Advertisement

Specificity of Psychiatric Polygenic Risk Scores and Their Effects on Associated Risk Phenotypes

  • Amanda L. Rodrigue
    Correspondence
    Address correspondence to Amanda Rodrigue, Ph.D.
    Affiliations
    Department of Psychiatry, Boston Children’s Hospital, Harvard Medical School, Boston, Massachusetts
    Search for articles by this author
  • Samuel R. Mathias
    Affiliations
    Department of Psychiatry, Boston Children’s Hospital, Harvard Medical School, Boston, Massachusetts
    Search for articles by this author
  • Emma E.M. Knowles
    Affiliations
    Department of Psychiatry, Boston Children’s Hospital, Harvard Medical School, Boston, Massachusetts
    Search for articles by this author
  • Josephine Mollon
    Affiliations
    Department of Psychiatry, Boston Children’s Hospital, Harvard Medical School, Boston, Massachusetts
    Search for articles by this author
  • Laura Almasy
    Affiliations
    Department of Genetics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania

    Penn-CHOP Lifespan Brain Institute, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for articles by this author
  • Laura Schultz
    Affiliations
    Department of Genetics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania

    Penn-CHOP Lifespan Brain Institute, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for articles by this author
  • Jessica Turner
    Affiliations
    Department of Psychology, Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, Georgia

    Neurosciences Institute, Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, Georgia

    Tri-Institutional Center for Translational Research in Neuroimaging and Data Science, Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, Georgia
    Search for articles by this author
  • Vince Calhoun
    Affiliations
    Department of Psychology, Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, Georgia

    Neurosciences Institute, Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, Georgia

    Tri-Institutional Center for Translational Research in Neuroimaging and Data Science, Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, Georgia

    Departments of Neurosciences and Psychiatry and Behavioral Sciences, University of New Mexico School of Medicine, Albuquerque, New Mexico

    Mind Research Network, Departments of Psychiatry and Electrical and Computer Engineering, University of New Mexico, Albuquerque, New Mexico
    Search for articles by this author
  • David C. Glahn
    Affiliations
    Department of Psychiatry, Boston Children’s Hospital, Harvard Medical School, Boston, Massachusetts

    Olin Neuropsychiatry Research Center, Institute of Living, Hartford HealthCare, Hartford, Connecticut
    Search for articles by this author
Open AccessPublished:June 05, 2022DOI:https://doi.org/10.1016/j.bpsgos.2022.05.008

      Abstract

      Background

      Polygenic risk scores (PRSs) are indices of genetic liability for illness, but their clinical utility for predicting risk for a specific psychiatric disorder is limited. Genetic overlap among disorders and their effects on allied phenotypes may be a possible explanation, but this has been difficult to quantify given focus on singular disorders and/or allied phenotypes.

      Methods

      We constructed PRSs for 5 psychiatric disorders (schizophrenia, bipolar disorder, major depressive disorder, autism spectrum disorder, attention-deficit/hyperactivity disorder) and 3 nonpsychiatric control traits (height, type II diabetes, irritable bowel disease) in the UK Biobank (N = 31,616) and quantified associations between PRSs and phenotypes allied with mental illness: behavioral (symptoms, cognition, trauma) and brain measures from magnetic resonance imaging. We then evaluated the extent of specificity among PRSs and their effects on these allied phenotypes.

      Results

      Correlations among psychiatric PRSs replicated previous work, with overlap between schizophrenia and bipolar disorder, which was distinct from overlap between autism spectrum disorder and attention-deficit/hyperactivity disorder; overlap between psychiatric and control PRSs was minimal. There was, however, substantial overlap of PRS effects on allied phenotypes among psychiatric disorders and among psychiatric disorders and control traits, where the extent and pattern of overlap was phenotype specific.

      Conclusions

      Results show that genetic distinctions between psychiatric disorders and between psychiatric disorders and control traits exist, but this does not extend to their effects on allied phenotypes. Although overlap can be informative, work is needed to construct PRSs that will function at the level of specificity needed for clinical application.

