Assessing the Validity Domains of Graphical Gaussian Models in Order to Infer Relationships among Components of Complex Biological Systems

摘要: The study of the interactions cellular components is an essential base step to understand structure and dynamics biological networks. Various methods were recently developed for this purpose. While most them combine different types data a priori knowledge, based on graphical Gaussian models are capable learning network directly from raw data. They consider full-order partial correlations which between two variables given remaining ones, modeling direct links variables. Statistical estimating these when number observations larger than However, rapid advance new technologies that allow simultaneous measure genome expression, led large-scale datasets where far observations. To get around dimensionality problem, strategies statistical proposed. In we focused published. All fact relationships very small in regards possible relationships, p(p-1)/2. context, assumption not always satisfied over whole graph. It precisely know behavior characteristics studied object before applying them. For purpose, evaluated validity domain each method wide-ranging simulated datasets. We then illustrated our results using published