The value of polygenic analyses in psychiatry

The last decade of genetics research in psychiatry (and in other fields) has been dominated by genome‐wide association (GWA) studies, in which common variants across the genome are tested for association with a trait or disorder. These studies have shown that polygenicity is the rule, i.e., psychiatric disorders are influenced by many (likely thousands of) genetic variants, each with a small effect1.

This is best illustrated by the flagship GWA meta‐analysis on schizophrenia, which is the first disorder that has achieved the sample size needed to detect the effect sizes that have been dealt by nature's hand. By analysing 37,000 cases and 113,000 controls, 108 associated regions were identified2. However, the significant variants together only explained 3.4% on the liability scale for schizophrenia, indicating there are many more variants involved. This high degree of polygenicity means that everyone harbours risk variants, but those affected likely carry a higher, and possibly unique burden of risk factors, which is fully consistent with the spectrum of clinical presentations.

Still, because of the small effect sizes, the usefulness of the results of GWA analyses has been questioned. In this paper, we show the value of the identification of genetic variants in psychiatric disorders and illustrate how analyses of GWA data have further advanced our knowledge, beyond the identification of associated genetic variants.

One major problem in psychiatry is that there have hardly been any new drugs developed in the last decades3. Even though effect sizes are small, significantly associated genetic variants can point to new drug targets, as shown for other diseases3. With 108 regions associated and no immediate knowledge about the functional effects of the far majority of the hits, further analyses are necessary, but could lead to new targets.

Functional annotation of genetic variants associated with psychiatric disorders using bioinformatic analyses is an active area of research4. This includes analyses that aim to explore which trait‐associated genetic variants are also associated with inter‐individual variation in gene expression levels, and gene‐based analyses investigating which biological pathways are enriched with genes harbouring associated genetic variants3. For psychiatric disorders, neuronal, immune and histone pathways are reported to be involved5, and these analyses will become more informative with new technologies, such as single cell gene expression studies.

GWA data can also be used to increase knowledge on the mechanisms underlying the frequent comorbidity within psychiatric disorders or between psychiatric disorders and other traits. This is interrogated by polygenic analyses, investigating the joint effect of genetic variants1, 4. Traditionally, to demonstrate a genetic relationship between disorders was difficult, especially for the rarer disorders, because recording of psychiatric diagnoses was needed on large samples of twins or families to demonstrate the increased risk of a second disorder in family members of those affected by a first disorder. However, direct measurement of DNA variants has allowed direct measures of genetic sharing using independently collected case‐control samples.

It has become apparent that psychiatric disorders not only share genetic risk with other psychiatric disorders, but also with somatic diseases and traits such as educational attainment6. If genetic correlations between disorders and traits are identified, a key question is whether the association reflects shared biological pathways (pleiotropy) or if there is a causal relationship. Using two‐step Mendelian randomization, it has been shown that cannabis initiation does result in a small increase in risk to develop schizophrenia, but that schizophrenia leads to a larger increase in risk for cannabis initiation7. More insight into directions of effect and causality can direct the development of prevention programs.

Such knowledge is also important for research aiming to develop treatments targeted to children at high risk that their disorder develops into an adult psychiatric disorder, either the same or a different one. A polygenic risk score is an estimate of the cumulative genetic risk of an individual. In schizophrenia research, polygenic risk scores have been found to predict various psychiatric traits during childhood and adolescence, indicating that genetic variants play a role in the transition from internalizing or externalizing symptoms during childhood or adolescence to schizophrenia later in life8.

These polygenic risk scores cannot be used as diagnostic predictors of psychiatric disease, as risk to psychiatric disorders is only partly explained by genetic risk factors, and, to date, only a small proportion of genetic risk has been identified. Nonetheless, out‐of‐sample prediction explains about 7% in liability to schizophrenia2, so those with highest polygenic risk scores have an increased risk approximately equivalent to having a first‐degree relative affected.

While this has little clinical utility in the general population, it may have clinical application in the context of prodromal presentation at a mental health clinic. Recently, an individualized risk calculator has been developed that with reasonable accuracy could predict the conversion to psychosis9. Predictors included were already existing symptoms and poorer functioning on cognitive tests. Possibly, risk prediction can be improved by adding further variables to the model, including, but not limited to, genetic risk scores10. Based on these profiles, individuals could be stratified into high and low risk groups for transition into a severe mental illness10 and the effects of different treatment programs for these groups could be tested.

Overall, the progress in genetic research has substantially increased our insight into the etiology of psychiatric disorders. Genetic discoveries in schizophrenia have been achieved by large sample sizes, and the current data show that, with larger samples, similar results can be obtained for other disorders. Genotyping technologies are no longer the limiting factor (500,000 DNA variants can be measured for less than $100/person). The limiting factors are availability of large samples with consistently measured clinical symptoms and environmental risk factors. International collaborations, such as the Psychiatric Genomics Consortium (PGC) (www.med.unc.edu/pgc) and the EArly Genetics Lifecourse Epidemiology consortium (EAGLE) (www.wikigenes.org/e/art/e/348.html), and long‐term planning are required for cost‐effective generation of the data sets needed to deliver on the promise of precision or stratified medicine in psychiatry.

The new genetic discoveries of the last five years are opening previously unknown avenues of research. If ultimately these lead to new treatments, as in other fields of medicine, these treatments could be specific to stratified patient groups.

Christel M. Middeldorp1‐3, Naomi R. Wray4,5
1Child Health Research Centre, University of Queensland, Brisbane, QLD, Australia; 2Child and Youth Mental Health Service, Children's Health Queensland Hospital and Health Services, Brisbane, QLD, Australia; 3Biological Psychology, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands; 4Institute for Molecular Bioscience, University of Queensland, Brisbane, QLD, Australia; 5Queensland Brain Institute, University of Queensland, Brisbane, QLD, Australia

C.M. Middeldorp acknowledges the Netherlands Organization for Health Research and Development grant “Genetic influences on stability and change in psychopathology from childhood to young adulthood” (ZonMW 912‐10‐020) and funding from the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska‐Curie grant agreement no. 721567. N.R. Wray acknowledges grants from the Australian National Health and Medical Research Council (1078901, 1113400, 1087889).