Genetic Biomarkers of Psychiatric Disease: Discriminating Bipolar Disorder From Major Depressive Disorder With Polygenic Risk Scores

Abstract

Polygenic risk summary scores (PRSs) based on genome-wide association studies for psychiatric diseases have shown promise in discriminating cases from controls, predicting disease course, and response to treatment. The research presented here has sought to investigate whether such an approach may improve differentiation between bipolar disorder (BD) and major depressive disorder (MDD), a clinical distinction with important prognostic and therapeutic implications, and one which is often challenging based on clinical grounds alone. More specifically, this study tested whether psychiatric polygenic risk scores (PRSs) for bipolar disorder (BD) and schizophrenia (SCZ) improve clinically based classification models of BD-MDD diagnosis. The test sample in the present analysis included 843 BD and 930 MDD subjects similarly genotyped and phenotyped using the same standardized interview. The primary analysis tested the association of clinical risk factors and PRSs with diagnosis of BD versus MDD. A secondary analysis used multivariate modeling and receiver operating characteristic analysis to test the additive effect of PRSs on a baseline model with clinical features known to associate with BD-MDD status. PRSs drawn from BD (R2=3.5%, p=4.94x10-12) and SCZ (R2 = 3.2%, p=5.71x10-11) GWAS meta-analyses associated with BD-MDD diagnosis. Individuals with top decile BD PRS had a significantly increased risk for BD versus MDD compared with those in the lowest decile (OR=3.39, CI=2.19-5.25). PRSs improved the discriminatory ability of a symptom-based model (DC=0.021, p=1.05x10-4) and a full model with symptoms and clinical features (DC = 0.011, p=6.48x10-4). This study demonstrates that psychiatric PRSs provide modest independent discrimination between BD and MDD cases, suggesting that PRS could ultimately have utility in subjects at the extremes of the distribution and subjects for whom clinical symptoms are poorly measured or yet to manifest.