A0622
Title: The graphical R2D2 estimator for the precision matrices
Authors: Dailin Gan - University of Notre Dame (United States)
Guosheng Yin - The University of Hong Kong (Hong Kong)
Yan Zhang - The University of Hong Kong (Hong Kong) [presenting]
Abstract: Biological networks are important for the analysis of human diseases, which summarize the regulatory interactions and other relationships between different molecules. Understanding and constructing networks for molecules, such as DNA, RNA and proteins, can help elucidate the mechanisms of complex biological systems. The Gaussian Graphical Models (GGMs) are popular tools for the estimation of gene regulatory networks because of their biological interpretability. Nonetheless, reconstructing GGMs from high-dimensional datasets is still challenging and current methods cannot handle the sparsity and high-dimensionality issues arising from datasets very well. Here we developed a new GGM, called the graphical R2D2 (R2-induced Dirichlet Decomposition), based on the R2D2 priors for linear models. When the true precision matrix is sparse and of high dimension, the graphical R2D2 provides the estimates with the smallest information divergence from the sampling model. Besides, we also provide a full Gibbs sampler for implementing the graphical R2D2 estimator. We also provide breast cancer gene network analysis using the graphical R2D2 estimator and the important genes recognized from the inferred gene regulatory networks are consistent with biological ground truth.
