Merging short and stranded long reads improves transcript assembly

New tools for improved long-read transcript assembly and coalescence with its short-read counterpart are required. Using our short- and long-read measurements from different cell lines with spiked-in standards, we systematically compared key parameters and biases in the read alignment and assembly of transcripts. We report a cDNA synthesis artifact in long-read datasets that impacts the identity and quantitation of assembled transcripts. We developed a computational pipeline to strand long-read cDNA libraries that markedly improves assembly of transcripts from long-reads. Incorporating stranded long-reads in a new hybrid assembly approach, we demonstrate its efficacy for improved characterization of challenging lncRNA transcripts. Our workflow can be applied to a wide range of transcriptomics datasets for superior demarcation of transcript ends and refined isoform structure, which can enable better differential gene expression analyses and molecular manipulations of transcripts. Nuclei from HL1 cells were isolated by detergent lysis, fractionated, then three chromatin and three nucleplasm RNA samples were converted to cDNA using NEBNext Ultra II Directional library kit, and sequenced on Illumina HiSeq 4000. In parallel, chromatin fraction RNA from two replicates were converted to cDNA using Oxford Nanopore direct cDNA sequencing kit and sequenced on MinION flowcell.