Polygenic Risk of Psychiatric Disorders Exhibits Cross-trait Associations in Electronic Health Record Data From European Ancestry Individuals

Abstract

Background: Prediction of disease risk is a key component of precision medicine. Common traits such as psychiatric disorders have a complex polygenic architecture, making the identification of a single risk predictor difficult. Polygenic risk scores (PRSs) denoting the sum of an individual's genetic liability for a disorder are a promising biomarker for psychiatric disorders, but they require evaluation in a clinical setting.

Methods: We developed PRSs for 6 psychiatric disorders (schizophrenia, bipolar disorder, major depressive disorder, cross disorder, attention-deficit/hyperactivity disorder, and anorexia nervosa) and 17 nonpsychiatric traits in more than 10,000 individuals from the Penn Medicine Biobank with accompanying electronic health records. We performed phenome-wide association analyses to test their association across disease categories.

Results: Four of the 6 psychiatric PRSs were associated with their primary phenotypes (odds ratios from 1.2 to 1.6). Cross-trait associations were identified both within the psychiatric domain and across trait domains. PRSs for coronary artery disease and years of education were significantly associated with psychiatric disorders, largely driven by an association with tobacco use disorder.

Conclusions: We demonstrated that the genetic architecture of electronic health record-derived psychiatric diagnoses is similar to ascertained research cohorts from large consortia. Psychiatric PRSs are moderately associated with psychiatric diagnoses but are not yet clinically predictive in naïve patients. Cross-trait associations for these PRSs suggest a broader effect of genetic liability beyond traditional diagnostic boundaries. As identification of genetic markers increases, including PRSs alongside other clinical risk factors may enhance prediction of psychiatric disorders and associated conditions in clinical registries.

Keywords: Biomarkers; Cross-trait; Electronic health records; Genetics; Polygenic risk scores; Psychiatric disorders.

Figure 1.

Case prevalence (%) by PRS quintile for schizophrenia (SCZ, phecode 295), bipolar disorder (BD, phecode 296.1), major depressive disorder (MDD, phecode 296.22), cross disorder (CROSS, phecodes 295, 296.1, 296.22, 313.1, 313.3), attention deficit/hyperactivity disorder (ADHD, phecode 313.1), and anorexia nervosa (AN, phecode 305.2) PRS.

Figure 2:

Phenome-wide association of psychiatric polygenic risk scores. A phenome-wide association plot is shown for each psychiatric PRS (schizophrenia, bipolar disorder, major depressive disorder, cross disorder, attention deficit/hyperactivity disorder, and anorexia nervosa). Phenotypes with >100 cases were tested (n=512) in 10,182 individuals from the Penn Medicine Biobank. Phenotypes are grouped by trait domain along the horizontal axis, the significance of association between the PRS and phenotype is shown on the vertical axis (−log₁₀ p; two-tailed). The blue line indicates phenome-wide significance following Bonferroni correction (p<9.77×10⁻⁵). Phenotypes reaching phenome-wide significance are labelled.

Figure 3.

Polygenic risk for other traits. SCZ=schizophrenia, BD=bipolar disorder, MDD=major depressive disorder, CROSS=cross disorder, ADHD=attention deficit/hyperactivity disorder, AN=anorexia nervosa, chronotype=chronotype, sleepduration=sleep duration, EduYears=years of education, IGAP=Alzheimer’s, HDL=high-density lipoproteins, LDL=low-density lipoproteins, TG=triglycerides, TC=total cholesterol, Diabetes=type 2 diabetes, COPD_FEV1=chronic obstructive pulmonary disease forced expiratory volume, COPD_FVC= chronic obstructive pulmonary disease forced vital capacity, CAD=coronary artery disease, MI=myocardial infarction, BMI=body mass index, Birthweight=birthweight, TAG_former=current vs. former smoker, TAG_CPD=cigarettes per day. A: Correlational structure between psychiatric PRS and PRS for other traits. Positive correlations are denoted in red, negative correlations are denoted in blue, white squares are non-significant. B: Association between all PRS and psychiatric disorder, broadly defined. PRS are grouped by trait domain. Odds ratios and 95% confidence intervals are shown for the association between each PRS and psychiatric disorder. Psychiatric disorder was defined as an individual having any psychiatric disorder phecode.