Abstract

How different methodologies in the construction of various Polygenic Risk Scores affects the ability to explain mood symptoms

Khalil H, Turner CA, Murphy-Weinberg V, Gates L, Grysko R, Arakawa K, Dokas T, Fang Y, Lopez J, Sen S, Watsons SJ, Akil H
51st Annual Meeting of the Society for Neuroscience. 2022.

Abstract

Methods for generating Polygenic Risk Scores (PRS) fall into two broad categories based on methods for Linkage Disequilibrium (LD), namely clumping and beta shrinkage. Within these two broad categories, there are a number of smaller decision points which can also lead to very different results. In this study we aim to highlight how these decision points affect the final PRS values and what impact this has on understanding the role of genetics in partly explaining mood. We detail the construction of the PRS, from the quality control on our own data to the use of imputation (or not) and the choice of method to calculate the PRS. We have longitudinal data on a sample of University of Michigan freshmen, from 2015 to 2021. The freshmen were genotyped and characterized at baseline using clinical variables. This included anxiety (GAD-7) and depression (PHQ-9) symptoms, as well as measures of neuroticism, childhood trauma, resilience, social support / isolation, and positive and negative suicide ideation. Their daily physical activity was captured using Fitbits, and their mood levels were sampled at multiple timepoints throughout the freshmen year. The breadth and longitudinal nature of this data means that we can use it to test PRSs for impulsivity, neuroticism, anxiety, depression and circadian rhythm. We can also test how the methods for calculating the PRS changes its explanatory power. As a test case, the PRS for major depressive disorder (MDD-PRS) was calculated using clumping and thresholding. The PRS with no p-value thresholding and using imputed genetic data gave us the most explanatory power. Using our own data, we will show how much the explanatory power of a PRS changes given how it is generated and how this applies to different PRSs.