bridge

bridge performs robust Bayesian hierarchical modeling to identify differentially expressed genes from gene expression microarray data.

Key Features:

• Outlier Management: Models microarray measurement errors and outliers explicitly using a t-distribution to account for experimental imperfections.
• Hierarchical Bayesian Framework: Uses an exchangeable prior for gene variances to allow gene-specific variances while shrinking extreme empirical variances toward shared values.
• Multiple-Sample Analysis: Tests differential expression among multiple samples and distinguishes different patterns of differential expression when three or more samples are compared.
• Parameter Estimation: Employs a novel variant of Markov Chain Monte Carlo (MCMC) optimized for models that allocate mass on subspaces of the full parameter space.

Scientific Applications:

• Differential Expression Discovery: Identification of genes differentially expressed under varying experimental conditions in gene expression microarray studies.
• Multi-Condition/Time-Course Studies: Detection and characterization of expression patterns across three or more conditions or time points.
• Disease and Treatment Studies: Analysis of experiments investigating disease mechanisms or treatment effects, including benchmarking on datasets such as HIV gene expression data.

Methodology:

Implements a robust Bayesian hierarchical model with a t-distribution for outliers, an exchangeable prior for gene variances, and a novel MCMC algorithm for parameter estimation.