To quantify the component of PSI (Ψ) determined by cis-acting genetic variants, we established a computational model based on the genotype and PSI derived from RNA-seq data of 991 CMC subjects [21]. For modeling, we required that each splicing event must have: (i) more than one SNV located in the transcribed region of the gene; (ii) number of support samples ≥100; and (iii) interquartile range (IQR) of the calculated Ψ across all support samples >10%. Specifically, support samples were defined as those having genotype information available and total junction read counts (including both inclusion and exclusion events) ≥10. For each splicing event, we retrieved all SNVs within the transcribed region spanning from the transcription start site (TSS) to the end of 3’-untranslated region (UTR) of the host gene with MAF ≥ 0.01 and genotype imputation score ≥0.6. To maximize the probability that the SNVs included in each model were informative and to reduce the computation complexity for events with large numbers of SNVs in the transcribed region, we selected up to 20 top SNVs based on the ranking of their
