| dc.description.abstract | Glioblastoma is the most common malignant primary brain tumor and is rapidly fatal despite aggressive treatment with chemoradiation and therapy. Single cell approaches to uncovering intra- and inter-tumor heterogeneity have begun to illuminate new strategies for stratifying patients and identifying new therapeutic targets. In this dissertation, I describe the generation of a first of its kind brain tumor mass cytometry dataset which contains measurements of more than 30 protein and phospho-protein features on millions of cells from individual patient tumors, and an algorithm for unsupervised identification of cell phenotypes and subsets that are correlated with clinical variables of interest. By applying this new tool to the mass cytometry data set, we identified two novel glioblastoma cell phenotypes associated with either longer or shorter overall survival. These subsets can stratify patients and point to new therapeutic targets and combinatorial therapeutic strategies. Furthermore, this dataset was used to probe the differences between tumors based on their location within the brain, relative to the normal neural stem cell niche, the ventricular sub-ventricular zone. This work also describes technical and computational tools for studying solid tumors and high dimensional data that can be applied to many other human diseases. Overall, this work provides approaches for the study of proteins in single cells from solid tissues and tumors, describes the generation of a new tool that can identify prognostic cell subsets in an automatic and unsupervised manner, and highlights two new phenotypes in glioblastoma tumors that are associated with patient outcomes and can be further investigated to inform new therapeutic approaches. | |
