Building a quantitative and predictive model of 5SS selection by human U1 snRNP using RNA-map

Alternative splicing of pre-mRNA greatly expands the protein coding potential of our genome and accounts for extensive cell and tissue-specific variation in gene expression. Spliceosome assembly depends on the recognition of the 5′ splice site (5′SS) at the exon-intron boundary by the U1 small nuclear ribonucleoprotein (U1 snRNP). Critically, mutations in either U1 or the 5′SS can lead to human disease, including cancer. Despite decades of investigation, accurately predicting 5′SS selection and alternative splicing outcomes is difficult, suggesting new approaches are needed. Here, we present our progress in developing a quantitative and predictive model of 5′SS selection by human U1 snRNP grounded in biochemical understanding. We used the RNA on a massively parallel array (RNA-MaP) platform to measure equilibrium binding constants of reconstituted human U1 snRNP to >60,000 unique RNAs. Our designed RNA library features all possible GU and GC variants in addition to thousands of identified mutations, allowing us to build a comprehensive and predictive model of 5′SS/U1 interactions at the transcriptome scale. Overall, our results will provide a first principles understanding of 5′SS selection and will serve as a foundation for future work to systematically determine the effects of other splicing factors and splice site usage in vivo.