For each splicing event, the model performance was further evaluated by leave-one-out cross-validation, in which the model was established using n-1 samples and the splicing outcome of the nth sample was predicted. We calculated the Pearson’s correlation r between the cross-validated \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\hat{\mathrm \Psi }}}}$$\end{document}Ψ^ and the observed Ψ. The p-value of the correlation was used as an indicator of how reliably the genetic variants explained the Ψ to the extent measured by r. Importantly, since r is the correlation between prediction and observation, a zero or negative value indicates that the model is non-explanatory. Thus, to test whether the Ψ of each splicing event is explainable by genetic variants, the p-value was calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P\left( {H_0:r \le 0, \cdot H_a:r \, > \, 0} \right)$$\end{document}PH0:r≤0,⋅Ha:r>0. Any splicing event with a greater significance than the Bonferroni adjusted 5%-threshold was accepted for later analysis. In addition, the coefficient of determination, R2, was calculated as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2 =
