Disentangling genetic nurture : novel methods and application to depression

Abstract

Recent interest has grown in detecting and estimating the effects of genetic nurture, the influence of parental genotypes above and beyond the direct effects of the variants they pass on to their offspring. Since genetic nurture is mediated through the parentally-provided environment, identifying such effects can be useful for informing prevention and intervention efforts aimed at reducing risk for negative outcomes in offspring across the lifespan. This thesis explores the statistical properties of genetic nurture models, introduces methodological improvements to increase power, and applies these methods to a range of phenotypes, with a focus on genetic nurture effects on depression. Chapter 2 highlights an improvement to recently proposed models of genetic nurture using polygenic scores (PGS). Instead of partitioning genetic variants into transmitted and non-transmitted PGS, we show that jointly modeling the PGS of offspring and parents produces increased power, reduced computational burden, and directly interpretable estimates of direct and indirect genetic effects. We apply this simplified modeling approach to longitudinal data on BMI trajectories across development, identifying a statistically significant interaction effect between offspring age and their mother’s PGS such that the influence of maternal genetic nurture on her offspring’s BMI increases over time. In Chapter 3, recognizing the common limitation that many family-based datasets only have genotype data on mother-offspring pairs, we then considered the consequences of genetic nurture models where data from only one parent is included. We show through derivation and simulation that even when assortative mating is assumed to be absent such that parental genotypes are uncorrelated, failure to model one parent can bias not only the child’s effect estimate, but the modeled parent’s estimate as well. Chapter 4 introduces a solution to the issue of missing parental data in nuclear family datasets, reducing bias and greatly expanding the data available for estimating genetic nurture. Combining Mendelian rules of inheritance and Bayes’ rule, we propose a method for imputing missing parental genotype data from available parent-offspring duos or sibling pairs. We show through derivation and simulation that such models are unbiased and apply our approach to estimate genome-wide direct and indirect effects on birthweight and BMI using data on families from the UK Biobank. In Chapter 5, we combine these methodological insights and approaches to explore the influence of genetic nurture on depression and neuroticism in the UK Biobank. We find evidence for the presence of genetic nurture effects on both depression and neuroticism. The combined effect size of the parental depression PGS on offspring depression is estimated to be about 35% the size of the offspring’s own PGS effect. We also find evidence for an effect of the maternal PGS for bipolar disorder which is nearly equivalent to the effect of the child’s bipolar PGS. Overall, this thesis contributes to the growing toolkit for researchers interested in disentangling direct and indirect genetic effects, while providing novel insights across a variety of offspring traits which are worthy of further study.