The functional and anatomical characteristics of cancer cells vary among patients. Additionally, therapeutic approaches display varying responses in different individuals and cancer types due to the anatomical and functional complexity of tumor. Prognosis, and responsiveness to therapy depends on the tissue architecture of the tumor microenvironment (TME). TME cells, including immune cells, endothelial cells, stromal cells, and their subtypes, coexist with cancer cells. The cellular and spatial architecture of the TME show significant variation across and within individuals. There is an important correlation between cell function and its spatial organization in the tissue. To unravel this organization, typical clustering is applied on spatial omics data to find discrete clusters based on local cellular abundance. Alternatively, graph-based methods are used to define clusters of cells that are closest to each other, using community-detection methods. In order to better understand the rules governing the design, formation, and interactions of the TME, the Niche-Phenotype Mapping (NIPMAP) analysis pipeline, developed based on ideas from community ecology and machine learning, was reimplemented on a new and different type of data called Hyperplexed Immunofluorescence Imaging (HIFI) from mouse glioblastoma multiforme cancer (GBM) tissue sections. NIPMAP identified cellular niches and their interactions in this dataset. Niche abundance and their cell type composition were dynamic in response to ionizing radiation (IR) treatment and relapseHausser. Testing different numbers of archetypes, resulted in different optimal niche numbers for different condition groups. So, the optimal niche is not only specific to different tissue and cancer types but also to the treatment and other experimental conditions.