Journées Statistiques Du Sud 2008

A current challenge in system biology is to infer the regulation network of a family of pgenes from a n-sample of microarrays, with n (much) smaller than p. Gaussian graphicalmodels are simple models to describe these regulation networks. We propose a procedurethat performs Gaussian graph estimation by model selection. We introduce a collection ofcandidate graphs and then select one of them by minimizing a penalized empirical risk. Wepay a special attention to the maximum degree of the graphs that we can handle and asses theperformance of the procedure in a non-asymptotic setting. The good theoretical propertiesof the procedure are confirmed on numerical examples.