Assembly and quantification of transcripts from noisy long reads with NIFFLR

Background Long-read RNA sequencing technologies can produce complete or near-complete transcript sequences. Recently introduced methods for direct RNA and cDNA sequencing can provide a high-throughput strategy for the discovery of novel and rare gene isoforms. However, the high error rates in ONT sequences limit the ability to exactly pinpoint splice site boundaries when aligning reads to the genome. Methods In this paper, we present a novel tool called NIFFLR (Novel IsoForm Finder using Long Reads) that identifies and quantifies both known and novel isoforms using long-read RNA sequencing data. NIFFLR recovers known transcripts and assembles novel transcripts present in the data by aligning exons from a reference annotation to the long reads. Results NIFFLR effectively recovers correct transcripts from simulated reads based on known transcript annotations, achieving higher sensitivity and precision compared to several previously published tools. On real data, NIFFLR shows high accuracy as measured by concordance of isoform counts to the counts computed from Illumina data for the same sample. We applied NIFFLR to a set of 92 GTEx long-read samples and produced transcript counts for both novel and known isoforms. In total, we identified and quantified 119,928 isoforms present in the RefSeq annotation of GRCh38 and 42,868 novel isoforms across 10,487 genes, more than previous studies identified in this dataset. Conclusions NIFFLR is an effective tool aimed at assembly and quantification of transcripts present in the long, high error transcriptome reads. NIFFLR is released under an open-source license (GPL 3.0) and is available on GitHub at https://github.com/alguoo314/NIFFLR/releases, and from BioConda as “nifflr”.