Reverse Engineering of Genome-wide Gene Regulatory Networks from Gene Expression Data

Abstract

Transcriptional regulation plays vital roles in many fundamental biological processes. Reverse engineering of genome-wide regulatory networks from high-throughput transcriptomic data provides a promising way to characterize the global scenario of regulatory relationships between regulators and their targets. In this review, we summarize and categorize the main frameworks and methods currently available for inferring transcriptional regulatory networks from microarray gene expression profiling data. We overview each of strategies and introduce representative methods respectively. Their assumptions, advantages, shortcomings, and possible improvements and extensions are also clarified and commented.

Keywords: Computational model; Gene expression data; Genome-wide inference; Reverse engineering; Transcriptional regulatory network.

Fig. (1)

The general framework of reverse engineering transcriptional regulatory networks. (A) The framework of inferring regulatory network from gene expression profiles. There are various sample types of gene expression data, such as condition-specific, perturbation and time series data. A reverse engineering algorithm takes the input of the gene expression profiles and outputs the inferred gene regulatory relationships in form of a network. (B) The interrelated four levels of regulatory parameter information should be determined in the reverse engineering. The algorithm addresses the gene regulatory questions at one or several combined levels. (C) The regulatory pair and system in the modeling. The decision-making of regulatory relationship of an individual pair is in an isolated manner. However, the regulatory system consists of complicated regula-tions of combination and cooperation, such as the indirect regulation from gene G1 to gene G2 conditioned upon gene G3, which needs to be modeled in a systematic manner.

Fig. (2)

The framework of building gene coexpression regulatory network [67]. (A) The array data. (B) The correlation analysis of these genes. (C) Pairwise gene correlation matrix. The bold numbers are those over a defined threshold 0.80. (D) The built gene coexpression network.

Fig. (3)

The reverse engineering diagram of PCA-CMI (path consistency algorithm based on conditional mutual information) [60]

Fig. (4)

An example of Boolean network. (A) A Boolean network G(V,F). (B) The corresponding wiring graph of G(V,F) (C) The logic operations and state transition table. The possible input at time point and the corresponding output at time t+1 are listed in the table. Boolean network models the regulatory relationships in the logical operating scheme [106].

Fig. (5)

The graphical representation of Bayesian network and dynamic Bayesian network. (A) An example of a Bayesian network. By recursive de-composition, the joint probability distribution of the network is . The condi-tional independence simplifies the conditional probability distributions of these nodes in the decomposition. (B) The graphical representation of a dynamic Bayesian network (DBN). The static and dynamic representations are shown respectively. Assuming the temporal regulations are from time t+1 to , cyclic structures are apparently permitted in the DBN framework.