SDEAP

SDEAP implements splice-graph-based differential transcript expression (DTE) analysis on population data by estimating latent biological conditions and detecting alternative splicing to enable biomarker discovery such as cancer subtypes and cell-cycle phases.

Key Features:

• Condition Estimation: Estimates the number of latent biological conditions directly from input samples using a Dirichlet mixture model, avoiding predefined condition specification.
• Splice Graph Representation: Employs splice graphs as the underlying data structure for transcript-level analysis.
• Alternative Splicing Analysis: Identifies alternative splicing events via a graph modular decomposition algorithm, extending analysis beyond differential exon usage.
• Performance and Validation: Validated on simulated data and real datasets with qPCR and evaluated against other DTE methods for tasks including cancer subtype and cell-cycle phase prediction.

Scientific Applications:

• Cancer subtype classification: Facilitates discovery of biomarkers that classify cancer samples into previously unknown subtypes.
• Alternative splicing studies: Enables detection and interpretation of alternative splicing patterns across population samples beyond differential exon usage.
• Cell-cycle and differentiation analysis: Supports prediction of cell-cycle phases and studies of cellular differentiation and cancer heterogeneity.
• Biomarker discovery for personalized medicine: Assists identification of transcriptomic biomarkers relevant to personalized diagnostic and therapeutic strategies.

Methodology:

SDEAP uses splice graphs with a Dirichlet mixture model to estimate the number of latent conditions and a graph modular decomposition algorithm to identify alternative splicing events.