A1418
Title: Bayesian chain graph model for microbiome data
Authors: Claudia Solis-Lemus - (United States) [presenting]
Abstract: A novel Bayesian chain graph model is introduced to infer a sparse network structure with nodes for responses and for predictors with applications to microbiome research. Directed edges between a predictor and a response represent conditional links, and undirected edges among responses represent correlations. Specifically, the model can estimate a microbial network that represents the dependence structure of a multivariate response (e.g. abundances of microbes) while simultaneously estimating the effect of a set of predictors that influence the network (e.g. diet, weather, experimental treatments). In addition, the method produces a sparse interpretable graph via LASSO penalization, which can become adaptive so that different shrinkage can be applied to different edges. Furthermore, the model is able to equally handle small and big data and is computationally inexpensive through an efficient Gibbs sampling algorithm. With hierarchical structure, the model is extended to binary, counting and compositional responses by adding an appropriate sampling distribution to the core Normal model.
