Deep indel mutagenesis reveals the regulatory and modulatory architecture of alternative exon splicing

While altered pre-mRNA splicing is a frequent mechanism by which genetic variants cause disease, the regulatory architecture of human exons remains poorly understood. Antisense oligonucleotides (AONs) that target pre-mRNA splicing have been approved as therapeutics for various pathologies including patient-customised treatments for rare diseases, but AON discovery is currently slow and expensive, limiting the wider adoption of the approach. Here we show that deep indel mutagenesis (DIM) –which can be made experimentally at very low cost – provides an efficient strategy to chart the regulatory landscape of human exons and rapidly identify candidate splicing-modulating oligonucleotides. DIM reveals autonomous effects of insertions, while systematic deletion scans delineate the checkerboard architecture of sequential enhancers and silencers in a model alternative exon. The results also suggest a mechanism for repression of transmembrane domain-encoding exons and for the generation of microexons. Leveraging deep learning tools, we provide a resource, DANGO, that predicts the splicing regulatory landscape of all human exons and can help to identify effective splicing-modulating antisense oligonucleotides. Altered pre-mRNA splicing frequently causes disease, yet how sequence variants alter splicing remains enigmatic. Here the authors use deep indel mutagenesis and deep learning tools to reveal the regulatory architecture of human exons and identify splicing-modulating antisense oligonucleotides.