of mutations arise at the cellular level – they directly affect the processes of proliferation and differentiation that drive the disease. Identifying a primary genetic mutation can thus directly implicate a particular biochemical pathway as a suitable drug target. Even when that is the case it is still a hugely difficult task to actually develop a new drug that works.

For psychiatric disorders, the difficulty of that task will be multiplied many-fold, because the answer will not come at the level of molecular biology or biochemistry. We must figure out how changes at those levels lead to alterations at the level of neural circuits and systems, which emerge only via cascading and indirect effects through the complex and dynamic processes of neural and cognitive development.

Predictions of new treatments in the short term should thus be tempered with a good dose of humility. Genetics will not suggest new pharmaceutical approaches by itself. What it will do is enable more insightful neuroscience by providing crucial entry points to elucidate the underlying pathobiology. This is especially true for high-risk mutations which can be modelled directly in cells or animals to elucidate the pathways by which mutation of specific genes lead ultimately to neural circuit and system dysfunction.

Genetics is thus the crucial first step on what will be a long journey to better understand the causes of mental illness. Indeed, not only do I think a genetic approach will be productive, I think it is the only thing that will be. Certainly nothing else has worked – psychiatry has made essentially no progress otherwise over the last sixty years or more. Maybe that’s a bit harsh, but we have certainly had no biological insights that have yielded new drugs with new mechanisms of action over that time-frame.

A key reason why is that we have had no way to dissect the cryptic heterogeneity of these disorders. Genetics not only illustrates that heterogeneity but provides the means to distinguish patients based on underlying etiology. This will entail a real paradigm shift in psychiatry – from treating all patients with similar symptoms as monolithic groups, to recognising that such symptoms can have very diverse causes and grouping patients instead on the basis of genetic etiology.

Thus, while I agree with Allen Frances that the attendant hype around genetics in psychiatry is misplaced and unhelpful, I am much more optimistic than he is that this approach will pay off. These are genetic disorders. Their inheritance may be complex but is not infinitely so or intractably so. We can and will make progress – but it will take a lot of subsequent research to turn that progress into new treatments.


  1. Appreciate your final points. One needn't back off the severe strategic weakness of psychiatry's tendency to avoid biology. You're merely embedding psychiatric imbalance in its context, and stating that there are clear, new potential handholds. I'm confused what it even means to be 'skeptical', in that sense. .5% more clarity, 40% more clarity– who the fuck cares what the number is? It gets us off the dime. There's no doubt a there, there. If one must be bloodless, fine: there's billions of dollars involved on the back end of this business we're messing up at the moment. Skepticism is best left to the context of framing the null.

    I am guardian of a bipolar son who has the usual blizzard of highly idiosyncratic personality variations, combined with the grindingly predictable tendencies of many other bipolar 1's, the overlaps with schizophrenia, the impulse control/addiction elements, etc. The emphasis on mystery and individual drug experiments is still understandable in 2016, but it won't be in 2025. One apparently has to sit on the wrong side of the desk to fully appreciate how lame and incurious the field is; how quickly the prescription pad comes out; how urgently we need even incremental understanding and improvement.

    Re your excellent point regarding the influence of wiring on personality: operationally, it seems highly likely that we're dealing with a dynamical system, in that small differentials send us all off on developmental arcs of great import. 'Stochastic' captures randomness and diffusion processes adequately, but I'd submit it's not a likely candidate for what's going on, even in nonlinear incarnations; not as an explanation for half of ourselves. Certainly there are dampening and offsetting effects in the sense you allude to, via random inputs, and there are random walks; but nonlinear effects that are driven by sensitivity just make sense intuitively, in that wiring would seem to involve stepwise alterations of great import. Framing development in terms of nonlinearity is a much clearer way to intimate the large size of potential effects, and helps explain the size of our 'dark energy' problem of other-than-heritable personality.

