A minimum-labeling approach for reconstructing protein networks across multiple conditions

Background: The sheer amounts of biological data that are generated in recent years have driven the development of network analysis tools to facilitate the interpretation and representation of these data. A fundamental challenge in this domain is the reconstruction of a protein-protein subnetwork that underlies a process of interest from a genome-wide screen of associated genes. Despite intense work in this area, current algorithmic approaches are largely limited to analyzing a single screen and are, thus, unable to account for information on condition-specific genes, or reveal the dynamics (over time or condition) of the process in question.Results: We propose a novel formulation for the problem of network reconstruction from multiple-condition data and devise an efficient integer program solution for it. We apply our algorithm to analyze the response to influenza infection and ER export regulation in humans. By comparing to an extant, single-condition tool we demonstrate the power of our new approach in integrating data from multiple conditions in a compact and coherent manner, capturing the dynamics of the underlying processes.