Fig. 2: Mendelian randomization (MR)-based analysis of COGA.
From: RNA alternative splicing impacts the risk for alcohol use disorder
A Overview of the MR-based analysis. Genetic variant (X) is the instrumental variable. The intermediate molecular trait (i.e., exposure) is the genetically predicted PSI, \({{{\hat{\mathrm \Psi }}}}_{\left( x \right)}\), for RNA splicing, and the phenotypic variable is the trait (Y). \({{{\hat{\mathrm \Psi }}}}\) is inferred from X using the elastic net (EN) models and the association between \({{{\hat{\mathrm \Psi }}}}\) and Y is evaluated by generalized estimating equation (GEE). Splicing events showing significant associations with the trait Y are putatively causal for the trait. This MR pipeline was run in the discovery cohort COGA and repeated in the replication cohort OZ-ALC. The number of subjects for each phenotype is provided. B, C Manhattan plots of significant splicing events. Chromosomal distribution of significance of association for all splicing events (\({{{\hat{\mathrm \Psi }}}}\), Y), with respect to the DSM-IV (B) and SXCT (C). Blue line, -log10 of the p-value corresponding to FDR = 0.05; green dots, significant events in discovery cohort; red dots, replicated significant events. D, E Effect sizes of the replicated events. Forest plots of the effect sizes of the six replicated events. The estimates of effect sizes (beta) in COGA (D) and OZ-ALC (E) are consistent. The rectangles are the estimates and hashed lines represent the 95% CI.
