Bridging Scales: a Hybrid Model to Simulate Vascular Tumor Growth and Treatment Response

TL;DR

This paper presents a hybrid computational model combining agent-based and PDE approaches to simulate breast cancer tumor growth and treatment response, capturing complex biological interactions in 3D with high scalability.

Contribution

The work introduces a scalable hybrid model for vascular tumor growth and treatment response, integrating agent-based and PDE components, applicable to breast cancer and potentially other scenarios.

Findings

Model qualitatively matches pre-clinical data on therapy effects.

Successfully simulates large tumor volumes with over 92 million agents.

Demonstrates scalability and efficiency of the C++ implementation.

Abstract

Cancer is a disease driven by random DNA mutations and the interaction of many complex phenomena. To improve the understanding and ultimately find more effective treatments, researchers leverage computer simulations mimicking the tumor growth in silico. The challenge here is to account for the many phenomena influencing the disease progression and treatment protocols. This work introduces a computational model to simulate vascular tumor growth and the response to drug treatments in 3D. It consists of two agent-based models for the tumor cells and the vasculature. Moreover, partial differential equations govern the diffusive dynamics of the nutrients, the vascular endothelial growth factor, and two cancer drugs. The model focuses explicitly on breast cancer cells over-expressing HER2 receptors and a treatment combining standard chemotherapy (Doxorubicin) and monoclonal antibodies with anti-angiogenic properties (Trastuzumab). However, large parts of the model generalize to other scenarios. We show that the model qualitatively captures the effects of the combination therapy by comparing our simulation results with previously published pre-clinical data. Furthermore, we demonstrate the scalability of the model and the associated C++ code by simulating a vascular tumor occupying a volume of 400mm3 using a total of 92.5 million agents.