Curating a large-scale regulatory network by evaluating its consistency with expression datasets

Abstract. The analysis of large-scale regulatory models using data is-sued from genome-scale high-throughput experimental techniques is an actual challenge in the systems biology field. This kind of analysis faces three common problems: the size of the model, the uncertainty in the expression datasets, and the heterogeneity of the data. On that ac-count, we propose a method that analyses large-scale networks with small – but reliable – expression datasets. Our method relates regula-tory knowledge with heterogeneous expression datasets using a simple consistency rule. If a global consistency is found, we predict the changes in gene expression or protein activity of some components of the net-work. When the whole model is inconsistent, we highlight regions in the network where the regulatory knowledge is incomplete. Confronting our predictions with mRNA expression experiments allows us to determine the missing post-transcriptional interactions of our model. We tested this approach with the transcriptional network of E. coli.