Interpretable machine learning enables de novo mapping of cell type-specific RNA splicing regulation from scRNA-seq data
TL;DR
CASREL is developed, a machine learning framework that infers candidate regulatory circuitry between RNA-binding proteins and alternative splicing from single-cell data, enabling de novo mapping of putative cell-specific regulatory networks.
Abstract
Context-dependent regulation of alternative splicing, largely mediated by RNA-binding proteins, is a key post-transcriptional mechanism shaping diverse biological processes. Experimental approaches for probing splicing regulation, such as crosslinking-immunoprecipitation assays and RNA-binding protein perturbations, suffer from low throughput, poor physiological relevance, and bulk resolution that overlooks cellular heterogeneity. Here, we present CASREL, a machine learning framework that reconstructs candidate regulatory circuitry between RNA-binding proteins and alternative splicing directly from single-cell RNA sequencing data without reliance on prior protein-RNA binding annotations. CASREL integrates ensemble learning with model interpretation based on Shapley additive explanations to infer associations between RNA-binding proteins and alternative splicing, leveraging distinctive features of single-cell splicing profiles, including polarized isoform usage, minimal averaging of regulatory programs, and resilience to expression noise. Applications across diverse tissues and cell types demonstrate its robustness, accuracy, and biological relevance, supporting CASREL as an effective method for de novo mapping of putative cell-specific RNA splicing regulation in physiological contexts. The authors develop CASREL, a machine learning framework that infers candidate regulatory circuitry between RNA-binding proteins and alternative splicing from single-cell data, enabling de novo mapping of putative cell-specific regulatory networks.