tradeSeq

tradeSeq performs trajectory-based differential expression analysis of single-cell RNA-seq (scRNA-seq) data to identify genes that change along inferred cellular trajectories and between distinct lineages.

Key Features:

• Generalized additive model (GAM) with negative binomial: Implements a GAM framework using a negative binomial distribution to model count-based scRNA-seq data.
• Overdispersion handling: Accounts for overdispersion common in count-based gene expression data.
• Non-linear trajectory modeling: Fits regression models that capture complex, non-linear relationships between gene expression and cellular states along inferred trajectories.
• Within-lineage differential expression: Tests for genes whose expression changes as cells progress within a single lineage.
• Between-lineage differential expression: Tests for genes whose expression differs between distinct lineages.
• Observation-level weights and zero inflation: Incorporates observation-level weights to address zero inflation in scRNA-seq data.

Scientific Applications:

• Developmental biology: Identify genes with dynamic expression across differentiation trajectories to study developmental processes.
• Disease progression: Detect lineage-specific or trajectory-associated expression changes relevant to disease progression.
• Cellular response studies: Characterize gene expression dynamics in cellular responses to environmental changes.
• Regulatory gene and pathway discovery: Reveal key regulatory genes and pathways by pinpointing trajectory-associated and between-lineage differential expression, including patterns missed by traditional methods.
• Data types: Applied to simulated and real datasets from droplet-based and full-length single-cell RNA sequencing protocols.

Methodology:

Fits generalized additive models (GAMs) with a negative binomial likelihood, uses regression-based modeling to capture non-linear expression–state relationships along trajectories, and incorporates observation-level weights to address zero inflation and overdispersion.