Inferring direct genetic effects in family-based designs

Sanchez Ortega, Marco Antonio (2022) Inferring direct genetic effects in family-based designs. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (655kB)

Abstract

In some genetic association studies, the same genetic loci have been found to be associated with different complex phenotypes. Such associations may be induced by direct genetic effects or indirect genetic effects through some intermediate phenotypes. In order to make valid statistical inference on direct genetic effects on a target phenotype, it is important to distinguish the causative genetic associations. A directed acyclic graph (DAG) can be used to describe possible relationships between genetic variants, primary and intermediate phenotypes, and confounding factors. There are several statistical methods for the aim of estimating direct genetic effects. This study mainly focuses on a novel approach called causal inference based on estimating equations (CIEE) with robust Huber-White sandwich variance estimator. The primary objective of this dissertation is to provide an extension of the CIEE method to family-based designs. Essentially, we are interested in estimating and testing direct genetic effects when some dependence exists among pairs of family members. We consider two statistical models for the analysis of continuous primary phenotypes, which are either completely observed or subject to censoring. In the first modelling approach, we assume independence between phenotypes of family members given measured factors and other phenotypes, while in our second approach the dependence between family members is modelled by a copula function. We formulate unbiased estimating equations to remove the indirect effect and obtain the direct genetic effect. The standard error of the direct effect estimator is obtained by using the so-called robust Huber-White sandwich variance estimator. We use large-sample Wald-type test statistic for testing the absence of direct genetic effect on the target phenotype. We evaluate the performance of our method by conducting Monte Carlo simulation studies for the analysis of quantitative and time-to- event primary traits. The results show that both methods are competitive and provide unbiased direct effect estimates and valid hypothesis testing. However, in general, the CIEE method under the copula model has better performance than the working independence model. It yields more accurate point estimates, valid empirical type I error rates for testing the absence of direct effect and higher empirical power rates across the considered simulation scenarios.

Actions (login required)

View Item