No abstract

No abstract

The high complexity in the gene regulation mechanism and the prevalent noise in high-throughput detection experiments are considered to be the two major obstacles in discovering transcriptional regulation with high accuracy from experimental gene expression data. In this paper, we study a model based on dynamic Bayesian networks to predict gene regulation by integrating transcription factor binding site data and proteinprotein interaction data with gene expression data. The knowledge of genetic interactions between proteins and the presence of transcription factors binding site at the promoter region of a gene have been used to restrict the number of potential regulators of each gene. We show the effectiveness of combining multiple data sources in the prediction of transcriptional regulation through the analysis of Saccharomyces cerevisiae (Yeast) cell cycle data. Experiments conducted on real microarray datasets show that the proposed model is significantly more efficient and topologically more accurate compared to other existing models based on dynamic Bayesian networks. We also demonstrate the scalability of the proposed model through the analysis of a large dataset with a sustainable performance level.

The experimental microarray data has the potential application in determining the underlying mechanisms of transcription regulation in a living cell. The inference of this regulation circuitry with computational methods suffers from two major challenges: the low accuracy of inferring true positive connections and the excessive computation time. In this paper, we show that models based on Dynamic Bayesian Networks which exploit the biological features of gene expression are more computationally efficient and topologically accurate compared to the other existing models. Using two experimental microarray datasets of the yeast cell cycle, we also evaluate how successfully the available models can address the current challenges with the increasing size of the datasets.