bgx

bgx performs Bayesian analysis of expression data from Affymetrix 3' GeneChips (single-labeled extracts with match and mismatch probes), modeling probe-level effects and producing posterior distributions to quantify gene expression and uncertainty.

Key Features:

• Integrated Bayesian Framework: Integrates background correction, normalization, and summarization within a unified Bayesian model that uses all probe-level data.
• Handling Uncertainty: Produces posterior distributions for each gene's expression level rather than point estimates, capturing probe-response uncertainty.
• Error Modeling: Accounts for additive and multiplicative errors, non-specific hybridization, replicate summarization, and probe affinity effects using sequence information.
• Adaptive MCMC Algorithm: Uses an adaptive Markov chain Monte Carlo algorithm to efficiently sample posterior distributions and estimate expression levels and fold changes between conditions.
• Comprehensive Analysis Functions: Includes functions for ranking genes by expression and estimating the number of upregulated and downregulated genes under different experimental conditions.

Scientific Applications:

• Comparative Gene Expression Studies: Enables comparison of gene expression between biological states with uncertainty-aware differential expression estimates.
• Transcriptomic Profiling: Provides nuanced transcriptome-level expression estimates and variability across experimental setups.
• Functional Genomics and Systems Biology: Supports identification of regulatory genes and pathways using uncertainty-informed differential expression outputs.

Methodology:

Implements a unified Bayesian model of probe-level data with posterior inference via an adaptive MCMC sampler, modeling additive and multiplicative errors, non-specific hybridization, probe affinity effects using sequence information, and replicate summarization to estimate expression levels, posterior distributions, and fold changes.