The benefit of using prior knowledge. Reconstruction of 20 networks based on MCMC simulations using prior knowledge and gene expression data. The reconstruction is evaluated using receiver operating characteristic (ROC) curves and the area under the ROC curve (AUC) as a measure for the accuracy of the reconstruction. (a) The prior knowledge model used to generate prior gene-gene interaction probabilities for true (N+) and false edges (N-) respectively. The parameter δ controls the accuracy of the prior knowledge by determining the separation of the means of both truncated Gaussian probability distributions. (b) Average ROC curves of a network reconstruction based on prior gene-gene interaction probabilities alone. Each curve represents a different level of δ. The dashed curve indicates the performance of a random prediction. It corresponds with an accuracy (AUC) of 0.5. (c) Average ROC curves of a network reconstruction based on prior knowledge and gene expression data. The size of the expression data is 10 data points per gene. The posterior probabilities are calculated by MCMC simulations. Each curve represents a different level of accuracy of the prior knowledge and corresponds with the prior knowledge used in panel (b). (d) Average accuracy of the prior networks versus average accuracy of the posterior networks for different data sizes. The dashed-dotted line indicates the equivalence between prior and posterior accuracy.