Abstract
RNA-seq protocols that focus on transcript termini are well suited for applications in which template quantity is limiting. Here we show that, when applied to end-sequencing data, analytical methods designed for global RNA-seq produce computational artifacts. To remedy this, we created the End Sequence Analysis Toolkit (ESAT). As a test, we first compared end-sequencing and bulk RNA-seq using RNA from dendritic cells stimulated with lipopolysaccharide (LPS). As predicted by the telescripting model for transcriptional bursts, ESAT detected an LPS-stimulated shift to shorter 3'-isoforms that was not evident by conventional computational methods. Then, droplet-based microfluidics was used to generate 1000 cDNA libraries, each from an individual pancreatic islet cell. ESAT identified nine distinct cell types, three distinct β-cell types, and a complex interplay between hormone secretion and vascularization. ESAT, then, offers a much-needed and generally applicable computational pipeline for either bulk or single-cell RNA end-sequencing.