Date of Award

Winter 12-15-2018

Author's School

Graduate School of Arts and Sciences

Author's Department

Biology & Biomedical Sciences (Computational & Systems Biology)

Degree Name

Doctor of Philosophy (PhD)

Degree Type

Dissertation

Abstract

Distinguishing driver mutations from passenger mutations within tumor cells continues to be a major challenge in cancer genomics. Many computational tools have been developed to address this challenge; however, they rely heavily on primary protein sequence context and frequency/mutation rate. Rare driver mutations not found in many cancer patients may be missed with these traditional approaches. Additionally, the structural context of mutations on tertiary/quaternary protein structures is not taken into account and may play a more prominent role in determining phenotype and function. This dissertation first presents a novel computational tool called HotSpot3D, which identifies regions of protein structures that are enriched in proximal mutations from cancer patients and identifies clusters of mutations within a single protein as well as along the interface of protein-protein complexes. This tool gives insight to potential rare driver mutations that may cluster closely to known hotspot driver mutations as well as critical regions of proteins specific to certain cancer types. A small subset of predictions from this tool are validated using high throughput phosphorylation data and in vitro cell-based assay to support its biological utility. We then shift to studying the druggability of mutations and apply HotSpot3D to identify potential druggable mutations that cluster with known sensitive actionable mutations. We also demonstrate how utilizing integrative omics approaches better enables precision oncology; Combining multiple data types such as genomic mutations or mRNA/protein expression outliers as biomarkers of druggability can expand the druggable cohort, better inform treatment response, and nominate novel combinatorial therapies for clinical trials. Lastly, we improve driver predictions of HotSpot3D by creating a supervised learning approach that integrates additional biological features related to structural context beyond just positional clustering. Overall, this dissertation provides a suite of computational methods to explore mutations in the context of protein structure and their potential implications in oncogenesis.

Language

English (en)

Chair and Committee

Li Ding

Committee Members

Barak Cohen, Greg Bowman, Cynthia Ma, Chris Maher,

Comments

Permanent URL: https://doi.org/10.7936/wrwz-0e65

Share

COinS