ABSTRACT OF THE DISSERTATION Modeling p53 Transcriptional Regulation

The p53 protein has been called the “gatekeeper ” of the cell. After DNA damage, p53 transcriptionally activates downstream pathways to prevent cancer from occurring. Target genes are activated to cause cell cycle arrest, DNA repair, cell senescence, and apoptosis. However, the exact transcriptional program that determines the specific outcome of a certain cell stress is not completely understood. We present an analysis to help shed light on the mechanisms of transcriptional regulation by the p53 protein. First, we present a detailed analysis of the known modes of p53-regulation, and a dataset of 160 functional human p53-binding sites curated from the literature. Second, we present a new method (called p53HMM) to model p53-binding sites using Profile Hidden Markov Models (PHMMs). This new method is the most accurate predictor of functional p53 single-sites and cluster-sites to date. Third, we show that functional sites with low estimated relative affinity scores are highly correlated with distances from the TSS. Fourth, we show that the capability to fold into a non-linear cruciform DNA structure is an important predictor in estimating the overall binding affinity of a functional p53-binding site. We ii use UNAFold to calculate free energies and probabilities of p53-binding sites folding into different non-linear cruciform structures. Fifth, we present a new motif-finding algorithm (called PURE) that uses relative entropy to find over- and under-represented motifs near functional p53-binding sites. The goal is to find possible motifs (like co-factor motifs) that can help designate functional p53-binding sites, thereby reducing the false positive rate that currently plagues motif-finding algorithms. ii