      Keywords

      Recent large-scale genome-wide association (GWA) studies have greatly expanded our understanding of the role common genetic variation plays in psychiatric diseases (
      • Gibson G.
      Rare and common variants: Twenty arguments.
      ,
      • Katsanis N.
      The continuum of causality in human genetic disorders.
      ,
      • Schork N.J.
      • Murray S.S.
      • Frazer K.A.
      • Topol E.J.
      Common vs. rare allele hypotheses for complex diseases.
      ). Polygenic risk scores (PRSs) are meant to index a person’s genetic liability for a trait across common variants. A PRS is a scalar value calculated from a weighted sum of allele effects over an individual’s genotype data (target data), in which weights are determined by summary statistics from a previously performed and independent GWA analysis on the trait of interest (discovery data). PRSs have successfully predicted disease risk in nonpsychiatric disorders, and in some cases, have predictive value on par with monogenic risk variants (
      • Khera A.V.
      • Chaffin M.
      • Aragam K.G.
      • Haas M.E.
      • Roselli C.
      • Choi S.H.
      • et al.
      Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations.
      ). However, PRS applications in psychiatry have not yet been clinically useful in determining risk for a specific psychiatric disorder (
      • Ikeda M.
      • Saito T.
      • Kanazawa T.
      • Iwata N.
      Polygenic risk score as clinical utility in psychiatry: A clinical viewpoint.
      ).
      A contributing factor may be the lack of illness-specific information captured by psychiatric PRS (
      • Zheutlin A.B.
      • Dennis J.
      • Karlsson Linnér R.
      • Moscati A.
      • Restrepo N.
      • Straub P.
      • et al.
      Penetrance and pleiotropy of polygenic risk scores for schizophrenia in 106,160 patients across four health care systems.
      ). Evidence suggests that there is genetic overlap among psychiatric disorders, with more overlap among psychotic/affective disorders (schizophrenia [SZ], bipolar disorder [BPD], and major depressive disorder [MDD]) (
      • Bulik-Sullivan B.
      • Finucane H.K.
      • Anttila V.
      • Gusev A.
      • Day F.R.
      • Loh P.R.
      • et al.
      An atlas of genetic correlations across human diseases and traits.
      ,
      • Lee S.H.
      • Ripke S.
      • Neale B.M.
      • Faraone S.V.
      • Purcell S.M.
      • et al.
      Cross-Disorder Group of the Psychiatric Genomics Consortium
      Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs.
      ) and among developmental disorders (autism spectrum disorder [ASD] and attention-deficit/hyperactivity disorder [ADHD]) (
      • Baselmans B.M.L.
      • Yengo L.
      • van Rheenen W.
      • Wray N.R.
      Risk in relatives, heritability, SNP-based heritability, and genetic correlations in psychiatric disorders: A review.
      ). Shared genetic variance between disorders across these categories has been reported, albeit less frequently and consistently (
      • Lee S.H.
      • Ripke S.
      • Neale B.M.
      • Faraone S.V.
      • Purcell S.M.
      • et al.
      Cross-Disorder Group of the Psychiatric Genomics Consortium
      Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs.
      ,
      • Chang S.
      • Yang L.
      • Wang Y.
      • Faraone S.V.
      Shared polygenic risk for ADHD, executive dysfunction and other psychiatric disorders.
      ,
      • Hamshere M.L.
      • Stergiakouli E.
      • Langley K.
      • Martin J.
      • Holmans P.
      • Kent L.
      • et al.
      Shared polygenic contribution between childhood attention-deficit hyperactivity disorder and adult schizophrenia.
      ). A second consideration is that the PRS represents more than diagnostic risk and is related to other allied phenotypes like symptomatology (
      • Mistry S.
      • Escott-Price V.
      • Florio A.D.
      • Smith D.J.
      • Zammit S.
      Genetic risk for bipolar disorder and psychopathology from childhood to early adulthood.
      ,
      • Hamshere M.L.
      • Langley K.
      • Martin J.
      • Agha S.S.
      • Stergiakouli E.
      • Anney R.J.
      • et al.
      High loading of polygenic risk for ADHD in children with comorbid aggression.
      ), substance use (
      • Reginsson G.W.
      • Ingason A.
      • Euesden J.
      • Bjornsdottir G.
      • Olafsson S.
      • Sigurdsson E.
      • et al.
      Polygenic risk scores for schizophrenia and bipolar disorder associate with addiction.
      ,
      • Wimberley T.
      • Agerbo E.
      • Horsdal H.T.
      • Ottosen C.
      • Brikell I.
      • Als T.D.
      • et al.
      Genetic liability to ADHD and substance use disorders in individuals with ADHD.
      ,
      • Andersen A.M.
      • Pietrzak R.H.
      • Kranzler H.R.
      • Ma L.
      • Zhou H.
      • Liu X.
      • et al.
      Polygenic scores for major depressive disorder and risk of alcohol dependence.
      ), cognition (
      • Wang S.H.
      • Hsiao P.C.
      • Yeh L.L.
      • Liu C.M.
      • Liu C.C.
      • Hwang T.J.
      • et al.
      Polygenic risk for schizophrenia and neurocognitive performance in patients with schizophrenia [published correction appears in Genes Brain Behav 2018; 17:93].
      ,
      • Aguilar-Lacasaña S.
      • Vilor-Tejedor N.
      • Jansen P.R.
      • López-Vicente M.
      • Bustamante M.
      • Burgaleta M.
      • et al.
      Polygenic risk for ADHD and ASD and their relation with cognitive measures in school children.
      ,
      • Yüksel D.
      • Dietsche B.
      • Forstner A.J.
      • Witt S.H.
      • Maier R.
      • Rietschel M.
      • et al.
      Polygenic risk for depression and the neural correlates of working memory in healthy subjects.
      ), and brain structure (
      • van der Merwe C.
      • Passchier R.
      • Mufford M.
      • Ramesar R.
      • Dalvie S.
      • Stein D.J.
      Polygenic risk for schizophrenia and associated brain structural changes: A systematic review.
      ), some of which are present across multiple disorders. However, quantification of shared and/or unique variance among PRSs and their effects on clinical risk factors (allied phenotypes) is difficult, given that studies often focus on a single disorder or a single allied phenotype.
      A third consideration affecting psychiatric PRS specificity could be the method of constructing the PRS itself. Traditionally, PRS are constructed from variants passing a particular p value threshold in the discovery GWA summary statistics, where the optimal threshold is unknown (
      • Choi S.W.
      • Mak T.S.-H.
      • O’Reilly P.F.
      Tutorial: A guide to performing polygenic risk score analyses.
      ). Therefore, PRS are often constructed at multiple p value thresholds in a forward selection process (from strict to more lenient p values) until an optimal association with case/control status is reached. However, this is only possible if one has access to individual-level data from the discovery sample. In addition to threshold selection, variants are also pruned to account for linkage disequilibrium (LD) to ensure only independent effects remain, but this too requires setting an arbitrary correlation threshold (
      • Choi S.W.
      • Mak T.S.-H.
      • O’Reilly P.F.
      Tutorial: A guide to performing polygenic risk score analyses.
      ). Furthermore, it has been shown that including all single nucleotide polymorphisms (SNPs) and incorporating LD information improves PRS prediction (
      • Vilhjálmsson B.J.
      • Yang J.
      • Finucane H.K.
      • Gusev A.
      • Lindström S.
      • Ripke S.
      • et al.
      Modeling linkage disequilibrium increases accuracy of polygenic risk scores.
      ).
      For the PRS to provide clinically actionable information about risk for disease, it needs to contain disease-specific information and should be constructed in way that avoids arbitrary decisions and accounts for dependency across the genome. Understanding PRS specificity, or lack thereof, will be important in determining what the role of psychiatric PRS will be in clinical practice. We constructed separate PRSs for 5 psychiatric disorders (SZ, BPD, MDD, ASD, and ADHD) in a population-based sample of individuals from the UK Biobank (N = 31,616). To calculate PRSs, we used PRS-CS (https://github.com/getian107/PRScs) (
      • Ge T.
      • Chen C.Y.
      • Ni Y.
      • Feng Y.C.A.
      • Smoller J.W.
      Polygenic prediction via Bayesian regression and continuous shrinkage priors.
      ), a method that incorporates all SNPs in the GWA discovery sample while accounting for LD information. To understand whether psychiatric PRSs contain disease-specific information, we evaluated overlap in genetic liability among disorders as well as overlapping effects of PRSs on allied phenotypes or traits commonly and consistently associated with risk for multiple psychiatric disorders including symptomatology (
      • Allsopp K.
      • Read J.
      • Corcoran R.
      • Kinderman P.
      Heterogeneity in psychiatric diagnostic classification.
      ,
      • Mayes S.D.
      • Calhoun S.L.
      • Mayes R.D.
      • Molitoris S.
      Autism and ADHD: Overlapping and discriminating symptoms.
      ,
      • Bambole V.
      • Johnston M.
      • Shah N.
      • Sonavane S.
      • Desouza A.
      • Shrivastava A.
      Symptom overlap between schizophrenia and bipolar mood disorder: Diagnostic issues.
      ), cognition (
      • McTeague L.M.
      • Huemer J.
      • Carreon D.M.
      • Jiang Y.
      • Eickhoff S.B.
      • Etkin A.
      Identification of common neural circuit disruptions in cognitive control across psychiatric disorders.
      ,
      • Snyder H.R.
      • Miyake A.
      • Hankin B.L.
      Advancing understanding of executive function impairments and psychopathology: Bridging the gap between clinical and cognitive approaches.
      ), trauma (
      • Copeland W.E.
      • Shanahan L.
      • Hinesley J.
      • Chan R.F.
      • Aberg K.A.
      • Fairbank J.A.
      • et al.
      Association of childhood trauma exposure with adult psychiatric disorders and functional outcomes [published correction appears in JAMA Netw Open 2020; 3:e207276].
      ), and brain structure (
      • Thompson P.M.
      • Jahanshad N.
      • Ching C.R.K.
      • Salminen L.E.
      • Thomopoulos S.I.
      • Bright J.
      • et al.
      ENIGMA and global neuroscience: A decade of large-scale studies of the brain in health and disease across more than 40 countries.
      ). Where possible, we also quantified whether overlap in PRS effects on allied phenotypes were reflective of overlap in genetic liability. Finally, to ensure that results were specific to psychiatric disorders, we performed all analyses with PRS calculated for 3 nonpsychiatric control traits that have been associated with psychiatric clinical risk phenomenon (
      • Biessels G.J.
      • Reijmer Y.D.
      Brain changes underlying cognitive dysfunction in diabetes: What can we learn from MRI?.
      ,
      • Chua C.S.
      • Bai C.H.
      • Shiao C.Y.
      • Hsu C.Y.
      • Cheng C.W.
      • Yang K.C.
      • et al.
      Negative correlation of cortical thickness with the severity and duration of abdominal pain in Asian women with irritable bowel syndrome.
      ,
      • Fang J.
      • Li S.
      • Li M.
      • Chan Q.
      • Ma X.
      • Su H.
      • et al.
      Altered white matter microstructure identified with tract-based spatial statistics in irritable bowel syndrome: A diffusion tensor imaging study.
      ,
      • Lam N.C.Y.
      • Yeung H.Y.
      • Li W.K.
      • Lo H.Y.
      • Yuen C.F.
      • Chang R.C.C.
      • Ho Y.S.
      Cognitive impairment in irritable bowel syndrome (IBS): A systematic review.
      ,
      • Mollon J.
      • Curran J.E.
      • Mathias S.R.
      • Knowles E.E.M.
      • Carlisle P.
      • Fox P.T.
      • et al.
      Neurocognitive impairment in type 2 diabetes: Evidence for shared genetic aetiology.
      ): height, type II diabetes (T2D), and irritable bowel disease (IBD).

      Methods and Materials

      Figure 1 shows a flowchart of all methods and procedures.
      Figure thumbnail gr1
      Figure 1Methods flowchart. First, polygenic risk score (PRS) for 5 psychiatric traits and 3 control traits were constructed via PRS-CS for 31,616 individuals in the UK Biobank passing genetic and imaging quality control filters. Discovery data used to determine allele weights for PRS construction were from 8 separate genome-wide association analyses (complete citations in in ). We then evaluated overlap in 2 contexts: in step 1, we evaluated overlap among genetic liability for all pairs of traits by correlating each pair of PRSs; in step 2 we evaluated overlap in PRS effects on allied phenotypes by regressing each PRS against several behavioral and brain measures. (∗ indicates that ICV is included as a covariate.) Finally, for phenotype categories containing all continuous measures (noted by †), we asked whether overlap in PRS effects on allied phenotypes were similar to the overlap in genetic liability among traits. We did this by formally testing the similarity between correlations of standardized PRS effect sizes (betas) from models in step 2 and correlations of PRSs from step 1 (via Mantel tests). ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BPD, bipolar disorder; GWA, genome-wide association; HWE, Hardy-Weinberg equilibrium; IBD, irritable bowel disease; ICV, intracranial volume; INFO, imputation quality; MAF, minor allele frequency; MDD, major depressive disorder; QC, quality control; SZ, schizophrenia; T2D, type II diabetes; VBM, voxel-based morphometry.

