We established new splicing models to use with classic MR in identifying the specific splicing events instead of conducting sQTL analysis together with commonly used summary-based approaches as used in previous studies [13, 18–20]. Our strategy provides greater precision and specificity in terms of selecting the explanatory SNVs and dissecting the genetically determined component of RNA splicing. On one hand, summary-based studies infer a causal splicing event by co-localizing the SNV (Z)-splicing (X) association βzx and the SNV (Z)-trait (Y) association βzy. In such approaches, the explanatory SNVs for splicing outcome are not identified, and the genetic component of the splicing outcome cannot be directly assessed. As a result, a challenge arises in evaluating the true validity of the model. In this regard, models established herein are verifiable directly through either heritability analysis or an independent RNA-seq dataset. On the other hand, identifying the SNVs responsible for alternative splicing events that impact disease susceptibility enables prediction of the disease risk based on individual genotyping information. Therefore, these models described herein facilitate future studies on personalized health care including AUD.
