Dynamic Bayesian network optimized by particle filtering in gene regulatory networks

With the development of bioinformatics, gene regulatory network research has gained growing more attention. The process of transcription regulation research has played a crucial role in the biomedical research. In a recent research work, the dynamic Bayesian network has become a powerful gene regulatory network modeling tool, which can show the power of the description of the relationship between complex gene regulations. However, because the vast majority of existing works simultaneously use all of the observational data on the reconstruction of the network structure and parameters optimization, so micro-array expression data in the timing characteristics have not been fully tapped. To solve the above problem, in this paper, the particle filter method is introduced into the framework of dynamic Bayesian networks algorithm, learning gene regulatory networks from the micro array expression data in sequential. By the test on brewer´s yeast cell cycle microarray expression data, the algorithm is proven to be successful in capturing the dynamic characteristics of expression data. Experimental results show that compared some other works, the algorithm has higher accuracy, and can be more accurately expressed gene regulatory network structure.