      Participants

      Participants were part of the UK Biobank (http://www.ukbiobank.ac.uk), a population-based prospective cohort of approximately 500,000 individuals, ages 40 to 69 years, recruited between 2006 and 2010 throughout Great Britain. Recruitment procedures for the UK Biobank are described elsewhere (
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • et al.
      UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      ). Analyses were performed on a subsample of biobank participants (N = 31,616, mean age [SD] = 64.1 (7.4), 46% male, White British ancestry) with magnetic resonance imaging data who passed our genetic filters and imaging quality control procedures (Supplemental Methods in Supplement 1). The UK Biobank received ethical approval from the Research Ethics Committee (REC reference 11/NW/0382).

      PRS Calculation

      PRSs were calculated for 5 psychiatric disorders (SZ, BPD, MDD, ASD, and ADHD) and 3 nonpsychiatric control traits (height, T2D, and IBD). Control traits were chosen based on 1) comparability to psychiatric traits in terms of polygenicity (
      • Zhang Y.
      • Qi G.
      • Park J.H.
      • Chatterjee N.
      Estimation of complex effect-size distributions using summary-level statistics from genome-wide association studies across 32 complex traits.
      ) and heritability (Table S1 in Supplement 1), and 2) availability of GWA results with comparable sample sizes while excluding UK Biobank participants. PRSs were calculated using PRS-CS (
      • Ge T.
      • Chen C.Y.
      • Ni Y.
      • Feng Y.C.A.
      • Smoller J.W.
      Polygenic prediction via Bayesian regression and continuous shrinkage priors.
      ), which uses Bayesian regression to place continuous shrinkage priors on SNP effect sizes from GWA summary statistics. PRS-CS incorporates an external LD reference panel (1000 Genomes Project European samples; N = 503) to model local LD patterns and update effect sizes jointly within LD blocks, allowing accommodation of diverse genetic architectures and avoiding decisions related to pruning and GWA threshold selection (
      • Vilhjálmsson B.J.
      • Yang J.
      • Finucane H.K.
      • Gusev A.
      • Lindström S.
      • Ripke S.
      • et al.
      Modeling linkage disequilibrium increases accuracy of polygenic risk scores.
      ). For each trait, PRS-CS was used to generate weights for overlapping variants between a discovery GWA sample, the LD reference panel, and the quality controlled and filtered target UK Biobank sample (Supplemental Methods and Table S1 in Supplement 1). Resulting weights were applied to imputed genotypes in the UK Biobank to calculate trait PRS via the PLINK score command. Finally, PRS were z scored and residualized for 10 ancestry principal components (Supplemental Methods in Supplement 1).
      For binary traits, we validated PRS scores using a Welch two-sample t test to compare the average PRS of those with an ICD-10 code for the respective trait to individuals with no chapter V ICD-10 code (Mental Health) or those with no code for T2D or IBD. Owing to the small number of cases for some disorders, we generated several ICD-10 categories including a wider scope of ICD-10 diagnoses or used self-reported diagnoses (Supplemental Methods and Table S2 in Supplement 1 for details). For height, we performed linear regression between height PRS and standardized standing height in the UK Biobank, using age and sex as covariates.

      Psychiatric-Associated Phenotypes

      We considered 2 categories of allied phenotypes: behavior and brain. Each category contained classes of phenotypes with multiple measures in each. Behavioral phenotypes included self-reported psychiatric symptoms (items for psychosis, mania, depression, anxiety, addiction, and self-harm), trauma (items for traumatic events, childhood trauma, and adulthood trauma), and cognition (performance on fluid intelligence test, time to complete numeric and alpha-numeric paths for the trails test, number of puzzles solved correctly for matrix reasoning test, and longest number string recalled during the numeric memory test). Information regarding the derivation of all measures in the behavioral phenotype category are detailed in Supplemental Methods in Supplement 1 and Table S3 in Supplement 1.
      Brain phenotypes included gray matter anatomy quantified via surface area and cortical thickness extracted from 31 bilateral regions from the Desikan-Killany-Tourville Atlas (
      • Desikan R.S.
      • Ségonne F.
      • Fischl B.
      • Quinn B.T.
      • Dickerson B.C.
      • Blacker D.
      • et al.
      An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest.
      ) in FreeSurfer (
      • Fischl B.
      ) and voxelwise gray matter volume via voxel-based morphometry (VBM) (
      • Ashburner J.
      • Friston K.J.
      Voxel-based morphometry—The methods.
      ) using FSL v.6.0.3 (
      • Jenkinson M.
      • Beckmann C.F.
      • Behrens T.E.
      • Woolrich M.W.
      • Smith S.M.
      FSL.
      ) (Supplemental Methods in Supplement 1). We also quantified white matter integrity using skeletonized fractional anisotropy (FA) measures from TBSS (
      • Smith S.M.
      • Jenkinson M.
      • Johansen-Berg H.
      • Rueckert D.
      • Nichols T.E.
      • Mackay C.E.
      • et al.
      Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data.
      ) in 21 bilateral and 6 unilateral tracts of the ICBM-JHU-Atlas (
      • Mori S.
      • Wakana S.
      • Van Zijl P.C.
      • Nagae-Poetscher L.
      MRI Atlas of Human White Matter.
      ). FreeSurfer and FA measures were downloaded directly from the UK Biobank showcase (Table S4 in Supplement 2 for a complete list). For lateralized FreeSurfer and white matter measures, hemispheres were averaged.

      Analysis

      Overlap Among PRSs

      To quantify the specificity of psychiatric PRSs, it was necessary to assess the relative independence of each PRS score. To do this, we correlated each PRS pair. To ensure correlations between PRSs were attributable to comparable weighting of similar variants, we also correlated the SNP weights assigned by PRS-CS and compared them with the score correlation matrix using a Mantel test (
      • Mantel N.
      The detection of disease clustering and a generalized regression approach.
      ).

      Overlap of PRS Effects on Allied Phenotypes

      For each measure in our allied phenotype categories (excluding VBM), we performed separate general linear models (GLMs) (logistic regression for binary phenotypes, linear regressions for continuous phenotypes) with PRS as a continuous predictor and covariates for age and sex. All brain phenotypes included scanning site as an additional covariate. Models for FreeSurfer regions also included a covariate for intracranial volume. All continuous outcomes and covariates were transformed to z scores. Multiple testing correction was performed across all phenotypes regardless of category (N = 824) (excluding VBM) by controlling the false discovery rate (
      • Benjamini Y.
      • Hochberg Y.
      Controlling the false discovery rate: A practical and powerful approach to multiple testing.
      ); effects with false discovery rate–corrected p values < .05 were considered significant.
      For the VBM analysis, voxelwise output (N = 202,898 voxels) from data preprocessing (Supplemental Methods in Supplement 1) was used as the dependent variable in a GLM with PRS as a predictor and covariates for age, sex, and site (intracranial volume effects are accounted for in the modulation step of the VBM pipeline). Voxels significantly associated with each PRS were determined using FSL’s randomise (
      • Winkler A.M.
      • Ridgway G.R.
      • Webster M.A.
      • Smith S.M.
      • Nichols T.E.
      Permutation inference for the general linear model.
      ) and Threshold-Free Cluster Enhancement (
      • Smith S.M.
      • Nichols T.E.
      Threshold-free cluster enhancement: Addressing problems of smoothing, threshold dependence and localisation in cluster inference.
      ), which implements cluster-based thresholding and returns a familywise error–corrected statistical image. For each PRS, voxels with a familywise error–corrected p value < .05 were considered significant.
      Finally, we compared overlap in PRS effects on associated phenotypes with overlap in genetic liability. To do this, we first correlated standardized effect sizes (betas) from each PRS’s regression on individual measures within a phenotype class. This was only possible for phenotype classes with exclusively continuous variables (e.g., cognition, surface area, cortical thickness, VBM volume, white matter FA). We then statistically compared the correlation in PRS effect sizes with correlations among PRSs themselves using Mantel tests. Significant Mantel tests indicate a statistically significant similarity between the 2 matrices being tested.

