NBAMSeq

NBAMSeq implements a generalized additive model framework to perform differential expression analysis of RNA-Seq gene counts by modeling the logarithm of mean counts as sums of smooth functions to capture linear and nonlinear associations with covariates.

Key Features:

• Generalized Additive Model (GAM) Framework: Models the logarithm of mean gene counts as sums of smooth functions to allow complex linear and nonlinear relationships between gene expression and covariates.
• Information Sharing Across Genes: Uses Bayesian shrinkage to share information across genes for improved variance estimation.
• Simultaneous Estimation of Parameters: Employs a nested iterative method to simultaneously estimate smoothing parameters and coefficients.
• Detection of Nonlinear Effects: Demonstrates superior performance in detecting nonlinear effects while maintaining equivalent performance for linear effects according to simulations and case studies.

Scientific Applications:

• Differential Expression Analysis: Identifies genes with linear and nonlinear associations to covariates in RNA-Seq datasets.
• Genomic Biomarker Discovery: Supports detection of disease-related gene expression patterns for biomarker identification.
• Studies of Nonlinear Phenotypes: Applicable to analyses where phenotypes exhibit nonlinear relationships with gene expression.

Methodology:

Model the logarithm of mean gene counts as sums of smooth functions via a generalized additive model; apply Bayesian shrinkage for variance estimation by sharing information across genes; and use a nested iterative method to simultaneously estimate smoothing parameters and coefficients.