A Self-organizing State Space Approach to Inferring Time-Varying Causalities between Regulatory Proteins

摘要

A number of methods based on time-dependent state space models have been proposed for inferring time-varying gene regulatory networks. These methods are capable of detecting a relatively small number of topological changes in gene regulatory networks. However, they are insufficient since there is a greater number of changes in the gene regulatory mechanisms; the function of a regulatory protein frequently changes due to post-translational modification, such as protein phosphorylation and ATP-binding. We propose a self-organizing state space approach to inferring consecutive changes in causalities between regulatory proteins from gene expression data. Hidden regulatory proteins are identified using a test-based method from genome-wide protein-DNA binding data. Application of this approach to cell cycle data demonstrated its effectiveness.