      Results

      Validation of PRS

      For binary traits, individuals with an ICD-10 code or self-reported diagnosis had a higher average PRS for that trait than those who did not (Table S5 in Supplement 2). This difference was significant for SZ, BPD, MDD, and T2D. The PRS for height significantly predicted standing height in the UK Biobank (β = 0.35, p < 2 × 10−16).

      Relationships Among PRSs

      Correlations between PRS weights and PRS themselves were highly similar (r = 0.97, p = .0004) (Figure 2A), indicating that similar weighting of overlapping genetic variants accounted for correlations in overall scores. Correlations ranged from 0.01 to 0.36 and were mostly positive, with the exception of those between height and all other trait PRS, which were at 0 or negative and near 0. Correlations were modest among SZ and BPD PRS (r = 0.36) and between PRS for ASD and ADHD (r = 0.34) although correlations between these 2 groups of disorders did not exceed 0.04. MDD PRS was similarly correlated with all psychiatric disorders (SZ: r = 0.14, BPD: r = 0.16, ASD: r = 0.19, ADHD: r = 0.24). PRS for control traits (height, T2D, and IBD) were not strongly correlated with the PRS for any psychiatric disorders (all correlations < 0.06).
      Figure thumbnail gr2
      Figure 2Relationships among polygenic risk score (PRS) and PRS effects. (A) Pairwise correlations between PRSs (top) and PRS-CS weights (bottom). (B–F) Pairwise correlations between general linear models effect sizes (standardized betas) of PRS across cognition (B), surface area (C), cortical thickness (D), unthresholded beta maps from the voxel-based morphometry (VBM) general linear models analysis (E), and white matter fractional anisotropy (FA) (F) measures. Correlation coefficients and p values located under each correlation matrix represent the results of Mantel tests comparing each matrix to the correlation among PRS scores (A). ∗p values < .05 indicate that there is a significant relationship between the 2 correlation matrices. ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BPD, bipolar disorder; IBD, irritable bowel disease; MDD, major depressive disorder; SZ, schizophrenia; T2D, type II diabetes.

      PRS Effects on Allied Phenotypes

      Descriptive data and model summaries of PRS effects on all allied phenotypes are listed in Table S3 and Figure S1 in Supplement 1 and Tables S4, S6, and S7 in Supplement 2.

      Behavior

      PRS effects showed significant overlap on behavioral phenotypes. In some cases, this overlap followed a similar pattern as overlap in genetic liability among disorders. For example, higher PRS for SZ, BPD, or MDD significantly increased the probability of reporting psychotic experiences, anxiety, and addiction, whereas these associations were not significant for ASD and ADHD nor any control traits (Figure 3). However, divisions between mood/psychotic disorders and neurodevelopmental disorders were not observed for most behavioral measures. Higher PRS for any psychiatric disorder (SZ, BPD, MDD, ASD, ADHD) was significantly associated with an increased probability of self-reporting depression and self-harm, experiencing a traumatic event, higher levels of reported trauma in both childhood and adulthood, and slower processing speed. In some instances, there also was overlap in PRS effects between psychiatric and control traits. This was especially true for ASD and height PRSs, which had similar effects on IQ and matrix reasoning, and BPD, MDD, T2D, and IBD PRSs, which had similar effects on matrix reasoning.
      Figure thumbnail gr3
      Figure 3Polygenic risk score effects on behavioral phenotypes. Bars represent effect sizes (betas) for the effect of trait polygenic risk scores on behavioral phenotypes resulting from either a linear regression (standardized continuous phenotypes) or logistic regression (binary phenotypes). Asterisks indicate significant effects after false discovery rate correction. ∗p < .05, ∗∗p < .01, ∗∗∗p < .001. ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BPD, bipolar disorder; IBD, irritable bowel disease; MDD, major depressive disorder; SZ, schizophrenia; T2D, type II diabetes.
      Overall, overlap in PRS effects on cognitive phenotypes was not consistent with patterns of overlap in genetic liability. In fact, when comparing correlations in PRS effect sizes across cognitive measures to correlations in PRSs themselves the relationship was weak (r = 0.29, p = .11) (see Figure 2B), indicating that the correlation matrices were unrelated.

      Brain: FreeSurfer

      Overlap in PRS effects on brain phenotypes was measure specific. For surface area, significant effects of PRS were seen across SZ and ADHD in the superior temporal gyrus (negative), MDD and ADHD in the inferior temporal gyrus (negative), and BPD and ADHD in the lateral and medial orbitofrontal, superior frontal, and rostral anterior cingulate gyri (positive for BPD and negative for ADHD). Psychiatric and control PRS also shared significant associations: higher ADHD and T2D PRS were associated with smaller surface area of the inferior and middle temporal gyri, superior frontal gyrus, and pars orbitalis, while higher BPD and height PRS was associated with larger surface area of the medial and lateral orbitofrontal cortices, rostral anterior cingulate, and precuneus. Trait-specific psychiatric PRS effects (effects in nonoverlapping regions) were observed for BPD in the caudal middle frontal gyrus and caudal anterior cingulate (positive) and for ADHD in the rostral middle frontal gyrus (negative). Overall, GLM effect sizes were modestly correlated across psychiatric disorders, with the exception of ADHD and BPD/ASD (r = −0.08 and r = −0.09, respectively) and ASD and MDD (r = 0.02) (Figure 2C). The pattern of overlap in PRS effect sizes for surface area across the cortex were unrelated to overlap among PRSs (r = 0.25, p = .10).
      For cortical thickness, significant effects of PRS were seen across MDD and ADHD in the lingual gyrus (negative) and SZ and BPD in the medial and lateral orbitofrontal gyri (negative) (Figure 4; Figure S2 in Supplement 1; Table S8 in Supplement 2). Overlap in psychiatric and control PRS effects on cortical thickness was limited, but present for MDD and height in the lateral occipital gyrus (positive for MDD, negative for height), SZ and height in the lateral orbitofrontal gyrus (negative), and BPD and IBD in the pars orbitalis (negative for BPD, positive for IBD). Trait-specific psychiatric PRS effects were observed for SZ in the superior temporal gyrus (negative), ADHD in the pericalcarine area (positive), and BPD in the pars triangularis, posterior middle frontal, and superior frontal gyri (all negative). Overall, GLM effect sizes were modestly correlated across psychiatric and nonpsychiatric traits, but were more similar to correlations between PRSs (r = 0.63 p = .003) (Figure 2D) than those for surface area.
      Figure thumbnail gr4
      Figure 4Polygenic risk score effects on FreeSurfer measures. Medial and lateral view of brain regions showing significant associations between trait polygenic risk scores and surface area (left) or cortical thickness (right) measures after false discovery rate correction. Colors represent standardized beta values resulting from the general linear models using polygenic risk scores to predict FreeSurfer measures. Warm colors indicate positive associations; cool colors indicate negative associations. A, anterior; ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BPD, bipolar disorder; IBD, irritable bowel disease; MDD, major depressive disorder; P, posterior; RH, right hemisphere; SZ, schizophrenia; T2D, type II diabetes.

      Brain: VBM

      Overlap in significant psychiatric PRS effects on brain volume was observed between SZ, ADHD, and/or MDD in the cerebellum (Figure 5A; Figure S3 in Supplement 1 and; Table S9 in Supplement 2). Voxels significantly associated with any psychiatric PRS heavily overlapped with the PRS effects of at least 1 control trait (height, T2D, or IBD) (Figure 5B). Overlap in PRS effects on VBM volume was somewhat comparable to the pattern of overlap among PRSs themselves (r = 0.62, p = .004) (see Figure 2E).
      Figure thumbnail gr5
      Figure 5Polygenic risk score (PRS) effects on voxel-based morphometry (VBM) volume. (A) Sagittal view of voxels showing significant associations between trait PRS and VBM volume after familywise error correction. Colors represent standardized beta values resulting from the general linear models using PRS to predict voxelwise brain volume. Warm colors indicate positive associations; cool colors indicate negative associations. (B) Sagittal view of spatial overlap of significant PRS effects across psychiatric traits (top) and across psychiatric and control traits (bottom). ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BPD, bipolar disorder; IBD, irritable bowel disease; MDD, major depressive disorder; SZ, schizophrenia; T2D, type II diabetes.

