Nanopore sequencing enables the efficient and unbiased measurement of transcriptomes. Current methods for transcript identification and quantification rely on mapping reads to a reference genome, which precludes the study of species with a partial or missing reference or the identification of disease-specific transcripts not readily identifiable from a reference. We present RATTLE, a tool to perform reference-free reconstruction and quantification of transcripts using only Nanopore reads. Using simulated data and experimental data from isoform spike-ins, human tissues, and cell lines, we show that RATTLE accurately determines transcript sequences and their abundances, and shows good scalability with the number of transcripts.

We describe ReorientExpress, a method to perform reference-free orientation of transcriptomic long sequencing reads. ReorientExpress uses deep learning to correctly predict the orientation of the majority of reads, and in particular when trained on a closely related species or in combination with read clustering. ReorientExpress enables long-read transcriptomics in non-model organisms and samples without a genome reference without using additional technologies and is available at https://github.com/comprna/reorientexpress.

Single-molecule long-read sequencing provides an unprecedented opportunity to measure the transcriptome from any sample 1-3 . However, current methods for the analysis of transcriptomes from long reads rely on the comparison with a genome or transcriptome reference 2,4,5 , or use multiple sequencing technologies 6,7 .These approaches preclude the cost-effective study of species with no reference available, and the discovery of new genes and transcripts in individuals underrepresented in the reference. Methods for the assembly of DNA long-reads 8-10 cannot be directly transferred to transcriptomes since their consensus sequences lack the interpretability as genes with multiple transcript isoforms. To address these challenges, we have developed RATTLE, the first method for the reference-free reconstruction and quantification of transcripts from long reads. Using simulated data, transcript isoform spike-ins, and sequencing data from human and mouse tissues, we demonstrate that RATTLE accurately performs read clustering and error-correction.Furthermore, RATTLE predicts transcript sequences and their abundances with accuracy comparable to reference-based methods. RATTLE enables rapid and cost-effective long-read transcriptomics in any sample and any species, without the need of a genome or annotation reference and without using additional technologies.