IsoEM

IsoEM estimates isoform-specific expression levels from RNA-Seq data using an expectation-maximization algorithm that leverages insert-size distributions to resolve reads mapping to alternative splicing gene isoforms.

Key Features:

• Expectation-Maximization algorithm: Uses an EM algorithm tailored to infer isoform and gene-specific expression levels from RNA-Seq reads.
• Insert-size distribution disambiguation: Exploits the distribution of insert sizes from sequencing library preparation to disambiguate reads that map to multiple isoforms.
• Integration of sequencing metadata: Incorporates base quality scores, strand specificity, and read pairing information when available to refine expression estimates.
• Short-read RNA-Seq focus: Addresses challenges associated with short read lengths produced by current sequencing technologies.
• Empirical validation and scalability: Performance has been evaluated on synthetic and real RNA-Seq datasets, demonstrating scalability and improved accuracy over existing methods.

Scientific Applications:

• Alternative splicing analysis: Enables estimation of expression levels for alternative splicing isoforms to study splicing variation.
• Gene and isoform expression profiling: Supports gene-level and isoform-level expression quantification from RNA-Seq experiments.

Methodology:

Processes RNA-Seq read data including insert sizes and base quality scores, and applies iterative expectation and maximization cycles that use insert-size distributions and read pairing/strand information to refine isoform expression estimates and resolve multi-mapping reads.