Reconstructing the metabolic landscape from histology images using a hybrid agent-based model

Abstract: The tortuous tumor vasculature and irregular cellular architecture can modify the metabolic landscape resulting in heterogeneities in tissue oxygenation. This can influence tumor response to anticancer therapies, i.e., chemotherapy, radiotherapy, and immunotherapy. To examine the distribution of intratumoral metabolites, we developed an in-silico hybrid agent-based model with Michaelis-Menten kinetics for oxygen uptake and a constant influx of oxygen from the vasculature. This model uses digitized tumor histology images from pancreatic and bladder cancers as the base for simulations. The images were digitized with either ImageJ or Aperio software to obtain the individual tumor, immune, and stromal cells, as well as blood vessels. The Clark-Evans test and Ripley’s K analyses were used to investigate the spatial patterns of clumped vs. dispersed cells and vessels. The simulations for tissue oxygenation were run for different pancreatic tumor grades (benign, premalignant, or invasive tumors) to determine the distributions of oxygenated vs. hypoxic tumor cells. Additionally, the distributions of infiltrating immune cells (CD8+, CD4+, and MDSC) in untreated vs. treated tumors (gemcitabine and/or adoptive T cell therapy) were examined for bladder tumors and normal tissues. These cell-scale simulations based on histology images are the first step in correlating the oxygen distribution patterns with 1) pancreatic tumor grades and 2) bladder tumor-T cell infiltration potential. In the future, this approach will be combined with acidity and glucose maps to identify tumor niches of specific phenotypes from metabolic landscapes.

Back to Search Research Symposium Abstracts