Computational complexity of agent-based multi-scale cancer modeling

The computational cost of agent-based multi-scale tumor growth simulations usually grows explosively with increasing the system size. Current models have to comprise a lot between model complexity and model fidelity. In this paper, this problem is investigated using a tumor model developed by us. Our model takes into account the discrete and the continuous nature involved in tumor development by integrating the discrete and the continuum approaches. Each individual tumor cell is modeled as a single agent that interacts with its surroundings biochemically and biomechanically. Cells move in an over-damped manner, producing a system of linear equations. The transport and metabolism of nutrients are described by reaction-diffusion equations. Cell proliferation is performed according to the biochemical and biomechanical conditions of the environment. Several methods for numerically solving the linear system of cell motion are compared in terms of performance and stability. After that, our model is analyzed referring to its computational bottlenecks. In general, this work can provide important instructing information for people who intend to work on multi-scale cancer modeling using the agent-based method.