GR2D2

GR2D2 estimates precision matrices for Gaussian Graphical Models to reconstruct biological networks from high-dimensional, sparse data.

Key Features:

• Graphical R2-induced Dirichlet Decomposition: Implements Graphical R^2-induced Dirichlet Decomposition derived from R2D2 priors for linear models.
• R2D2 priors: Leverages R2D2 priors for linear models to inform precision-matrix estimation.
• Estimation algorithm: Computes posterior estimates via a data-augmented block Gibbs sampler.
• Handling sparsity and high dimensionality: Tailored to reconstruct GGMs from sparse, high-dimensional precision matrices.
• Estimation accuracy: Demonstrates superior precision-matrix estimation with minimal information divergence relative to existing techniques across simulation settings.
• Biological network focus: Targets reconstruction of regulatory interactions among DNA, RNA, and proteins within GGMs.

Scientific Applications:

• Biological network reconstruction: Reconstructs GGMs to capture regulatory interactions and relationships among DNA, RNA, and proteins from high-dimensional data.
• Cancer transcriptomics: Applied to five cancer RNA-seq gene expression datasets spanning three cancer types to identify common and cancer-specific pathways.

Methodology:

Implements R2D2 priors for linear models within a Graphical R^2-induced Dirichlet Decomposition framework and computes posterior estimates using a data-augmented block Gibbs sampler.