Widespread naturally variable human exons aid genetic interpretation

Most mammalian genes undergo alternative splicing. The splicing of some exons has been acquired or lost in specific mammalian lineages, but differences in splicing within the human population are poorly characterized. Using GTEx tissue transcriptomes from 838 individuals, we identified 56,415 exons which are included in mRNAs in some individuals but entirely excluded from others, which we term 'naturally variable exons' (NVEs). NVEs impact three quarters of protein-coding genes, occur at all population frequencies, and are often absent from reference annotations. NVEs are more abundant in genes depleted of genetic loss-of-function mutations and aid in the interpretation of causal genetic variants. Genetic variants modulate the splicing of many NVEs, and 5'UTR and coding-region NVEs are often associated with increased and decreased gene expression, respectively. Together, our findings characterize abundant splicing variation in the human population, with implications for a range of human genetic analyses.