A Computational Framework for Deriving Mechanistic Hypotheses about Immune Interactions in the Tumor Microenvironment from Multiplex Immunohistochemistry Images | Investigating Factors Contributing to the Implementation of Personalized Medicine Practices in Clinical Oncology

The complex network of interactions in the tumor microenvironment (TME) has yet to be fully elucidated. Spatial proteomic techniques like multiplex immunohistochemistry (mIHC) have made it possible for researchers to investigate numerous protein biomarkers simultaneously. While mIHC has been used to investigate tumor infiltration and some intercellular interactions on a spatial basis, the field has yet to establish a means of harnessing the power of observing and comparing the spatial interactions between all available markers for analysis simultaneously. To accomplish this, we propose a framework for analysis that incorporates a custom algorithm quantifying cell neighbors and machine learning techniques to elucidate the most important intercellular relationships investigated by immunologists and clinical researchers. To develop this framework, we used samples of tumor biopsies from a cohort of small cell lung cancer (NSCLC) patients. We developed neighborhood profiles surrounding each defined cell as features for analysis. Through orthogonalized Partial Least Squared Discriminant Analysis (OPLSDA), we connected the neighborhood profile of cells with their phenotype with high accuracy. Features with high Variable Importance in Projection (“VIP”) scores highlighted intercellular relationships which have significantly strong associations to a given phenotype, implicating a relationship to be explored further by detailed pathway analysis. In a case study focusing on helper T cells, we found strong associations between interferon-gamma (IFNγ) expression and colocalization with activated T lymphocytes and Natural Killer (NK) cells. Additionally, we have established a similar pipeline that uses the cumulative neighbor scores between many cell types within each tissue region to separate different tumors based on broad clinical features. In an example distinguishing tumors by pathologic grade, VIP scores highlighted interactions between MHC-I expressing tumor cells and T lymphocytes. These methods have demonstrated to be useful in preliminary investigations of mechanism of action by immune cells in the TME and predictive power of tumor profiling. In the future, this is intended to be used towards implementing precision medicine techniques into immunotherapy treatment.

Chemotherapy and radiation have been long standing treatments for a wide array of cancers and have largely been harmful to patients. Their major aim has been to kill malignant cells by means of cytotoxicity.
However, this leaves healthy cells in the body to be killed or harmed alongside them. Their use is toxic to patients and overall contributes to a decreased quality of life and has limited efficacy in extending survival and curing patients. In the last twenty years, immunotherapy gained traction as a feasible and effective form of therapy. Instead of using cytotoxic mechanisms to kill cells, it breaks down barriers that cancerous cells put up to suppress the immune system. Upon activation, the immune system effectively destroys the tumor without destroying uninvolved tissue. While immunotherapy has the power to completely transform the current landscape of cancer treatment, there are technical and sociotechnical barriers to immunotherapy completely transforming cancer treatment as we know it. My technical problem looks to improve the informed selection of immunotherapy treatment to improve clinic use. My STS project focuses on the cultural and organizational factors surrounding the uptake of similar implementations of precision medicine cancer therapy into the clinic, so that precision immunotherapy may be modeled in an educated way.

School of Engineering and Applied Science

Bachelor of Science in Biomedical Engineering

Technical Advisor: Dolatshahi Sepideh

STS Advisor: Wylie Caitlin

Technical Team Members: Goundry Katherine