Mendelian randomization is commonly used to infer causal relationships among complex traits, but the approach is confounded by pleiotropic effects. Here the authors develop a new statistical approach that can distinguish between genetic correlation and full or partial genetic causation. They apply both their model and several MR methods to GWAS summary statistics for 52 diseases and complex traits and show that their algorithm decreases the number of false positive causal relationships identified.
Distinguishing causality from correlation
A latent causal variable model improves power to detect causal relationships among genetically correlated pairs of complex traits, compared with Mendelian Randomization
A monthly journal covering the science and business of biotechnology, with new concepts in technology/methodology of relevance to the biological, biomedical, agricultural and environmental sciences.