      Brain: Diffusion

      PRS effects on FA between SZ and BPD were highly similar, however there was also overlap among PRS effects on FA for these disorders and MDD in the posterior thalamic radiation, external capsule, and cingulum-cingulate gyrus portion (all negative), ASD in the corpus callosum body and genu and fornix cresstria terminalis (both negative), and ADHD in the superior cerebellar peduncle and fornix cresstria terminalis (both negative) (Figure 6). Furthermore, a majority of significant effects of psychiatric PRS on tract FA overlapped with that of control traits (height and/or T2D). Tract-specific negative associations between PRS and FA were observed for SZ in the cingulum-hippocampal tract and sagittal stratum and MDD in the superior coronal radiata. Correlations between GLM effect sizes on FA measures was less cohesive among psychiatric disorders than for other brain traits (Figure 2F). Overlap in PRS effects on white matter FA were weak but significantly associated with overlap between PRSs (r = 0.38, p = .04).
      Figure thumbnail gr6
      Figure 6Polygenic risk score effects on white matter fractional anisotropy. Sagittal, axial, and coronal view of white matter tracts showing significant associations between trait polygenic risk scores and fractional anisotropy. Colors represent standardized beta values resulting from the general linear models using polygenic risk scores to predict tractwise fractional anisotropy. Warm colors indicate positive associations; cool colors indicate negative associations. Note: Figure shows full tracts for visualization purposes, but fractional anisotropy measures were extracted from skeletonized tracts using FSL’s TBSS procedure. ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; BPD, bipolar disorder; IBD, irritable bowel disease; MDD, major depressive disorder; SZ, schizophrenia; T2D, type II diabetes.

