SparseIso

SparseIso applies a Bayesian approach to identify and quantify alternatively spliced isoforms from RNA-seq data for improved transcript-level inference.

Key Features:

• Bayesian framework: Implements a Bayesian model that incorporates a spike-and-slab prior to enforce sparsity in isoform selection and mitigate overfitting.
• Gibbs sampling: Uses a specialized Gibbs sampling procedure to estimate the joint distribution of candidate transcripts and to perform simultaneous identification and quantification.
• Low-expression detection: Enhances sensitivity for detecting lowly expressed transcripts and multiple isoforms within genes through its joint inference approach.
• Validation and performance: Demonstrates improved performance in transcript assembly and isoform identification based on simulation studies and real RNA-seq data analyses.

Scientific Applications:

• Transcriptome reconstruction: Reconstructs full transcriptomes from RNA-seq data to resolve alternatively spliced isoforms.
• Alternative splicing analysis: Detects and quantifies alternative splicing events at the isoform level.
• Gene expression studies: Supports genomic and molecular biology investigations that require isoform-level expression estimates, including studies of rare transcripts.

Methodology:

SparseIso performs Bayesian inference using a spike-and-slab prior and employs Gibbs sampling to estimate the joint distribution of candidate transcripts while simultaneously identifying and quantifying isoforms.