Modeling gene regulatory network is a difficult task. Experiments often fail at giving enough information for parameterizing existing models. We propose herein a dedicated hybrid modeling. To overcome the lack of quantitative information, our modeling focuses on (i) the biological compound product signs and (ii) the temporal properties associated with the biological effects of the interactions. Both these constraints are easy to extract from experimental data. They aims at reasoning on a hybrid system adapted to large gene regulatory networks, which is suitable for emphasizing biological properties.