For the gene ELOVL7, we extracted all genetic variants (x) located in the transcribed region (from TSS to 3’UTR) from the CMC genotyped subjects. The effect size βxy of each variant (x) on the AUD trait (y) was retrieved from the GWAS summary statistics. The effect size βxΨ of each variant (x) on PSI (Ψ) of the ELOVL7 exon skipping event was calculated based on CMC data, using a linear regression model adjusted for the demographic covariates of sex and ethnic group. To infer the causality of splicing (Ψ) on trait (y), we co-localized βxy with βxΨ by Generalized Summary data-based Mendelian Randomization (GSMR), in which the causal effect size of Ψ on y, i.e., \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat \beta _{{{\Psi }}y}$$\end{document}β^Ψy was estimated by a least-square (LS) regression model [43].
