C Oates and S Mukherjee
Joint estimation of multiple networks from time course data
Date: 8 February, 2013
Abstract: Graphical models are widely used to make inferences concerning interplay in multivariate systems, as modeled by a conditional independence graph or network. In many applications, data are collected from multiple individuals whose networks may differ but are likely to share many features. Here we present a hierarchical Bayesian formulation for joint estimation of such networks. The formulation is general and can be applied to a number of specific graphical models. Motivated by applications in biology, we focus on time-course data with interventions and introduce a computationally efficient, deterministic algorithm for exact inference in this setting. Application of the proposed method to simulated data demonstrates that joint estimation can improve ability to infer individual networks as well as differences between them. Finally, we describe approximations which are still more computationally efficient than the exact algorithm and demonstrate good empirical performance.