Discipline: Biological Sciences
Subcategory: Cancer Research
Session: 1
Room: Harding
Chris Andersen - University of Puget Sound
Cancer is a pervasive disease that has largely been studied with a focus on oncogene function in individual cells despite its well documented heterogeneity. As such, our understanding of many solid tumor malignancies is limited, as several recent studies have described the significance of understanding intercellular signaling and, more broadly, spatial relationships for the tumor microenvironment. Developing novel approaches to understanding the spatial relationships between cancer cells and the tumor microenvironment has clinical applications, such as patient outcomes and resistance to drug therapy and may clarify some discrepancies in patient diagnoses. The goal of this research was to apply a novel computational pipeline to a 356-member cohort of colorectal cancer patients, consisting of spatially resolved in situ images using hyperplexed staining of more than 50 known cancer biomarkers, and assess the effect of heterogeneity in the tumor microenvironment on patient outcomes. If cancer cells utilize their spatial proximity to neighboring cells via intercellular signaling, then the spatial correlations between cells in the tumor microenvironment will be reflected in the patient outcomes. To understand what spatial relationships may arise within our cohort, we began by generating simulated datasets with known spatial relationships and applied a spatial statistical model, known as the multitype Strauss model, to these datasets. Our results were corroborated by ground truth values in the literature, indicating that our model can be applied to our experimental datasets. Current work at the University of Puget Sound is aimed at identifying the most heterogeneous regions of our images (i.e. where cancer is most pervasive) and applying our spatial model to the biomarker-based phenotypes of each cell across our images. Additionally, an alternative model assessing clustered spatial interactions (i.e. when cells are found closer to one another on average) using a Markov point process has demonstrated improved theoretical performance in assessing spatial clustering patterns, though its application in imaging contexts is limited. This spatial statistics approach is not limited to this cohort or form of cancer. The approach can be generalized to any solid tumor malignancy or biological structure resembling the TME. Future projects may implement a similar workflow as an exploratory tool for the discovery and quantification of spatial interactions in in situ models. References: Weinstein IB. Addiction to Oncogenes–the Achilles Heal of Cancer. Science. 2002;297: 63-64. Baddeley AJ, van Lieshout MNM. Area-interaction point processes. Ann Inst Stat Math. 1995;47: 601-619. Gerdes MJ, Sevinsky CJ, Sood A, Adak S, Bello MO, Bordwell A, et al. Highly multiplexed single-cell analysis of formalin-fixed, paraffin-embedded cancer tissue. Proceedings of the National Academy of Sciences. 2013;110: 11982-11987.
Funder Acknowledgement(s): Funding for the TECBio REU program at Pitt was provided by the National Science Foundation under Grant DBI-1659611.
Faculty Advisor: Chakra Chennubhotla, chakracs@pitt.edu
Role: I developed and curated the simulated datasets in Matlab and I applied the multitype Strauss model using RStudio to the simulated data. I am currently working on applying the aforementioned model to the real datasets (which I did not generate).