RSEM

RSEM estimates gene and isoform expression from RNA-Seq reads using a generative statistical model that accounts for read mapping uncertainty.

Key Features:

• Handling Read Mapping Uncertainty: Uses a probabilistic generative model to incorporate reads that map to multiple genes or isoforms rather than discarding ambiguous reads or assigning them heuristically.
• Gene Expression Estimation: Reports gene-level expression as the sum of estimated isoform expression levels to reflect isoform contributions to gene activity.
• Modeling Non-Uniform Read Distributions: Models non-uniform read distributions across transcripts to improve accuracy of expression estimates.
• Optimal Read Length Determination: Uses simulations parameterized with real RNA-Seq data to identify that 20–25 base read lengths are optimal for gene-level estimation in mouse and maize under fixed sequencing throughput.

Scientific Applications:

• Transcriptomics studies: Quantifies isoform and gene expression from RNA-Seq datasets for transcriptome characterization.
• Differential expression analysis: Provides expression estimates suitable for comparing gene and isoform abundance across conditions.
• Functional genomics: Supplies quantified expression inputs for downstream functional and regulatory analyses across organisms including mouse and maize.

Methodology:

Implements a generative statistical framework with an expectation-maximization algorithm to infer isoform and gene expression, models non-uniform read distributions across transcripts, and performs simulations parameterized with real RNA-Seq data for read-length evaluation.