"Developing polygenic risk scores to characterize a longitudinal phenotype"
Polygenic risk scores are commonly used to estimate the multi-gene effects on a single phenotype such as disease status in a case-control study. These scores are the weighted sums of individual single nucleotide polymorphism (SNP) effects used to predict the phenotype of interest. There has been little work on the estimation of polygenetic risk scores when the phenotype is a longitudinal trajectory. We develop a linear mixed modeling framework for estimating polygenic risk scores for characterizing the genetic effects on the baseline and trajectory of a longitudinal continuous trait. The sets of random effects are crossed since the genetic effects vary over genome-location and the longitudinal effects vary over individual. We propose an EM algorithmic approach for parameter estimation, discuss computational challenges, and consider robustness of the model to key assumptions. We illustrate the methodology by examining the genetic effects on the prostate-specific antigen (PSA) level trajectory of male controls from the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial.
Additional authors: Doron Levy