Abstract
Motivation: Gene expression changes over time in response to perturbations. These changes are coordinated into functional modules via regulatory interactions. The genes within a functional module are expected to be differentially expressed in a manner coherent with their regulatory network. This perspective presents a promising approach to increase power to detect differential signals as well as for describing regulated modules from a mechanistic point of view. Results: We present an effective procedure for identifying differentially activated subnetworks in molecular interaction networks. Differential gene expression coherent with the regulatory nature of the network is identified. Sequentially controlling error on genes and links results in more efficient inference. By focusing on local inference, our method is ignorant of the global topology, and as a result equally effective on exponential and scale-free networks. We apply our procedure both to systematically simulated data, comparing its performance to alternative methods, and to the transcription regulatory network in the context of particle-induced pulmonary inflammation, recapitulating and proposing additional candidates to some previously obtained results. Availability: The R package is made available at git.oru.se/bioinfo/diffactivation. Contact: stephen.rush@oru.se, dirk.repsilber@oru.se Supplementary information: Supplementary data are available at -----------
