An integrated gene regulatory network inference pipeline

Rapidly accumulated gene expression data put forward the development of numerous methods for inferring gene regulatory networks and the efforts for critical performance assessment of these methods. In this paper, we propose an integrated pipeline for gene regulatory network inference motivated by the results and follow-up analysis of a blinded, community-wide challenge DREAM (Dialogue on Reverse Engineering Assessment and Methods) project. In particular, we categorize the gene expression data into three types, i.e., steady-state gene expression profile of knockout or knockdown experiments, steady-state gene expression profiles after multi-factorial perturbations, and time-series data after multi-factorial perturbations. Then we analyze the three types of gene expression data by using the combination of fold change and t-test, the path consistency algorithm based on conditional mutual information, and the ordinary differential equation model, respectively. Finally we integrate the three procedures to a pipeline for gene regulatory network inference by considering their complementarities. Performance for the network inference will be improved in the proposed pipeline by maximally utilizing information in the available data, emphasizing the knock-out and knock-down data, and differentiating the direct and indirect regulatory interactions.