mgsa

mgsa implements a Bayesian model to identify biological categories (e.g., Gene Ontology (GO) terms) that are enriched for responder genes in genomics experiments.

Key Features:

• Bayesian Network Integration: Embeds all biological categories within a Bayesian network that models gene response as a function of category activation.
• Probabilistic Inference: Uses probabilistic inference to identify active categories while accounting for overlap between categories without requiring multiple testing corrections.
• Performance on Simulations: Demonstrates up to 95% precision at 20% recall under moderate noise and reports a tenfold precision improvement over single-category enrichment analyses on simulated data.
• High-Level Biological Summaries: Produces summarized core biological processes and reduces confounding associations when applied to real datasets such as gene expression data in yeast.

Scientific Applications:

• Gene set enrichment interpretation: Identification of biologically relevant pathways and GO terms enriched for responder genes in genomics experiments.
• Transcriptome summarization: Extraction of high-level biological process summaries from gene expression datasets, including yeast expression studies.

Methodology:

Implements the Model-Based Gene Set Analysis (MGSA) Bayesian modeling approach by embedding categories in a Bayesian network, modeling gene responses as functions of category activation, and applying probabilistic inference to infer active categories while accounting for category overlap.