AUC results for network reconstruction and hub identification on simulated data.

Data was simulated using the GeneNetWeaver package, subsampling networks of size 100, 250 and 500 from the yeast transcriptional network. Simulation was done using an ordinary differential equation model, with and without experimental noise added to the data. Network reconstruction was carried out using the model described, using an L1 and a hierarchical automatic relevance determination (ARD) prior, respectively. Shown are the area under the ROC curve values for the correct identification of edges (top) and hub identification (bottom). A value of 0.5 is equivalent to guessing, a value of 1 corresponds to perfect identification of hub genes.