Metastasis is the leading cause of cancer-related deaths. While cancer and metastasis have been studied for many years there is still much to learn, particularly with regards to how various biological pathways interact to instigate or prevent metastasis. The multitude of gene regulatory networks that govern cancer metastasis can be studied individually as “modules” that represent individual networks. While biological networks are not isolated in nature, using modules is a necessary first step to understand the mechanism of cancer metastasis. This thesis will discuss the core network of the epithelial-mesenchymal transition (EMT), the glucose metabolism pathway, the stemness network, and the Notch signaling pathway. The dysregulation of these networks leads to phenotypes with high metastatic potential. For instance, a partial EMT transition can allow cells to gain the ability to collectively migrate while metabolic reprogramming can increase the ability of cells to survive in differing microenvironments. By coupling the EMT and metabolism networks, we noticed the phenotypes with high metastatic potential are correlated. While coupling networks can bring insight into the crosstalk between the modules governing cancer, another key aspect of cancer is cellular heterogeneity. We utilized random circuit perturbation, a way to generate an ensemble of heterogeneous phenotypes, to understand the hierarchical decision making of the stemness network. Lastly, rather than incorporating cellular heterogeneity via changing kinetic parameters, we can introduce stochasticity to adjust the stability of cellular states. We used stochastic fluctuations to model extrinsic noise, such as signals in the tumor microenvironment, and determined they can promote ordered pattern formation in Notch-Delta mediated systems. Additionally, noise alters the stability of the cellular states and potentially destabilizes a phenotype associated with therapy resistance in the Notch-Delta-Jagged pathway. These projects show the importance of not stopping at a modular viewpoint of biology. Instead, models should incorporate crosstalk, cellular heterogeneity, and could even incorporate noise when the effect of the microenvironment is unknown.