All Times ET
Keywords: Genetic correlation, parametric bootstrap, heritability, confidence interval, linear mixed model
Genetics is one of the important factors in determining human phenotypes and the correlations between them. Heritability is the proportion of phenotypic variance in a population that is attributable to individual genotypes, whereas genetic correlation is the genetic component of phenotypic correlation. Understanding genetic correlations helps uncover gene functions and disease mechanisms, improve diagnosis and aid therapeutic interventions. Heritability and genetic correlation can be estimated when genetic relationship between individuals in a sample is available. The distributions of both heritability and genetic correlation estimates are typically non-normal and cannot be well-approximated by normal distributions near the boundaries of the parameter space. Several studies have investigated methods to construct confidence intervals (CI) for heritabilities. Here, we extend this research into constructing the CIs for genetic correlations. In this paper, we propose to construct CIs for genetic correlations via parametric bootstrap. Briefly, we first simulate N pairs of phenotype vectors under every potential scenario of true heritability and genetic correlation while relying on the genetic relationship between individuals in the sample. Second, we estimate the genetic correlation coefficients for each of these N pairs based on a closed form solution within the Haseman-Elston method-of-moment regression framework, and obtain a probability mass function from these N estimates. We then utilize Bayes theorem to reverse the conditions from conditioning on true genetic correlation to conditioning on estimated genetic correlation, from which we can construct CIs given any real-world estimates of genetic correlation in the sample. Our proposed parametric bootstrap method could yield more accurate CI estimates compared to the current Fisher transformation method, and at the same time retain computational efficiency when computing a high-dimensional set of genetic correlation.