Gene regulatory network is essential to regulate the biological functions of cells. With the rapid development of “omics” technologies, the network can be inferred for a certain biological function. However, it still remains a challenge to understand the complex network at a systematic level. In this thesis, we utilized quantitative modeling approaches to study the nonlinear dynamics and the design principles of these biological gene regulatory networks. We aim to explain the existing experimental observations with the model, and further propose reasonable hypothesis for future experimental designs. More importantly, the understanding of the circuit’s regulatory mechanism would benefit the design of a de novo gene circuit for a new biological function. We first studied the plasticity of cell migration phenotypes during cancer metastasis, which contains two key cellular plasticity mechanisms - epithelial-tomesenchymal transition (EMT) and mesenchymal-to-amoeboid transition (MAT). In this study, we quantitatively modeled the core Rac1/RhoA gene regulatory circuit for MAT and later connected it with the core regulatory circuit for EMT. We found four different stable states, consistent with the amoeboid (A), mesenchymal (M), the hybrid amoeboid/mesenchymal (A/M), and the hybrid epithelial/mesenchymal (E/M) phenotypes that are observed in the experiment. We also explored the effects of microRNAs and EMT-inducing signals like Hepatocyte Growth Factor (HGF), and provided a new insight for the transitions among these phenotypes. To improve the traditional modeling approaches, we developed a new computational modeling method called Random Circuit Perturbation (RACIPE) to explore the dynamic behavior of gene regulatory circuits without the requirement of detailed kinetic parameters. We applied RACIPE on several gene circuits, and found the existence of robust gene expression patterns even though the model parameters are wildly perturbed. We also showed the powerful aspect of RACIPE to decipher the operating principles of the circuits. This kind of quantitative models not only works for gene regulatory network, but also is capable to be extended to study the cell-cell interactions among cancer and immune cells. The results shown the co-occurrence of three cancer states: low risk cancer with intermediate immunity (L), intermediate risk cancer with high immunity (I) and high risk cancer with low immunity state (H). We further used the model to assess the different combinations of cancer therapies.