Background: The ultra intercellular heterogeneity in tumor is one major causes for the failure of cancer therapy, e.g.drug resistance and/or cancer relapse.How the two key properties, the robustness and gene expression noise, of gene regulatory network contribnte to the genesis and progress of heterogeneity largely remain to be characterized.Methods: Based on a well-studied framework of gene regulatory network, the progress of tumor heterogeneity was modeled by in silicon evolving of a cell population with initial isogenetic background.The roles of robustness and Gaussian noise in the gene regulatory network were examined within a wide range parameters.Results: We found that high robustness of a gene regulatory network promotes the high genotypic heterogeneity in the asexual cell population during the simulation.However, the genotypic heterogeneity only weakly associated with phenotypic heterogeneity, which highlighted the critical roles of non-genetic sources for the genesis and progress of phenotypic heterogeneity.We investigated one such non-genetic source, gene expression noise, and revealed that it is strongly negative connects to phenotypic heterogeneity with weak effect to genotypic heterogeneity.Conclusions: The two close relationships, between robustness and genotypic heterogeneity, and between gene expression noise and phenotypic heterogeneity, revealed in the present work suggested the divided mechanisms for the progress of two type of heterogeneities.Our model provides a theoretical framework to characterized more realistic situations in diagnose or cancer therapy .