Background: A single neuron cannot think, just as a single hepatocyte cannot filter blood. The functions of the tissues that make up our bodies are emergent, depending on both their constituent cells’ internal biology, as well as how they interact with the other cells around them. Tumors are no different. They interact with nearby cells through both mechanical and molecular mechanisms, adapting their own behaviors to suit local environmental conditions, and even hijacking the functions of nearby tissues for tumor-promoting purposes. As such, predicting how a given tumor will grow or respond to treatment requires more than just an understanding of the tumor’s intrinsic characteristics, it requires a spatially- and temporally-resolved description of the tumor’s shape, its surrounding tissues, and the milieu of diffusible molecules that drive its behavior and interactions.
Methods: SimBioSys TumorScope for Breast Cancer implements a multi-scale simulation methodology that couples genome-scale metabolic modeling with reaction-diffusion and material mechanics models in order to predict how individual patients’ breast tumors will respond to different chemotherapeutic regimens. The simulations explicitly track the 3D morphology of the tumor and surrounding adipose, fibroglandular, and vascular tissues (initialized using pre-treatment MRI images), as well as the concentrations of key nutrients, metabolic byproducts, and drugs as they change over time. The vasculature serves as both source and sink for the modeled chemicals; nutrients and drugs are able to diffuse out of the blood vessels, while metabolic byproducts can diffuse in. The vascular concentrations of the drugs are assumed to vary in accordance with their administration schedules and known pharmacokinetics. At each location within the simulation, the local metabolic response of the tissues to the available nutrients is modeled using flux balance analysis, yielding uptake and efflux rates for each chemical species, as well as local growth rates for the tissues. At the same time, drugs may also be taken up by the cells, leading to cell death at rates that depend on the local intracellular drug concentrations, as well as their pharmacodynamic properties. As different regions of the tissues grow or shrink in response to their microenvironmental conditions, the macroscopic tissues stretch and deform.
Results: SimBioSys TumorScope for Breast Cancer was retrospectively applied to over 100 patients that received neoadjuvant chemotherapy (NACT) culled from publicly-available datasets (via the TCIA). Simulations were initialized with pre-treatment MRI data, and run through the entirety of each patient’s specified treatment regimen. Predicted changes in tumor volume and longest dimension were then compared against measured values at several time-points after initiation of therapy, yielding Pearson correlations of over 0.93 for both.
Conclusions: Through accurate spatio-temporal modeling of drug and nutrient perfusion, metabolic behavior, and the physico-chemical interactions that arise between tissues, the SimBioSys TumorScope for Breast Cancer can accurately predict the response of patients treated with NACT.
