Computational analysis of single-cell alternative splicing
Alternative splicing (AS) generates isoform diversity critical for cellular identity and homeostasis in multicellular life. Although AS variation has been observed among single cells for a few events, little is known about the biological significance of such variation. We developed Expedition, a computational framework consisting of outrigger, a de novo splice graph transversal algorithm to detect AS; anchor, a Bayesian approach to assign modalities and bonvoyage, a visualization tool using non-negative matrix factorization to display modality changes. Applying Expedition to single iPSCs undergoing neuronal differentiation, we discover up to 20% of AS exons exhibit bimodality and are flanked by more conserved introns harboring distinct cis-regulatory motifs. Bimodal exons constitute the majority of cell-type specific splicing, are highly dynamic during cellular transitions, preserve translatability and reveal intricacy of cell states invisible to global gene expression analysis. Systematic AS characterization in single cells redefines our understanding of AS complexity in cell biology.