Geostatistical visualization of ecological interactions in tumors

Recent advances in our understanding sof cancer progression have highlighted the important role played by proteomic heterogeneity and the tumor microenvironment. Single-cell measurement technologies have enabled deep investigation of tumor heterogeneity. However, tools to visualize and interpret single-cell data have lagged behind experimental methods; currently, dimensionality reduction and clustering techniques, such as t -sne and SPADE, are the most prevalent visualization techniques. However, such techniques do not enable the visualization of the super-cellular structures that arise either via microenvironmental forces (e.g. hypoxia) or through cell-cell interactions. They additionally do not allow the visualization of ecological niches that are critical to understanding tumor behavior. The advent of novel experimental hyperplex immunostaining platforms, capable of measuring the in situ protein expression of dozens of proteins simultaneously, necessitates novel computational tools to quantify and visualize spatial patterns in the tumor microenvironment. Here, we introduce an approach to visualize tumor heterogeneity that integrates multiple geostatistics to capture both global and local spatial patterns. We assess the utility of this approach using an agent-based model of tumor growth under four ecological contexts; predation, mutualism, commensalism, and parasitism. Spatial patterns are visualized in real time as the models progress. This application introduces both an ecological framework for characterizing cellular interactions in cancer and novel way of quantifying and visualizing spatial patterns in cancer.