DRIMSeq

DRIMSeq models counts of transcripts, exons, or exonic bins with a Dirichlet-multinomial distribution to detect differential isoform usage and identify SNP-associated splicing quantitative trait loci (sQTLs).

Key Features:

• Differential Splicing Analysis: Identifies changes in isoform usage between conditions by modeling multivariate counts of transcripts, exons, or exonic bins.
• sQTL Analysis: Tests associations between SNPs and multivariate feature counts to discover sQTLs that affect splicing.
• Statistical Framework: Employs a Dirichlet-multinomial joint model of isoform expression to capture correlated isoform counts and to account for differential gene expression while preserving overall gene abundance information.
• Robust Parameter Estimation: Shares information across features and samples to improve parameter estimates, particularly with limited replicates.
• Versatility: Applicable to other multivariate genomics scenarios such as Poly-A-seq and differential DNA methylation analyses where beta-binomial models are traditionally used.

Scientific Applications:

• Alternative Splicing Studies: Quantifies and compares isoform ratios between conditions (e.g., normal vs disease) to study regulatory and phenotypic consequences.
• sQTL Mapping: Maps SNPs that influence splicing to link genetic variation to changes in transcript expression.
• Phenotype and Prognostic Studies: Assesses relationships between isoform ratio changes and phenotypic outcomes or disease prognostics.
• Multivariate Genomics Assays: Applies the multivariate count framework to assays such as Poly-A-seq and to analyses of differential DNA methylation.

Methodology:

Uses the Dirichlet-multinomial distribution to model counts of transcripts, exons, or exonic bins, jointly models isoform expression to account for correlated counts and overall gene abundance, and shares information across features to obtain robust parameter estimates with limited replicates.