Abstract: Modeling cell state transitions using regulatory networks
Cells undergo radical shifts in identity – such as stem cells differentiating into specific tissues, or healthy cells becoming diseased – by turning on and off specific gene expression programs. These programs are in turn controlled by DNA-binding proteins called transcription factors that interact together to form a gene regulatory network. We developed a series of computational methods based on modularity maximization, Bayesian matrix factorization, and probabilistic Boolean networks to analyze the structure and dynamics of these gene regulatory networks. Applying these methods to the aggressive skin cancer Merkel cell carcinoma, we identified genetic and drug interventions that can shift cells towards a less pathological state. We validated our predictions in tumor-derived cell lines and mouse models in the wet-lab. Our work provides a blueprint for how the mathematical modeling of networks can have biological and clinical applications.