CGBayesNets

CGBayesNets learns and performs inference on Conditional Gaussian Bayesian Networks to model relationships and predict outcomes from datasets containing both discrete and continuous variables.

Key Features:

• Mixed Data Type Modeling: Supports Bayesian network analysis of datasets containing both discrete and continuous variables without requiring discretization.
• Conditional Gaussian Bayesian Network Framework: Implements algorithms specifically designed for Conditional Gaussian Bayesian Networks to model dependencies in mixed genomic data.
• Multiple Network Learning Algorithms: Provides four network structure learning algorithms that balance computational cost with network likelihood.
• Probabilistic Inference Functions: Includes inference routines for predicting phenotypes and other outcomes using Bayesian network models.
• Model Validation Methods: Supports cross-validation, bootstrapping, and Area Under the Curve (AUC) evaluation for assessing predictive performance.

Scientific Applications:

• Genomic Predictive Modeling: Predicts clinical phenotypes and biological traits using mixed genomic datasets containing discrete and continuous variables.
• Multi-Omics Data Analysis: Analyzes integrated datasets from genomics, metabolomics, and gene expression studies to identify probabilistic relationships among variables.

Methodology:

CGBayesNets applies Conditional Gaussian Bayesian Network learning algorithms to mixed discrete and continuous datasets and performs probabilistic inference with validation procedures including cross-validation, bootstrapping, and AUC evaluation.