Specificity of Polygenic Scores for Psychiatric Disorders Beyond Transdiagnostic Genetic Risk.

Importance: Polygenic scores (PGSs) are increasingly used to estimate genetic liability for psychiatric disorders, but their limited disorder specificity constrains their clinical and research utility. Objective: To test whether associations between psychiatric PGSs and psychiatric outcomes are primarily driven by transdiagnostic or disorder-specific genetic liability. Design, Setting, and Participants: This cohort study used a population-based sample from the Twins Early Development Study of twins born in England and Wales between January 1994 and December 1996. Mental health data were collected from July 2021 through June 2023, when participants were aged 25 to 28 years. The analyses include participants who had available genetic data and at least 1 quantitative symptom score. Main Outcomes and Measures: The main outcomes were quantitative symptom scores and self-reported psychiatric diagnoses. Associations were tested using generalized estimating equations with 3 types of PGSs: (1) uncorrected; (2) a transdiagnostic score indexing the shared genetic liability across 11 major psychiatric disorders (general factor of psychopathology [p]); and (3) residual disorder-specific scores corrected for p (non-p). Results: Analyses included 6567 participants (mean [SD] age, 26.4 [0.93] years; 4220 female [64.3%]). For quantitative symptom scores, the PGS for p showed comparable or larger associations with outcomes than the corresponding uncorrected PGSs across most symptom domains. For example, the effect size for the association between the p PGS and posttraumatic stress disorder (PTSD) symptoms was 0.15 (95% CI, 0.11-0.18) compared with 0.10 (95% CI, 0.07-0.13) for the uncorrected PTSD PGS; associations for other traits were smaller in magnitude but directionally consistent. Associations with self-reported diagnoses were of similar magnitude for the p and uncorrected PGSs. Some disorder-specific (non-p) PGSs retained significant trait-specific associations, including with symptom scores, such as problematic alcohol use (β = 0.09; 95% CI, 0.07-0.12) and PTSD (β = 0.08; 95% CI, 0.05-0.11), and with self-reported diagnoses, such as PTSD (odds ratio, 1.43; 95% CI, 1.18-1.74) and anorexia nervosa (odds ratio, 1.28; 95% CI, 1.03-1.59), suggesting residual disorder specificity. Most uncorrected PGSs showed extensive cross-trait associations, which were substantially attenuated after accounting for p, suggesting that much of the genetic signal was associated with transdiagnostic genetic liability. Conclusions and Relevance: This cohort study found that the observed associations between PGSs and psychopathology outcomes were largely explained by transdiagnostic genetic liability, with some disorder-specific signals also detected. Accounting for shared genetic effects may enhance specificity of PGSs, while transdiagnostic scores may inform risk stratification and interventions targeting general psychopathology.