45th European Mathematical Genetics Meeting (EMGM) 2017, Tartu, Estonia, April 4 7, 2017: Abstracts

Recently, various novel statistical methods have been developed to work directly on summary statistics from Genome-Wide Association Studies (GWAS). This is a promising approach to utilize the increasing GWAS sample sizes while avoiding privacy concerns and logistics of sharing individual-level genotype data. Examples include estimation of heritability and genetic correlations, gene-level tests, risk prediction, z-score imputation and fine-mapping of causal variants.In addition to GWAS results, these approaches require estimates of genotype-genotype and/or phenotype-phenotype covariance structures to properly account for dependences between variables. Hence, key questions are from where to take these covariance estimates and how do they perform in practice. I consider these questions in the context of our recent multivariate methods metaCCA and FINEMAP.