Precise Transcript Reconstruction with End-Guided Assembly

Accurate annotation of transcript isoforms is crucial to understand gene functions, but automated methods for reconstructing full-length transcripts from RNA sequencing (RNA-seq) data remain imprecise. We developed Bookend, a software package for transcript assembly that incorporates data from different RNA-seq techniques, with a focus on identifying and utilizing RNA 5' and 3' ends. Through end-guided assembly with Bookend we demonstrate that correct modeling of transcript start and end sites is essential for precise transcript assembly. Furthermore, we discovered that utilization of end-labeled reads present in full-length single-cell RNA-seq (scRNA-seq) datasets dramatically improves the precision of transcript assembly in single cells. Finally, we show that hybrid assembly across short-read, long-read, and end-capture RNA-seq datasets from Arabidopsis, as well as meta-assembly of RNA-seq from single mouse embryonic stem cells (mESCs) can produce end-to-end transcript annotations of comparable quality to reference annotations in these model organisms. Overall design: Two PacBio Iso-seq libraries were generated each using 10 µg of total RNA from Arabidopsis inflorescences containing unopened floral buds. Total RNA was extracted with TRIzol following the method described in (Schon et al. 2018. Genome Research) to yield two biological replicates with an RNA integrity number (RIN) of 9.0 and 9.2, respectively. SMRTbell libraries were constructed by the Vienna BioCenter Core Facilities (VBCF) and sequenced on a Sequel SMRT Cell 1M.