This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| RSEM | |
|---|---|
| Name | RSEM |
| Author | Bo Li |
| Developer | University of California, Berkeley |
| Released | 2010 |
| Latest release | 1.3.0 (example) |
| Programming language | C++, Perl |
| Operating system | Unix-like |
| License | Academic |
RSEM
RSEM is a computational tool for estimating gene and isoform expression levels from RNA-Seq data. It was developed to model ambiguity in read mapping and to produce transcript abundance estimates suitable for downstream analysis in molecular biology, genomics, and biomedical research. RSEM has been cited across studies involving human, mouse, and plant transcriptomes and is commonly used alongside alignment, differential expression, and pathway analysis pipelines.
RSEM was introduced to address challenges in quantifying transcript-level expression from short-read sequencing platforms such as Illumina. Its model explicitly accounts for multi-mapping reads, fragment length distributions, and sequencing errors to provide maximum likelihood and posterior estimates of transcripts. RSEM is frequently used by research groups at institutions like Broad Institute, Wellcome Trust Sanger Institute, Johns Hopkins University, Stanford University, and Harvard University and appears in large consortia workflows including ENCODE Project, GTEx Project, 1000 Genomes Project, The Cancer Genome Atlas.
RSEM employs a generative model of RNA-Seq fragment generation and a statistical inference procedure to estimate expression. It uses the Expectation-Maximization (EM) algorithm to assign probabilistic weights to reads that map to multiple transcripts, iteratively updating transcript abundance estimates until convergence. The model incorporates transcript length correction and fragment length distribution estimation, and it can operate with or without a reference genome by using transcriptome assemblies from tools like Trinity (software), Cufflinks, or StringTie. RSEM supports input from aligners such as Bowtie, Bowtie 2, and STAR and integrates with quantification frameworks like DESeq2, edgeR, and limma for downstream differential expression testing.
RSEM accepts aligned read files in formats produced by common aligners: SAM, BAM, or aligner-specific outputs when run in its own alignment mode via Bowtie/Bowtie 2. It also accepts raw FASTQ reads when configured to perform alignment internally. Output includes gene- and isoform-level expected counts, TPM (Transcripts Per Million), and FPKM (Fragments Per Kilobase per Million) values, as well as posterior variance estimates and model diagnostics. Typical outputs are plain text tables compatible with platforms such as GenePattern, Galaxy (computational biology platform), and R-based analysis environments like Bioconductor.
RSEM's probabilistic assignment of multi-mapping reads generally improves accuracy over simple counting methods in benchmarking studies. Evaluations against simulated datasets and spike-in controls (e.g., ERCC controls used by groups at National Institutes of Health laboratories) demonstrate robust estimation of isoform abundance under varying read lengths and fragment sizes. However, runtime and memory usage scale with transcriptome complexity and sequencing depth; in practice, users combine RSEM with fast aligners such as STAR for large datasets like those from Sequence Read Archive studies. Comparative benchmarks often report favorable accuracy for transcript-level estimates, though gene-level aggregation may yield similar results across several tools.
RSEM is applied in diverse contexts: characterization of alternative splicing in model organisms from European Molecular Biology Laboratory collaborations, expression profiling in cancer cohorts from Memorial Sloan Kettering Cancer Center and Dana-Farber Cancer Institute, and quantitative analyses in developmental biology studies at institutions like Max Planck Society. It underpins workflows for biomarker discovery, expression quantitative trait loci (eQTL) mapping in projects similar to GTEx Project, and integration with pathway analysis using resources such as KEGG, Reactome, and Gene Ontology annotations curated by groups like UniProt.
RSEM is often compared to methods that either align-then-count or use pseudoalignment strategies. Popular alternatives include alignment-based quantifiers like Cufflinks and alignment-free or quasi-mapping tools such as Salmon (software) and Kallisto. Compared to Kallisto and Salmon (software), RSEM traditionally requires explicit alignments and is computationally heavier but provides detailed posterior estimates and variance measures. Against Cufflinks and older pipelines from groups like Tuxedo suite, RSEM's probabilistic read assignment and explicit modeling of fragment length distribution often yield improved isoform-level accuracy.
Critiques of RSEM focus on computational demands—CPU time and memory—especially for large eukaryotic transcriptomes, where tools like Kallisto and Salmon (software) offer substantial speedups. RSEM's reliance on an accurate reference transcriptome means assembly errors or incomplete annotations from databases like RefSeq, Ensembl, or GENCODE can bias estimates. Additionally, challenges remain in resolving highly similar paralogs and distinguishing transcriptional from post-transcriptional regulation without complementary experiments from laboratories such as EMBO-affiliated groups. Users must also consider batch effects documented in consortium studies like ENCODE Project and control for library preparation variability reported by sequencing centers at Illumina and others.
Category:Bioinformatics software