Recent observation techniques on the nucleus of living cells have revealed dynamic and self-organized aspects of multiprotein structures at very different spatial and temporal scales [Misteli 2007, Cell, 128:787-800]. Together with the evidences for the stochasticity of gene expression [Kaern et al. 2005, Nat.Rev.Genet., 6:451-64], these properties contradict our understanding of cellular processes. The difficulty for experimentalists to observe at the relevant spatio-temporal scales brings modeling and simulation approaches to the an imporant place. We focus here on modeling this dynamic structuration at different scales (from multiprotein complexes to nuclear organization) and to analyze the resulting stochastic properties, by mean of various computational methods: from individual-based (or multi-agents) molecular simulation to the analysis of stochastic biophysical processes.