    I do agree that wiring may explain much, even most of the missing causation, but there's little reason to suppose that non-traumatic events, especially when fetal and very young, can't manifest similar patterns of sensitive dependency. Your vision of what's going on seems too exclusively a mean-reversion/random one, and that seems partial to me. I'm in psych, and raised a lot of kids; I see sensitivities everywhere, from adoptive parents to friends to well-placed scoldings. Less about the bing-bong of billiard balls, and more about art, and the lack of it.

    If anything could create nonlinear alterations in wiring, these rare mutations would. If nonlinearities drive either wiring or environmental alterations of personality, a significant amount of personality variation may well be undetectable as genetic (for now), even if in fact purely genetic. I suspect that a relatively finite set of dimensions embed an attractor of a given mental health condition (and/or lack of the condition), including mostly genetic variables. Framing mental disease as a dynamical system of perhaps limited dimensionality offers hope that one can parse it. That should lend urgency to the argument to chase the biology involved.

    1. Scott - thanks very much for your thoughtful comments. You are right - there is a there there! Psychiatric disorders are clearly strongly genetic in nature - figuring out how will not be a waste of time or resources.

      I agree with you about nonlinearities in development - we know they exist and can have important effects on the outcome. That means that development is sensitive to small random variations that are intrinsic to the organism, not coming from the outside. (That's what I mean by stochastic, above). This is an important and often-ignored source of variance in many human traits, including whether people develop psychiatric disorders. By the time they are born, the brains of monozygotic twins are already quite different from each other.

      Of course development does not stop at birth and any initial differential sensitivities can also be amplified by further nonlinear processes of ongoing development, in response to experience and life events (as I also allude to above).

      That means exactly what you conclude - much of the variance in psychological traits and also in risk of psychiatric disorders lies in the "non-shared environment" term.

      I also think that the characterisation of brain states as attractors is a very useful way to think about pathophysiology, as discussed here:

  2. Kevin, thank you for sharing this compelling blog post. I’m not a neuroscientist but I do have one of those families plagued with “neuroweirdness”: ADHD, migraine, synaesthesia, bipolar II, anxiety, depression, Asperger’s, panic disorders, phobias, Tourette’s, narcolepsy, PTSD and at least a few more conditions. Your point that “clinical categories like autism or schizophrenia are, from an etiological point of view, umbrella terms encompassing hundreds of rare genetic disorders” elucidates what I’ve always suspected about my own family; clinical categories both compartmentalize mental illnesses and miss the point…there’s something profoundly complex and dynamic that underpins our neuroweirdness. I’m curious to see what the field of developmental neurobiology will reveal about families like mine and the genetics of mental health. And, I’m hopeful that future treatments will move beyond mere symptom management, which seems to be the best psychiatry can offer at this time. CC Hart

    1. Thanks CC! "Complex and dynamic" sums it up very well!

  3. Seems very long winded and if its this technical what will it be like in fifty years. Being scizophrenic or schizo affective ( depends on what symptoms im displaying on the day i was informed ) i get disheartened by the lack of simplicity. I discovered through addiction recovery mental health issues in my immediate family which i was completely oblivious too. The impression i got from the rhetoric i read was basically im a freak and the genetic pool is as well. I think its intellectual gamesmenship as the language used sounds very elitist. Why the annalytical mumbo jumbo but i once heard a shearing contractor say that your most productive years shearing are your early thirties. So is anyone trying to shear a incredibly wrinkly sheep with know style at all. The more blows you use the longer it takes to get around the sheep. If you can relax concentrate on your pattern and keep a flat hand the less demand it places on you. Mind you shearers can be competitive. And the discourses ive just read, will just put down to being to wrinkly for me and find sustenance in the fact acceptance can lead to a life of peace,purposefulness and tranquility too. Best of luck too you all

    1. Apologies if that was overly technical. It was more intended for professionals in the field, addressing an ongoing debate on how best to make progress in understanding the causes of mental illness. For what I hope is a more accessible exposition, you can see this talk: (it's long, though!)