Cancer accounted for 8.2 million deaths in 2012, a number that continues to grow despite the efforts of researchers and medical professionals alike (World Health Organization, 2014). The challenge in understanding cancer, and thereby developing successful therapies, stems from its inherent heterogeneity and the complexity of its microenvironment. It is therefore essential to study, and account for, the manner in which tumors interact with the environment. To address this need, we developed a tunable, multi-scale in silico model that predicts the emergent behavior of a tumor in response to a variety of environmental and therapeutic perturbations.
Continuous and deterministic models cannot account for the discrete, stochastic, and heterogeneous nature of tumor cells, prompting researchers to consider a different paradigm: bottom-up rather than top-down modeling. Bottom-up approaches characterize the low-level building blocks of a system and couple them together to predict dynamic responses. Here, these building blocks, or agents, are individual healthy or cancerous cells. These cells are coupled in a 3D microenvironment that accounts for crowding, nutrient resources, and cytokine signaling (both secretion and uptake). Unlike cellular automata (CA) models, the agent-based modeling (ABM) framework allows both spatial and temporal resolution so these cells can divide, invade, and apoptose across space and time.
Our multi-scale ABM incorporates inter- and intra-cellular interactions, nutrient diffusion, and paracrine signaling. Cell agents, both healthy and cancerous, are equipped with a simplified, 14-component EGFR network that interacts with the environment (Zhang et al., 2009). Each cell agent can exist in one of five phenotypes—quiescent, proliferative, invading, apoptotic, or necrotic—and follows rules defined by its phenotype, local environment, and neighboring cells. More specifically, the intracellular network is most responsible for proliferation and invasion, and—through the corresponding dynamics—couples the cell to the physical environment. Glucose availability and crowding is most responsible for a cell’s entrance to, and exit from, quiescence. Cancerous agents have a modified rule set based on the “hallmarks of cancer” described by Hanahan & Weinberg (2000). Examples of distinctions include increased tolerance for crowding and an accelerated cell cycle.
The flexibility of agent-based modeling is demonstrated by the ease with which new agents and perturbations can be introduced. A sensitivity analysis of the model demonstrated significant sensitivity of parameters affecting glucose conditions, highlighting glucose concentration as a potential control input. Variable glucose conditions, in combination with existing treatment strategies—excision, radiation, and chemotherapy—will be simulated to predict therapeutic efficacy. These tunable perturbations can be introduced at any time point and in any combination. Preliminary results quantify the effects of increasing excision radius under different glucose conditions, revealing important tradeoffs that should be explored in future experiments as well as strategies to be validated in the clinic. The development and analysis of tumor growth under different parameters and therapies demonstrate the versatility of ABMs for guiding translational cancer therapy and informing more personal medicine.
Hanahan, D., & Weinberg, R. A. (2000). The hallmarks of cancer. Cell, 100(1), 57-70.
World Health Organization. Global status report on noncommunicable diseases. 2014.
Zhang, L., Want, Z., Sagotsky, J. A., & Deisboeck, T. S. (2009). Multiscale agent-based cancer modeling. J Math Biol, 58, 545-559.
See more of this Group/Topical: Topical Conference: Emerging Frontiers in Systems and Synthetic Biology