Enhancing transcriptome expression quantification through accurate assignment of long RNA sequencing reads with TranSigner

Abstract

Recently developed long–read RNA sequencing technologies promise to provide a more accurate and comprehensive view of transcriptomes compared to short-read sequencers, primarily due to their capability to achieve full–length sequencing of transcripts. However, realizing this potential requires computational tools tailored to process long reads, which exhibit a higher error rate than short reads. Existing methods for assembling and quantifying long–read data often disagree on expressed transcripts and their abundance levels, leading researchers to lack confidence in the transcriptomes produced using this data. One approach to address the uncertainties in transcriptome assembly and quantification is by assigning reads to transcripts, enabling a more detailed characterization of transcript support at the read level. Here, we introduce TranSigner, a versatile tool that assigns long reads to any input transcriptome. TranSigner consists of three consecutive modules performing: read alignment to the given transcripts, transcripts, computation of compatibility scores based on alignment scores and positions, and execution of an expectation–maximization algorithm to probabilistically assign read fractions to transcripts while estimating transcript abundances. Using simulated and experimental datasets from three well studied organisms — Homo Sapiens, Arabidopsis thaliana and Mus musculus — we demonstrate that TranSigner achieves higher accuracy in transcript abundance estimation and read assignment compared to existing tools.

A replicable and modular benchmark for long-read transcript quantification methods

“…The results here also demonstrate that the tested methods display a wide spectrum of robustness to data quality and cleanness. The benchmark presented in this work, using the simulation protocol from Ji and Pertea (10), suggests that as the simulation setup changes, the absolute performance of different quantification tools changes considerably (e.g. lr-kallisto goes from performing similarly to other methods to considerably under-performing them).…”

“…We evaluate the above methods on 4 datasets. Two of these datasets are generated according to the methodology proposed by Ji and Pertea (10). Specifically, these datasets are generated with the NanoSim (11) simulator by learning the model parameters from experimental ONT sequencing datasets (both direct-RNA and 1D-cDNA) from Workman et al (12).…”

“…This is driven by several factors, including the (empirically derived) inclusion of "background" reads in the simulation, and higher error rates (though all model parameters are learned by NanoSim from the experimental data). We refer to these datasets as ts-dRNA and ts-cDNA, respectively, since the simulation follows the protocol that was originally introduced in the TranSigner paper (10). The next two datasets we consider were originally introduced in the IsoQuant paper (14) and recently used in the lr-kallisto paper (4).…”

“…S3. Notes with respect to evaluating bambu in the TranSigner-protocol simulations • In the RefSeq reference, we note some small discrepancies in the transcript set present in the available RNA fasta file 10 and what is present in the available GTF annotation 11 . Specifically, the latter has some extra transcripts, and additional variants of some transcripts present in the former.…”