RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome.

Li B, Dewey CN. · 2011

Read RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. by Li B, Dewey CN. (2011) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3163565.txt.

Introduction

Bo Li and Colin Dewey presented software for measuring gene activity from RNA sequencing in a 2011 paper in BMC Bioinformatics. The hard part is reads that match several genes or several versions of one gene, which naive counting misassigns. Their program distributes such shared reads by probability and reports uncertainty ranges for each estimate.

Uniquely, the tool needs no reference genome, so it pairs with freshly assembled transcript sets for species whose genomes were never sequenced. On simulated and real data it matched or beat genome-based rivals. Experiments with its simulation feature yielded design advice. Gene-level estimates favour many short single reads, while sorting versions within a gene can benefit from reads taken from both fragment ends. The package is user-friendly and free.

Catalog record

CREATORLi B, Dewey CN.
YEAR2011
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC3163565

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome By Bo Li, Colin N Dewey Published in: BMC Bioinformatics DOI: 10.1186/1471-2105-12-323 ABSTRACT Background RNA-Seq is revolutionizing the way transcript abundances are measured. A key challenge in transcript quantification from RNA-Seq data is the handling of reads that map to multiple genes or isoforms. This issue is particularly important for quantification with de novo transcriptome assemblies in the absence of sequenced genomes, as it is difficult to determine which transcripts are isoforms of the same gene. A second significant issue is the design of RNA-Seq experiments, in terms of the number of reads, read length, and whether reads come from one or both ends of cDNA fragments. Results We present RSEM, an user-friendly software package for quantifying gene and isoform abundances from single-end or paired-end RNA-Seq data. RSEM outputs abundance estimates, 95% credibility intervals, and visualization files and can also simulate RNA-Seq data. In contrast to other existing tools, the software does not require a reference genome. Thus, in combination with a de novo transcriptome assembler, RSEM enables accurate transcript quantification for species without sequenced genomes. On simulated and real data sets, RSEM has superior or comparable performance to quantification methods that rely on a reference genome. Taking advantage of RSEM's ability to effectively use ambiguously-mapping reads, we show that accurate gene-level abundance estimates are best obtained with large numbers of short single-end reads. On the other hand, estimates of the relative frequencies of isoforms within single genes may be improved through the use of paired-end reads, depending on the number of possible splice forms for each gene. Conclusions RSEM is an accurate and user-friendly software tool for quantifying transcript abundances from RNA-Seq data. As it does not rely on the existence of a reference genome, it is particularly useful for quantification with de novo transcriptome assemblies. In addition, RSEM has enabled valuable guidance for cost-efficient design of quantification experiments with RNA-Seq, which is currently relatively expensive. BACKGROUND RNA-Seq is a powerful technology for analyzing transcriptomes that is predicted to replace microarrays [1]. Leveraging recent advances in sequencing technology, RNA-Seq experiments produce millions of relatively short reads from the ends of cDNAs derived from fragments of sample RNA. The reads produced can be used for a number of transcriptome analyses, including transcript quantification [2-7], differential expression testing [8,9], reference-based gene annotation [6,10], and de novo transcript assembly [11,12]. In this paper we focus on the task of transcript quantification, which is the estimation of relative abundances, at both the gene and isoform levels. After sequencing, the quantification task typically involves two steps: (1) the mapping of reads to a reference genome or transcript set, and (2) the estimation of gene and isoform abundances based on the read mappings. A major complication in quantification is the fact that RNA-Seq reads do not always map uniquely to a single gene or isoform. Previously, we have shown that properly taking read mapping uncertainty into account with a statistical model is critical for achieving the most accurate abundance estimates [7]. In this paper, we present a user-friendly software package, RSEM (RNA-Seq by Expectation Maximization), which implements our quantification method and provides extensions to our original model. A key feature unique to RSEM is the lack of the requirement of a reference genome. Instead, it only requires the user to provide a set of reference transcript sequences, such as one produced by a de novo transcriptome assembler [11,12]. Extensions to our original methodology include the modeling of paired-end (PE) and variable-length reads, fragment length distributions, and quality scores. In addition, a 95% credibility interval (CI) and posterior mean estimate (PME) are now computed for the abundance of each gene and isoform, along with a maximum likelihood (ML) estimate. Lastly, RSEM now enables visualization of its output through probabilistically-weighted read alignments and read depth plots. Through experiments with simulated and real RNA-Seq data, we find that RSEM has superior or comparable quantification accuracy to other related methods. With additional experiments, we obtained two surprising results regarding the value of PE data and quality score information for estimating transcript abundances. Although a PE read provides more information than a single-end (SE) read, our experiments indicate that for the same sequencing throughput (in terms of the number of bases sequenced), short SE reads allow for the best quantification accuracy at the gene-level. And while one would assume that quality scores provide valuable information for the proper mapping of reads, we find that for RNA-Seq reads with Illumina-like error profiles, a model that takes into account quality scores does not significantly improve quantification accuracy over a model that only uses read sequences. Related work A simple quantification method that was used in some initial RNA-Seq papers [13,14] and that is still used today is to count the number of reads that map uniquely to each gene, possibly correcting a gene's count by the "mappability" of its sequence [15] and its length. The major problems with this type of method are that it: (1) throws away data and produces biased estimates if "mappability" is not taken into account, (2) produces incorrect estimates for alternatively-spliced genes [16], and (3) does not extend well to the task of estimating isoform abundances. A couple of methods were later developed that addressed the first problem by "rescuing" reads that mapped to multiple genes ("multireads") [17,18]. Some other methods addressed the latter two problems, but not the first, by modeling RNA-Seq data at the isoform level [5]. Later, we developed the methodology behind RSEM, which addressed all of these issues by using a generative model of RNA-Seq reads and the EM algorithm to estimate abundances at both the isoform and gene levels [7]. Since the publication of the RSEM methodology, a number of methods utilizing similar statistical methods have been developed [3,4,6,19-22]. Of the methods developed, only RSEM and IsoEM are capable of fully handling reads that map ambiguously between both isoforms and genes, which the authors of both methods have shown is important for achieving the best estimation accuracies [4,7]. In contrast with IsoEM, RSEM is capable of modeling non-uniform sequence-independent read start position distributions (RSPDs), such as 3'-biased distributions that are produced by some RNA-Seq protocols [1]. In addition, RSEM can compute PME and 95% CIs, whereas IsoEM only produces ML estimates. Lastly, RSEM is the only statistical method that we are aware of that is designed to work without a whole genome sequence, which allows for RNA-Seq analysis of species for which only transcript sequences are available. IMPLEMENTATION A typical run of RSEM consists of just two steps. First, a set of reference transcript sequences are generated and preprocessed for use by later RSEM steps. Second, a set of RNA-Seq reads are aligned to the reference transcripts and the resulting alignments are used to estimate abundances and their credibility intervals. The two steps are carried out by the user-friendly scripts rsem-prepare-reference and rsem-calculate-expression. The steps of the RSEM workflow are diagrammed in Figure 1 and described in more detail in the following sections. [Figure: The RSEM software workflow. The standard RSEM workflow (indicated by the solid arrows) consists of running just two programs (rsem-prepare-reference and rsem-calculate-expression), which automate the

Read the complete text