Phenotypical analysis of tumor microenvironment

Establishing dependency between the genotypes and the phenotypes in any model system is not straightforward especially for complex processes such as cancer. This is primarily due to the fact that the phenotype domain is inconceivably large owing to phenotypes occurring at multiple hierarchies. Although the entire genome sequences of many model systems are now available, the corresponding growth in the phenotype domain lags far behind. This lead to considerable growth in phenotyping technologies in the recent past across multiple hierarchies.;In this dissertation, I present a computational framework for quantitative phenotypical analysis primarily directed at cellular scale. It has been recently known in cancer biology that both the tumor cells and the neighboring normal cells collectively constituting the tumor microenvironment, work together in the initiation and progression of tumors. I describe how changes in the genetic make up of those neighboring normal cells of a specific type - fibroblasts, affect their physical features like nuclear morphology. Although the significance of nuclear morphology on tumor progression is well known, there is a strong need to incorporate other phenotypes like nuclear appearance and spatial context especially in the case of gene knockouts like P53 that plays a crucial role in DNA repair. I show how these phenotypes are computed and are affected to different degrees based on specific gene knockouts. Thus the proposed framework could be used as a hypothesis generation tool.;Besides characterizing sub-cellular features, I propose an algorithm to reconstruct larger structures like blood vessels and epithelial ducts based on a model-based approach that uses a tracking paradigm to perform the reconstruction.;Finally, I show ways to visualize both the tumor microenvironment as well as the computed features using volume and information visualization techniques, and propose an algorithmic technique to reconstruct isosurfaces by avoiding smaller area triangles and edges by extending the existing Marching Cubes lookup table.