PT  - JOURNAL ARTICLE
AU  - Arianna M. Gard
AU  - Erin B. Ware
AU  - Luke W. Hyde
AU  - Lauren Schmitz
AU  - Jessica Faul
AU  - Colter Mitchell
TI  - Assessing the Structure and Specificity of Polygenic Scores for Psychiatric Disorders in a Population-based Cohort of Older Adults
AID  - 10.1101/601609
DP  - 2019 Jan 01
TA  - bioRxiv
PG  - 601609
4099  - http://biorxiv.org/content/early/2019/04/07/601609.short
4100  - http://biorxiv.org/content/early/2019/04/07/601609.full
AB  - The underlying structure of psychiatric comorbidities has led researchers to investigate whether genetic factors can account for these patterns. Though psychiatric phenotypes are hypothesized to organize into a two-factor structure of internalizing and externalizing domains, few studies have evaluated the structure of psychopathology in older adults, and no studies have evaluated whether genome-wide polygenic scores (PGSs) organize in a similarly hierarchical structure. We used data from 7,157 individuals of European ancestry from the Health and Retirement Study, a large nationally-representative sample of older adults in the United States. Structural equation models utilized validated measures of psychopathology and genome-wide PGSs for multiple psychiatric outcomes. The data were best characterized by a two-factor internalizing-externalizing phenotypic model and a one-factor PGS model. The latent PGS factor (composed of PGSs for neuroticism, Major Depressive Disorder, anxiety disorders, Attention Deficit-Hyperactivity Disorder, and smoking) outperformed every individual PGS in predicting internalizing or externalizing outcomes, suggesting that future studies might construct a single latent PGS factor as a transdiagnostic measure of psychiatric genetic risk. However, neither the individual PGSs, nor the latent PGS factor exhibited specificity in their prediction of phenotypic outcomes, highlighting the tradeoff between polygenic population prediction and the utility of PGSs for clinical use. Several interpretations of the current results are provided: genetic risk for psychiatric disorders is transdiagnostic, GWAS-derived PGSs fail to capture genetic variation associated with disease specificity (e.g., rare variants), and blunt phenotypic measurement in GWAS precludes our ability to evaluate the structure and specificity of genetic contributions to psychiatric disorders.