      Discussion

      Clinical use of PRSs to predict risk for psychiatric disorders has unfortunately not yet come to fruition. Our results show that this is partly because of overlapping genetic influences among disorders even when using PRS construction methods that avoid arbitrary thresholding decisions and consider LD structure. While genetic similarity among select psychiatric disorders is novel, we show that psychiatric disorders with minimally correlated PRSs (e.g., ADHD and SZ) show similar effects on allied phenotypes. Moreover, we showed substantial overlap in PRS effects on allied phenotypes between psychiatric and control traits, again despite little correlation between any control PRS and any psychiatric PRS. This could indicate that the issue of PRS specificity may extend beyond psychiatric traits and that independence between trait PRSs do not ensure separable effects on allied phenotypes. Furthermore, genetic risk for nonpsychiatric traits may play a role in influencing factors that put people at risk for psychiatric problems.
      Consistent with previous reports of genetic overlap among psychiatric disorders, we showed modest correlations between SZ and BPD PRS and distinctions between psychotic/affective disorders (SZ, BPD) and developmental disorders (ASD, ADHD) (
      • Bulik-Sullivan B.
      • Finucane H.K.
      • Anttila V.
      • Gusev A.
      • Day F.R.
      • Loh P.R.
      • et al.
      An atlas of genetic correlations across human diseases and traits.
      ,
      • Lee S.H.
      • Ripke S.
      • Neale B.M.
      • Faraone S.V.
      • Purcell S.M.
      • et al.
      Cross-Disorder Group of the Psychiatric Genomics Consortium
      Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs.
      ,
      • Baselmans B.M.L.
      • Yengo L.
      • van Rheenen W.
      • Wray N.R.
      Risk in relatives, heritability, SNP-based heritability, and genetic correlations in psychiatric disorders: A review.
      ). We also showed comparable associations between individual psychiatric PRSs and single measures reported in previous work including those between SZ and cognition (
      • Mistry S.
      • Harrison J.R.
      • Smith D.J.
      • Escott-Price V.
      • Zammit S.
      The use of polygenic risk scores to identify phenotypes associated with genetic risk of schizophrenia: Systematic review.
      ), psychotic/affective disorders (SZ, BPD, MDD) and substance abuse (
      • Carey C.E.
      • Agrawal A.
      • Bucholz K.K.
      • Hartz S.M.
      • Lynskey M.T.
      • Nelson E.C.
      • et al.
      Associations between polygenic risk for psychiatric disorders and substance involvement.
      ), and all psychiatric disorders and childhood abuse (
      • Peyrot W.J.
      • Milaneschi Y.
      • Abdellaoui A.
      • Sullivan P.F.
      • Hottenga J.J.
      • Boomsma D.I.
      • Penninx B.W.
      Effect of polygenic risk scores on depression in childhood trauma.
      ,
      • Ratanatharathorn A.
      • Koenen K.C.
      • Chibnik L.B.
      • Weisskopf M.G.
      • Rich-Edwards J.W.
      • Roberts A.L.
      Polygenic risk for autism, attention-deficit hyperactivity disorder, schizophrenia, major depressive disorder, and neuroticism is associated with the experience of childhood abuse.
      ). In contrast to previous reports, which failed to find associations between PRS and brain phenotypes (
      • van der Merwe C.
      • Passchier R.
      • Mufford M.
      • Ramesar R.
      • Dalvie S.
      • Stein D.J.
      Polygenic risk for schizophrenia and associated brain structural changes: A systematic review.
      ,
      • Reus L.M.
      • Shen X.
      • Gibson J.
      • Wigmore E.
      • Ligthart L.
      • Adams M.J.
      • et al.
      Association of polygenic risk for major psychiatric illness with subcortical volumes and white matter integrity in UK Biobank.
      ,
      • Simões B.
      • Vassos E.
      • Shergill S.
      • McDonald C.
      • Toulopoulou T.
      • Kalidindi S.
      • et al.
      Schizophrenia polygenic risk score influence on white matter microstructure.
      ), we detected associations between certain psychiatric PRSs and regional surface area, cortical thickness, and FA, possibly owing to increased power provided by the current sample size of the UK Biobank. PRS/brain associations were also in contrast to work showing limited genetic correlation between psychiatric traits and cortical thickness and surface area measures (
      • Hofer E.
      • Roshchupkin G.V.
      • Adams H.H.H.
      • Knol M.J.
      • Lin H.
      • Li S.
      • et al.
      Genetic correlations and genome-wide associations of cortical structure in general population samples of 22,824 adults.
      ), although we did replicate select genetic associations between brain and SZ (superior temporal surface area and lateral orbitofrontal thickness), BPD (insula surface area), and ADHD (fusiform surface area). We do not see this as problematic because genetic correlations between traits is not equivalent to our analyses here, where we used PRS to predict phenotypes, which include both genetic and nongenetic variance. Studies considering PRS effects on VBM-associated volumes are rare (
      • Ranlund S.
      • Rosa M.J.
      • de Jong S.
      • Cole J.H.
      • Kyriakopoulos M.
      • Fu C.H.Y.
      • et al.
      Associations between polygenic risk scores for four psychiatric illnesses and brain structure using multivariate pattern recognition.
      ,
      • Spalthoff R.
      • Degenhardt F.
      • Awasthi S.
      • Heilmann-Heimbach S.
      • Besteher B.
      • Gaser C.
      • et al.
      Effects of a neurodevelopmental genes based polygenic risk score for schizophrenia and single gene variants on brain structure in non-clinical subjects: A preliminary report.
      ), but associations between SZ PRS and superior temporal gyrus volume have been reported. We replicated this association in both cortical thickness and surface area, but not within our VBM analysis, where most significant psychiatric PRS effects were located in the cerebellum.
      While replication of prior results and the discovery of new results is encouraging, the current study emphasized the lack of specificity among PRSs and their effects on allied phenotypes. This was especially true for behavioral measures, where there was little evidence of disorder-specific associations. Psychiatric PRS effect sizes were highly similar, differing in quantity rather than quality across phenotypes (Figure 3). The exception was ASD where higher ASD PRS was associated with better performance and other psychiatric PRSs were associated with worse performance on nonreaction time measures of cognition. Some disorder-specific associations were observed between PRS and brain phenotypes including BPD PRS and frontal/cingulate surface area (caudal middle frontal, caudal anterior cingulate) and thickness (pars triangularis, rostral middle frontal, and superior frontal), SZ PRS and superior temporal gyrus thickness and tractwise FA (cingulum-hippocampal portion, sagittal stratum), ADHD PRS and rostral middle frontal surface area, and MDD PRS and superior coronal radiata FA. Significant association between SZ PRS and superior temporal gyrus thickness was especially promising given its robust relationship with positive symptoms in SZ (
      • Walton E.
      • Hibar D.P.
      • van Erp T.G.M.
      • Potkin S.G.
      • Roiz-Santiañez R.
      • Crespo-Facorro B.
      • et al.
      Positive symptoms associate with cortical thinning in the superior temporal gyrus via the ENIGMA Schizophrenia consortium.
      ). Further specificity could be gleaned by looking across brain phenotypes, specifically for BPD, which was the only psychiatric PRS to show associations with global decreases in cortical thickness and global increases in surface area.
      Despite indications of disorder-specific PRS effects on individual brain regions, overall correlations between GLM effect sizes predicting brain traits from PRS were substantial between psychiatric disorders (e.g., MDD and ADHD: cortical thickness, r = 0.49) and between psychiatric disorders and control traits (e.g., T2D and MDD: FA, r = 0.58) (Figure 2B–E). Patterns of GLM effect size correlations were also phenotype-dependent because pairs of PRS with the most similar effects on brain (highest correlations) were different depending on the measure type (surface area, cortical thickness, VBM, FA). SZ and BPD PRS showed the most consistent effects on allied phenotypes, where correlations in GLM effect sizes were above 0.48 except for surface area (r = 0.25); this is somewhat expected given their PRS scores were modestly correlated (r = 0.36) and support from current genetic (
      • Bulik-Sullivan B.
      • Finucane H.K.
      • Anttila V.
      • Gusev A.
      • Day F.R.
      • Loh P.R.
      • et al.
      An atlas of genetic correlations across human diseases and traits.
      ,
      • Lee S.H.
      • Ripke S.
      • Neale B.M.
      • Faraone S.V.
      • Purcell S.M.
      • et al.
      Cross-Disorder Group of the Psychiatric Genomics Consortium
      Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs.
      ,
      • Prata D.P.
      • Costa-Neves B.
      • Cosme G.
      • Vassos E.
      Unravelling the genetic basis of schizophrenia and bipolar disorder with GWAS: A systematic review.
      ) and neuroimaging research (
      • Birur B.
      • Kraguljac N.V.
      • Shelton R.C.
      • Lahti A.C.
      Brain structure, function, and neurochemistry in schizophrenia and bipolar disorder-A systematic review of the magnetic resonance neuroimaging literature.
      ). However, there were unexpected correlations in GLM effect sizes between T2D and ADHD (r = 0.44, surface area; r = 0.33, VBM), SZ (r = 0.48, cortical thickness), and MDD (r = 0.58, FA), given correlations between T2D PRS and each of these disorders’ PRSs were 0.06, 0.02, and 0.03, respectively. Despite little to no relationship between height and psychiatric PRSs (Figure 2A), GLM effect sizes for height PRS were often negatively correlated with GLM effect sizes for all psychiatric PRSs (Figure 2B–F). This negative relationship, and others like it, may explain heterogeneity within psychiatric diagnoses, because distinct genetic factors could pull the same phenotype in opposing directions. Departures between PRS GLM effects and the PRS themselves highlights the extremely polygenic nature of allied phenotypes and provides a cautionary note that distinctions in genetic liability do not ensure independent effects on allied phenotypes.
      The UK Biobank is a population-based cohort with relatively few affected individuals for psychiatric disorders as well as biases for recruitment (
      • Fry A.
      • Littlejohns T.J.
      • Sudlow C.
      • Doherty N.
      • Adamska L.
      • Sprosen T.
      • et al.
      Comparison of sociodemographic and health-related characteristics of UK Biobank participants with those of the general population.
      ) and retention (
      • Tyrrell J.
      • Zheng J.
      • Beaumont R.
      • Hinton K.
      • Richardson T.G.
      • Wood A.R.
      • et al.
      Genetic predictors of participation in optional components of UK Biobank.
      ) that may limit the generalizability of our results. However, the UK Biobank is currently the largest sample with a breadth of phenotypes, especially neuroimaging, where samples rarely exceed 1000. Analyzing population cohorts is necessary when considering many phenotypes and multiple PRS scores in a single sample and can be a useful comparison to case-control studies. Another consideration is that while PRS effects sizes on some allied phenotypes were significant, they were small, as was the proportion of variance explained solely by the PRS in GLM models (Tables S6–S8). We also acknowledge that similar effects of PRSs on allied phenotypes may be because of correlations among the phenotypes themselves (e.g., phenotypic correlations among cognitive phenotypes in Figure S1 in Supplement 1 compared with strong correlations in PRS effect sizes on cognitive measures in Figure 2B), however, high phenotypic correlations among allied phenotypes did not always guarantee similarity in PRS effects. For example, cortical thickness within parietal regions and between parietal and frontal regions showed the greatest phenotypic correlations (Figure S1 in Supplement 1), however PRS effects on these regions were separable both within and across psychiatric traits (see differential SZ PRS effects within frontal and parietal regions and differential SZ and BPD PRS effects in those same regions in Figure S2 in Supplement 1). Different allied phenotypes may have larger effect sizes, larger proportion of variance explained, and/or provide better distinction between psychiatric PRSs. We focused on allied phenotypes with the most support in their associations to psychiatric disease (
      • McTeague L.M.
      • Huemer J.
      • Carreon D.M.
      • Jiang Y.
      • Eickhoff S.B.
      • Etkin A.
      Identification of common neural circuit disruptions in cognitive control across psychiatric disorders.
      ,
      • Snyder H.R.
      • Miyake A.
      • Hankin B.L.
      Advancing understanding of executive function impairments and psychopathology: Bridging the gap between clinical and cognitive approaches.
      ,
      • Copeland W.E.
      • Shanahan L.
      • Hinesley J.
      • Chan R.F.
      • Aberg K.A.
      • Fairbank J.A.
      • et al.
      Association of childhood trauma exposure with adult psychiatric disorders and functional outcomes [published correction appears in JAMA Netw Open 2020; 3:e207276].
      ,
      • Thompson P.M.
      • Jahanshad N.
      • Ching C.R.K.
      • Salminen L.E.
      • Thomopoulos S.I.
      • Bright J.
      • et al.
      ENIGMA and global neuroscience: A decade of large-scale studies of the brain in health and disease across more than 40 countries.
      ,
      • Feldman B.J.
      • Conger R.D.
      • Burzette R.G.
      Traumatic events, psychiatric disorders, and pathways of risk and resilience during the transition to adulthood.
      ) available to us in the UK Biobank. Finally, considering rare genetic variation in conjunction with the PRS (
      • Bergen S.E.
      • Ploner A.
      • Howrigan D.
      • O’Donovan M.C.
      • Smoller J.W.
      • et al.
      CNV Analysis Group and the Schizophrenia Working Group of the Psychiatric Genomics Consortium
      Joint contributions of rare copy number variants and common SNPs to risk for schizophrenia.
      ) may also lead to increases in effect sizes and more genetic specificity among disorders.
      A primary goal of the PRSs approach is to index genetic burden or disease risk. This is by no means their only application, but it is a substantial one. Our results show that this conception of PRSs is highly problematic as psychiatric PRSs contain little disease-specific information. Not only was there significant overlap among psychiatric disorders but also among psychiatric disorders and control traits in terms of genetic liability (correlations between PRSs) and their associations with allied phenotypes (GLMs predicting phenotypes from PRSs). As such, there is an even greater need to understand what psychiatric PRSs index besides pure diagnosis. Perhaps PRSs for mental illness simply index a general predisposition for psychiatric disease, rather than being linked to a specific illness. While this could limit the clinical utility of the approach, it could still be beneficial for assessing general risk. However, multi-illness PRS should be generated and tested against disease-specific indices.

      Acknowledgments and Disclosures

      This work was supported by the National Institute of Mental Health (Grant No. MH094524 [to JT, VC, DG]). All analyses were conducted under UK Biobank data application 4844.
      The authors report no biomedical financial interests or potential conflicts of interest.

