Wessel van Wieringen and Yao Chen (now at Ghent University) published an article on the reproducibility of reconstructed networks in Statistics in Medicine. The article starts from the observation that the reconstruction of the cohesion among the variates of a multivariate random variable by means of Gaussian graphical model usually takes only sampling variation into account. They then point out the consequences of this practice for the reconstruction of the underlying conditional independence graph. When replicates are included in the study, these consequences are overcome by the separation of sampling from other sources of variation. Hereto a simple `signal+noise' model for the description of the multivariate data has been put forward. A penalized EM algorithm for the estimation of the model's parameters has been presented, alongside a discussion of cross-validation for choosing the penalty parameter(s). Through simulation they investigate how much is won by the inclusion of replicates, and compare the presented method to obvious alternatives. Finally, in an illustration using oncogenomics studies with replicates they further investigate the effect of ignoring variation due to other sources than sampling variation and assess the reproducibility of the reconstruction of the conditional independence graph.