shinyBN

shinyBN provides interactive construction, inference, and visualization of Bayesian networks to analyze relationships in high-dimensional biological datasets produced by high-throughput omics technologies.

Key Features:

• Probabilistic graphical modeling: Implements Bayesian network (BN) framework and accounts for the conditional independence assumptions applicable to sparse omics datasets.
• Structure learning and parameter training: Performs BN structure learning and parameter estimation using the bnlearn R package.
• Network inference: Conducts probabilistic inference on trained networks using the gRain R package.
• Visualization and rendering: Renders and exports network visualizations using visNetwork with customizable rendering settings.
• Predictive performance analysis: Supports receiver operating characteristic (ROC) curve analysis via pROC and decision curve analysis (DCA) via rmda.
• Flexible data input and outputs: Accepts multiple input data formats and exports network plots, prediction results, and external validation outputs.

Scientific Applications:

• Disease risk assessment: Applies BN modeling to estimate and interpret risk relationships from omics-derived predictors.
• Prognostic prediction: Uses learned networks and inference to generate prognostic predictions from high-dimensional biological data.
• Mechanistic insight from omics data: Explores conditional dependencies in high-throughput omics datasets to inform hypotheses about disease mechanisms.

Methodology:

Performs BN structure learning and parameter training (bnlearn), probabilistic inference (gRain), network visualization (visNetwork), ROC analysis (pROC), and decision curve analysis (rmda).