      References

        • Gibson G.
        Rare and common variants: Twenty arguments.
        Nat Rev Genet. 2012; 13: 135-145
        • Katsanis N.
        The continuum of causality in human genetic disorders.
        Genome Biol. 2016; 17: 233
        • Schork N.J.
        • Murray S.S.
        • Frazer K.A.
        • Topol E.J.
        Common vs. rare allele hypotheses for complex diseases.
        Curr Opin Genet Dev. 2009; 19: 212-219
        • Khera A.V.
        • Chaffin M.
        • Aragam K.G.
        • Haas M.E.
        • Roselli C.
        • Choi S.H.
        • et al.
        Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations.
        Nat Genet. 2018; 50: 1219-1224
        • Ikeda M.
        • Saito T.
        • Kanazawa T.
        • Iwata N.
        Polygenic risk score as clinical utility in psychiatry: A clinical viewpoint.
        J Hum Genet. 2021; 66: 53-60
        • Zheutlin A.B.
        • Dennis J.
        • Karlsson Linnér R.
        • Moscati A.
        • Restrepo N.
        • Straub P.
        • et al.
        Penetrance and pleiotropy of polygenic risk scores for schizophrenia in 106,160 patients across four health care systems.
        Am J Psychiatry. 2019; 176: 846-855
        • Bulik-Sullivan B.
        • Finucane H.K.
        • Anttila V.
        • Gusev A.
        • Day F.R.
        • Loh P.R.
        • et al.
        An atlas of genetic correlations across human diseases and traits.
        Nat Genet. 2015; 47: 1236-1241
        • Lee S.H.
        • Ripke S.
        • Neale B.M.
        • Faraone S.V.
        • Purcell S.M.
        • et al.
        • Cross-Disorder Group of the Psychiatric Genomics Consortium
        Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs.
        Nat Genet. 2013; 45: 984-994
        • Baselmans B.M.L.
        • Yengo L.
        • van Rheenen W.
        • Wray N.R.
        Risk in relatives, heritability, SNP-based heritability, and genetic correlations in psychiatric disorders: A review.
        Biol Psychiatry. 2021; 89: 11-19
        • Chang S.
        • Yang L.
        • Wang Y.
        • Faraone S.V.
        Shared polygenic risk for ADHD, executive dysfunction and other psychiatric disorders.
        Transl Psychiatry. 2020; 10: 182
        • Hamshere M.L.
        • Stergiakouli E.
        • Langley K.
        • Martin J.
        • Holmans P.
        • Kent L.
        • et al.
        Shared polygenic contribution between childhood attention-deficit hyperactivity disorder and adult schizophrenia.
        Br J Psychiatry. 2013; 203: 107-111
        • Mistry S.
        • Escott-Price V.
        • Florio A.D.
        • Smith D.J.
        • Zammit S.
        Genetic risk for bipolar disorder and psychopathology from childhood to early adulthood.
        J Affect Disord. 2019; 246: 633-639
        • Hamshere M.L.
        • Langley K.
        • Martin J.
        • Agha S.S.
        • Stergiakouli E.
        • Anney R.J.
        • et al.
        High loading of polygenic risk for ADHD in children with comorbid aggression.
        Am J Psychiatry. 2013; 170: 909-916
        • Reginsson G.W.
        • Ingason A.
        • Euesden J.
        • Bjornsdottir G.
        • Olafsson S.
        • Sigurdsson E.
        • et al.
        Polygenic risk scores for schizophrenia and bipolar disorder associate with addiction.
        Addict Biol. 2018; 23: 485-492
        • Wimberley T.
        • Agerbo E.
        • Horsdal H.T.
        • Ottosen C.
        • Brikell I.
        • Als T.D.
        • et al.
        Genetic liability to ADHD and substance use disorders in individuals with ADHD.
        Addiction. 2020; 115: 1368-1377
        • Andersen A.M.
        • Pietrzak R.H.
        • Kranzler H.R.
        • Ma L.
        • Zhou H.
        • Liu X.
        • et al.
        Polygenic scores for major depressive disorder and risk of alcohol dependence.
        JAMA Psychiatry. 2017; 74: 1153-1160
        • Wang S.H.
        • Hsiao P.C.
        • Yeh L.L.
        • Liu C.M.
        • Liu C.C.
        • Hwang T.J.
        • et al.
        Polygenic risk for schizophrenia and neurocognitive performance in patients with schizophrenia [published correction appears in Genes Brain Behav 2018; 17:93].
        Genes Brain Behav. 2018; 17: 49-55
        • Aguilar-Lacasaña S.
        • Vilor-Tejedor N.
        • Jansen P.R.
        • López-Vicente M.
        • Bustamante M.
        • Burgaleta M.
        • et al.
        Polygenic risk for ADHD and ASD and their relation with cognitive measures in school children.
        Psychol Med. 2022; 52: 1356-1364
        • Yüksel D.
        • Dietsche B.
        • Forstner A.J.
        • Witt S.H.
        • Maier R.
        • Rietschel M.
        • et al.
        Polygenic risk for depression and the neural correlates of working memory in healthy subjects.
        Prog Neuropsychopharmacol Biol Psychiatry. 2017; 79: 67-76
        • van der Merwe C.
        • Passchier R.
        • Mufford M.
        • Ramesar R.
        • Dalvie S.
        • Stein D.J.
        Polygenic risk for schizophrenia and associated brain structural changes: A systematic review.
        Compr Psychiatry. 2019; 88: 77-82
        • Choi S.W.
        • Mak T.S.-H.
        • O’Reilly P.F.
        Tutorial: A guide to performing polygenic risk score analyses.
        Nat Protoc. 2020; 15: 2759-2772
        • Vilhjálmsson B.J.
        • Yang J.
        • Finucane H.K.
        • Gusev A.
        • Lindström S.
        • Ripke S.
        • et al.
        Modeling linkage disequilibrium increases accuracy of polygenic risk scores.
        Am J Hum Genet. 2015; 97: 576-592
        • Ge T.
        • Chen C.Y.
        • Ni Y.
        • Feng Y.C.A.
        • Smoller J.W.
        Polygenic prediction via Bayesian regression and continuous shrinkage priors.
        Nat Commun. 2019; 10: 1776
        • Allsopp K.
        • Read J.
        • Corcoran R.
        • Kinderman P.
        Heterogeneity in psychiatric diagnostic classification.
        Psychiatry Res. 2019; 279: 15-22
        • Mayes S.D.
        • Calhoun S.L.
        • Mayes R.D.
        • Molitoris S.
        Autism and ADHD: Overlapping and discriminating symptoms.
        Res Autism Spec Disord. 2012; 6: 277-285
        • Bambole V.
        • Johnston M.
        • Shah N.
        • Sonavane S.
        • Desouza A.
        • Shrivastava A.
        Symptom overlap between schizophrenia and bipolar mood disorder: Diagnostic issues.
        OJPsych. 2013; 3: 8-15
        • McTeague L.M.
        • Huemer J.
        • Carreon D.M.
        • Jiang Y.
        • Eickhoff S.B.
        • Etkin A.
        Identification of common neural circuit disruptions in cognitive control across psychiatric disorders.
        Am J Psychiatry. 2017; 174: 676-685
        • Snyder H.R.
        • Miyake A.
        • Hankin B.L.
        Advancing understanding of executive function impairments and psychopathology: Bridging the gap between clinical and cognitive approaches.
        Front Psychol. 2015; 6: 328
        • Copeland W.E.
        • Shanahan L.
        • Hinesley J.
        • Chan R.F.
        • Aberg K.A.
        • Fairbank J.A.
        • et al.
        Association of childhood trauma exposure with adult psychiatric disorders and functional outcomes [published correction appears in JAMA Netw Open 2020; 3:e207276].
        JAMA Network Open. 2018; 1e184493
        • Thompson P.M.
        • Jahanshad N.
        • Ching C.R.K.
        • Salminen L.E.
        • Thomopoulos S.I.
        • Bright J.
        • et al.
        ENIGMA and global neuroscience: A decade of large-scale studies of the brain in health and disease across more than 40 countries.
        Transl Psychiatry. 2020; 10: 100
        • Biessels G.J.
        • Reijmer Y.D.
        Brain changes underlying cognitive dysfunction in diabetes: What can we learn from MRI?.
        Diabetes. 2014; 63: 2244-2252
        • Chua C.S.
        • Bai C.H.
        • Shiao C.Y.
        • Hsu C.Y.
        • Cheng C.W.
        • Yang K.C.
        • et al.
        Negative correlation of cortical thickness with the severity and duration of abdominal pain in Asian women with irritable bowel syndrome.
