A dynamic bayesian network-based model for inferring gene regulatory networks from gene expression data

Driven by the need to uncover the vast information and understand the dynamic behaviour of biological systems, researchers are now garnering interests in inferring gene regulatory networks (GRNs) from gene expression data which is otherwise unfeasible in the past due to technology constraint. In this regard, the dynamic Bayesian network (DBN) has been broadly utilized for the inference of GRNs, thanks to its ability to handle time-series microarray data and model feedback loops. Unfortunately, the commonly found missing values in gene expression data, and the excessive computation time owing to the large search space whereby all genes are treated as potential regulators for a target gene, often impede the effectiveness of DBN in inferring GRNs. This paper proposes a DBN-based model with missing values imputation and potential regulators selection (ISDBN) which deals with the missing values and reduces the search space by selecting potential regulators based on gene expression changes. The performance of the proposed model is assessed by using S. cerevisiae cell cycle and E. coli SOS response pathway time-series expression data. The experimental results showed reduced computation time and improved accuracy in detecting gene-gene relationships when compared to conventional DBN. The results of this study showed that ISDBN performs better than conventional DBN in terms of accuracy and computation time for GRNs inference. Moreover, we foresee the applicability of the resultant networks from ISDBN as a framework for future gene intervention experiments.