TITLE: “Fast estimation of genetic correlation for Biobank-scale data”
ABSTRACT: Genetic correlation is an important parameter in understanding the shared genetic basis across pairs of complex traits with applications ranging across disease subtyping, genetic prediction, and causal inference. The availability of genome-wide genetic data has led to a number of methods that aim to estimate genetic correlation. Methods that analyze individual genotype data (typically using a bi-variate linear mixed model) are computationally expensive to be applied to large-scale datasets such as the UK Biobank. In contrast, methods that use GWAS summary statistics, such as LD-score regression (LDSC) and high-definition likelihood (HDL), are computationally efficient but tend to have large standard errors. Thus, it is critical to develop methods that can accurately estimate genetic correlation from large individual-level datasets.
In this talk, I will present on SCORE (SCalable genetic CORrelation Estimator), a randomized algorithm to estimate the genetic correlation of traits using individual-level genotypes that can scale to UK Biobank-size datasets.