Survey of Research on Non-homogeneous Gene Regulatory Network Models

Abstract

In the field of bioinformatics, the construction of gene regulatory networks is crucial. In recent years, non-homogeneous dynamic Bayesian networks have become a common modeling tool for learning gene regulatory networks from gene expression time-series data. Aiming at this research field, this paper combs the gene regulatory network modeling methods based on homogeneous dynamic Bayesian network, and summarizes the non-homogeneous dynamic Bayesian models proposed in the past ten years from the time slice division method and gene regulation network parameter learning. The main contents include: time slice division method, including free allocation, continuous changepoint process, discrete changepoint process and hidden Markov-based changepoint process; gene regulation network parameter learning, mainly including sequence coupling parameters and global coupling parameters. After that, the performance of the non-homogeneous dynamic Bayesian network models is analyzed, and the accuracy and reliability of these models for gene regulatory network modeling, as well as the differences and connections between the models are introduced. Finally, the difficulties and challenges of gene regulatory network construction and some future research directions of non-homogeneous dynamic Bayesian net-work models are pointed out.

Keywords

• dynamic bayesian networks; non-homogeneous dynamic bayesian networks; gene regulatory network; bioinformatics