COMPUTATIONAL AND BIOCHEMICAL APPROACHES FOR THE PREDICTION AND CHARACTERIZATION OF DNA REPAIR SYNERGY IN CANCERS

This thesis Cancer is the second leading cause of death in the modern world, only trailing congestive heart disease. Many factors contribute to the mortality rate, including the diversity of tumors, development of chemoresistance, and recurrence of metastatic tumors. Conventional chemotherapeutic approaches focus on universal features of cancer derived from its rapid proliferation. Rapid proliferation inherently produces stress on metabolic systems, but also on the limiting macromolecule to cell proliferation: DNA. DNA replication and overall genomic stability are negatively impacted by rapid cell proliferation and is mitigated by dysregulation of DNA damage repair (DDR) and apoptosis pathways. Somatic mutations and aberrant gene expression provide both avenues of therapy and resistance. Understanding tumor sensitivity to optimize care expeditiously can be furthered by investigating additional targeted molecular and integrated bioinformatic approaches.

Proliferating cell nuclear antigen (PCNA) is an essential gene to numerous tumor-dysregulated processes including DNA replication and repair. Conventional wisdom would prohibit targeting PCNA due to its status as an essential gene, as directly antagonizing it would cause toxic effects in healthy cells. However, multiple groups have created small molecule antagonists capable of targeting PCNA without affecting normal cells. Some of these antagonists have been qualified biochemically providing insights into their mechanisms of action. I sought out to define the different classes of PCNA antagonists to describe possible clinical utility. This work resulted in three defined classes that are separated by their effects on general DNA damage induction, selective inhibition of DDR pathways, and DNA replication processivity.

Network theory has been utilized to integrate disparate informatic approaches to extract multilevel data with greater explanatory power than the original source data. Network approaches utilizing differential gene expression and drug response profiles have led to the discovery of novel targets and disease subtypes. I sought to use a network approach to leverage differential gene expression (DGE), gene ontology (GO) terms, and protein-protein interactions (PPI) data to determine synergistic drug combinations in cancer cell lines with disparate DDR backgrounds. I limited the scope of this work to DDR, cell cycle, DNA replication, apoptosis, and MAPK associated genes. To power this approach, I created three novel metrics of PPI network connectivity through GO term and DGE: GO Impact, GO Cohesion, and GO Adhesion. From these novel metrics, I created a visualization technique dubbed a Process Network that recharacterizes a PPI network into a set of pathway interactions. Using gene removal as a model of inhibition, I measured resulting network disruption to determine synergistic relationships. I produced a method with 90.3% specificity and 90.3% sensitivity.