EcoSta 2017: Start Registration
View Submission - EcoSta2017
A0740
Title: Penalized estimation of sparse concentration matrices based on prior knowledge with applications to placenta metal data Authors:  Jiang Gui - Dartmouth College (United States) [presenting]
Abstract: Essential elements (P, S, K, Ca, Mg, Mn, Co, Fe, Cu, Zn, Se,) play critical cellular roles as structural components of bio-molecules, signalling molecules, catalytic co-factors and regulators of protein expression, and in biological systems their concentrations are homeostatically regulated. Altered metal homeostasis occurs in neurological diseases and cancer. The chemical architecture for metals may be complex, and advanced biostatistical methods are needed to infer the dependency structures of the metals. We introduce a weighted sparse Gaussian graphical model that can incorporate prior knowledge to infer the structure of the network of metal concentrations measured in the human term placenta. We present the L1 penalized regularization procedure for estimating the sparse precision matrix in the setting of Gaussian graphical models. Simulation results indicate that the proposed method yields a better estimate of the precision matrix than the procedures that fail to account for the prior knowledge of the network structure. We also applied this method to a New Hampshire Birth Cohort Study for inference of the chemical network in placenta biopsies.