A0249
Title: Detecting and leveraging node-level information in network inference
Authors: Xiaoyue Xi - Univerisity of Cambridge (United Kingdom) [presenting]
Helene Ruffieux - Univerisity of Cambridge (United Kingdom)
Abstract: Bayesian graphical models are powerful tools to infer complex relationships in high dimensions yet are often fraught with computational and statistical challenges. If exploited in a principled way, the increasing information collected alongside the data of primary interest constitutes an opportunity to mitigate these difficulties by guiding the detection of dependence structures. For instance, gene network inference may be informed using publicly available summary statistics on the regulation of genes by genetic variants. The aim is to present a novel Gaussian graphical modelling framework to identify and leverage information on the centrality of nodes in conditional independence graphs. Specifically, a fully joint hierarchical model is considered to simultaneously infer (1) sparse precision matrices and (2) the relevance of node-level information for uncovering the sought-after network structure. Such information is encoded as candidate auxiliary variables using a spike-and-slab sub-model on the propensity of nodes to be hubs, which allows hypothesis-free selection and interpretation of a sparse subset of relevant variables. A variational expectation conditional maximization algorithm is developed that scales inference to hundreds of samples, nodes and auxiliary variables. The advantages of the approach are illustrated and exploited in simulations and in a gene network study, which identifies hub genes involved in biological pathways relevant to immune-mediated diseases.