        PloS One. 2017; 12e0183960
        • Fang J.
        • Li S.
        • Li M.
        • Chan Q.
        • Ma X.
        • Su H.
        • et al.
        Altered white matter microstructure identified with tract-based spatial statistics in irritable bowel syndrome: A diffusion tensor imaging study.
        Brain Imaging Behav. 2017; 11: 1110-1116
        • Lam N.C.Y.
        • Yeung H.Y.
        • Li W.K.
        • Lo H.Y.
        • Yuen C.F.
        • Chang R.C.C.
        • Ho Y.S.
        Cognitive impairment in irritable bowel syndrome (IBS): A systematic review.
        Brain Res. 2019; 1719: 274-284
        • Mollon J.
        • Curran J.E.
        • Mathias S.R.
        • Knowles E.E.M.
        • Carlisle P.
        • Fox P.T.
        • et al.
        Neurocognitive impairment in type 2 diabetes: Evidence for shared genetic aetiology.
        Diabetologia. 2020; 63: 977-986
        • Sudlow C.
        • Gallacher J.
        • Allen N.
        • Beral V.
        • Burton P.
        • Danesh J.
        • et al.
        UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
        PLoS Med. 2015; 12e1001779
        • Zhang Y.
        • Qi G.
        • Park J.H.
        • Chatterjee N.
        Estimation of complex effect-size distributions using summary-level statistics from genome-wide association studies across 32 complex traits.
        Nat Genet. 2018; 50: 1318-1326
        • Desikan R.S.
        • Ségonne F.
        • Fischl B.
        • Quinn B.T.
        • Dickerson B.C.
        • Blacker D.
        • et al.
        An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest.
        Neuroimage. 2006; 31: 968-980
        • Fischl B.
        FreeSurfer. Neuroimage. 2012; 62: 774-781
        • Ashburner J.
        • Friston K.J.
        Voxel-based morphometry—The methods.
        Neuroimage. 2000; 11: 805-821
        • Jenkinson M.
        • Beckmann C.F.
        • Behrens T.E.
        • Woolrich M.W.
        • Smith S.M.
        FSL.
        Neuroimage. 2012; 62: 782-790
        • Smith S.M.
        • Jenkinson M.
        • Johansen-Berg H.
        • Rueckert D.
        • Nichols T.E.
        • Mackay C.E.
        • et al.
        Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data.
        Neuroimage. 2006; 31: 1487-1505
        • Mori S.
        • Wakana S.
        • Van Zijl P.C.
        • Nagae-Poetscher L.
        MRI Atlas of Human White Matter.
        Elsevier, Amsterdam, The Netherlands2005
        • Mantel N.
        The detection of disease clustering and a generalized regression approach.
        Cancer Res. 1967; 27: 209-220
        • Benjamini Y.
        • Hochberg Y.
        Controlling the false discovery rate: A practical and powerful approach to multiple testing.
        J R Stat Soc B (Methodological). 1995; 57: 289-300
        • Winkler A.M.
        • Ridgway G.R.
        • Webster M.A.
        • Smith S.M.
        • Nichols T.E.
        Permutation inference for the general linear model.
        NeuroImage. 2014; 92: 381-397
        • Smith S.M.
        • Nichols T.E.
        Threshold-free cluster enhancement: Addressing problems of smoothing, threshold dependence and localisation in cluster inference.
        Neuroimage. 2009; 44: 83-98
        • Mistry S.
        • Harrison J.R.
        • Smith D.J.
        • Escott-Price V.
        • Zammit S.
        The use of polygenic risk scores to identify phenotypes associated with genetic risk of schizophrenia: Systematic review.
        Schizophr Res. 2018; 197: 2-8
        • Carey C.E.
        • Agrawal A.
        • Bucholz K.K.
        • Hartz S.M.
        • Lynskey M.T.
        • Nelson E.C.
        • et al.
        Associations between polygenic risk for psychiatric disorders and substance involvement.
        Front Genet. 2016; 7: 149
        • Peyrot W.J.
        • Milaneschi Y.
        • Abdellaoui A.
        • Sullivan P.F.
        • Hottenga J.J.
        • Boomsma D.I.
        • Penninx B.W.
        Effect of polygenic risk scores on depression in childhood trauma.
        Br J Psychiatry. 2014; 205: 113-119
        • Ratanatharathorn A.
        • Koenen K.C.
        • Chibnik L.B.
        • Weisskopf M.G.
        • Rich-Edwards J.W.
        • Roberts A.L.
        Polygenic risk for autism, attention-deficit hyperactivity disorder, schizophrenia, major depressive disorder, and neuroticism is associated with the experience of childhood abuse.
        Mol Psychiatry. 2021; 26: 1696-1705
        • Reus L.M.
        • Shen X.
        • Gibson J.
        • Wigmore E.
        • Ligthart L.
        • Adams M.J.
        • et al.
        Association of polygenic risk for major psychiatric illness with subcortical volumes and white matter integrity in UK Biobank.
        Sci Rep. 2017; 742140
        • Simões B.
        • Vassos E.
        • Shergill S.
        • McDonald C.
        • Toulopoulou T.
        • Kalidindi S.
        • et al.
        Schizophrenia polygenic risk score influence on white matter microstructure.
        J Psychiatr Res. 2020; 121: 62-67
        • Hofer E.
        • Roshchupkin G.V.
        • Adams H.H.H.
        • Knol M.J.
        • Lin H.
        • Li S.
        • et al.
        Genetic correlations and genome-wide associations of cortical structure in general population samples of 22,824 adults.
        Nat Commun. 2020; 11: 4796
        • Ranlund S.
        • Rosa M.J.
        • de Jong S.
        • Cole J.H.
        • Kyriakopoulos M.
        • Fu C.H.Y.
        • et al.
        Associations between polygenic risk scores for four psychiatric illnesses and brain structure using multivariate pattern recognition.
        NeuroImage Clin. 2018; 20: 1026-1036
        • Spalthoff R.
        • Degenhardt F.
        • Awasthi S.
        • Heilmann-Heimbach S.
        • Besteher B.
        • Gaser C.
        • et al.
        Effects of a neurodevelopmental genes based polygenic risk score for schizophrenia and single gene variants on brain structure in non-clinical subjects: A preliminary report.
        Schizophr Res. 2019; 212: 225-228
        • Walton E.
        • Hibar D.P.
        • van Erp T.G.M.
        • Potkin S.G.
        • Roiz-Santiañez R.
        • Crespo-Facorro B.
        • et al.
        Positive symptoms associate with cortical thinning in the superior temporal gyrus via the ENIGMA Schizophrenia consortium.
        Acta Psychiatr Scand. 2017; 135: 439-447
        • Prata D.P.
        • Costa-Neves B.
        • Cosme G.
        • Vassos E.
        Unravelling the genetic basis of schizophrenia and bipolar disorder with GWAS: A systematic review.
        J Psychiatr Res. 2019; 114: 178-207
        • Birur B.
        • Kraguljac N.V.
        • Shelton R.C.
        • Lahti A.C.
        Brain structure, function, and neurochemistry in schizophrenia and bipolar disorder-A systematic review of the magnetic resonance neuroimaging literature.
        NPJ Schizophr. 2017; 3: 15
        • Fry A.
        • Littlejohns T.J.
        • Sudlow C.
        • Doherty N.
        • Adamska L.
        • Sprosen T.
        • et al.
        Comparison of sociodemographic and health-related characteristics of UK Biobank participants with those of the general population.
        Am J Epidemiol. 2017; 186: 1026-1034
        • Tyrrell J.
        • Zheng J.
        • Beaumont R.
        • Hinton K.
        • Richardson T.G.
        • Wood A.R.
        • et al.
        Genetic predictors of participation in optional components of UK Biobank.
        Nat Commun. 2021; 12: 886
        • Feldman B.J.
        • Conger R.D.
        • Burzette R.G.
        Traumatic events, psychiatric disorders, and pathways of risk and resilience during the transition to adulthood.
        Res Hum Dev. 2004; 1: 259-290
        • Bergen S.E.
        • Ploner A.
        • Howrigan D.
        • O’Donovan M.C.
        • Smoller J.W.
        • et al.
        • CNV Analysis Group and the Schizophrenia Working Group of the Psychiatric Genomics Consortium
        Joint contributions of rare copy number variants and common SNPs to risk for schizophrenia.
        Am J Psychiatry. 2019; 176: 29-35