DEXSeq

DEXSeq detects differential exon usage from RNA-seq exon counts to identify exon-level changes across experimental conditions and species.

Key Features:

• Input data: Operates on RNA-seq exon counts to perform exon-level analyses.
• Statistical modeling: Models exon counts with a negative binomial distribution to estimate variance among biological replicates.
• Inference framework: Uses generalized linear models to test for differences in exon usage between samples.
• Dual-mode differential usage model: Distinguishes "Weak Differences" where most exons show minor, likely neutral variability, from "Strong Differences" where a subset shows conserved, tissue-specific usage.
• Conserved exon identification: Identifies approximately 3,800 strongly conserved, tissue-dependent exons from around 1,643 genes reported conserved from humans to macaques.
• Comparative orthology: Targets 1:1 orthologous exons across primate species for cross-species comparisons.
• Enrichment analysis: Finds conserved exons enriched for regions encoding protein-disordered areas and untranslated segments.
• Genome-wide multi-tissue/species analysis: Enables transcriptome-wide surveys across multiple tissues and species.
• Result visualization: Provides visualization capabilities for exploring and interpreting exon usage results.

Scientific Applications:

• Alternative splicing analysis: Detects exon-level changes that reflect alternative isoform usage and its effects on transcript structure.
• Comparative evo-devo studies: Compares exon usage across primate species to investigate conserved and nonconserved patterns of isoform regulation.
• Functional inference: Identifies tissue-specific exons and candidate isoforms potentially linked to protein disorder, untranslated regions, and regulatory functions.
• Transcriptome-wide surveys: Performs genome-wide analyses of exon usage across multiple tissues and species to assess the prevalence and conservation of alternative exon usage.

Methodology:

Uses RNA-seq exon counts, models counts with a negative binomial distribution to estimate replicate variance, applies generalized linear models to test exon usage differences, employs a dual-mode model to classify weak versus strong differences, and analyzes 1:1 orthologous exons for cross-species comparisons.