DNA Methylation and Human Diseases: Applied and Methodological Studies

Abstract

DNA methylation plays important roles in regulating gene expression and chromosome integrity via addition of methyl groups to cytosine residues. A growing number of human diseases have been found to be associated with aberrant DNA methylation. DNA methylation also provides potential cancer biomarkers and therapeutic targets. Lung cancer remains the leading cause of cancer-related mortality worldwide, with an estimated 224,390 new cases and 158,080 deaths in the U.S. alone in 2016. In Chapter 1, the association between overall survival and DNA methylation of a tumor-suppressor gene named LRRC3B was investigated in 1,230 early-stage non-small cell lung cancer patients. It provides evidence of plausibility for building biomarkers on DNA methylation of LRRC3B for overall survival of early-stage non-small cell lung cancer, thus filling a gap between previous in vitro studies and clinical applications. Acute respiratory distress syndrome (ARDS) is a severe lung disease with a mortality rate of over 40% among moderate-to-severe patients. In Chapter 2, we conducted an epigenome-wide association study (EWAS) between DNA methylation and 28-day survival time in 185 moderate-to-severe ARDS patients from intensive care units (ICUs). We identified four CpG sites that were significantly associated with ARDS survival in two independent cohorts. By integrating all four statistically significant methylation sites, we built a methylation risk score for each patient. Patients with a higher methylation risk score had a significantly higher hazard of death within 28 days. As the next generation sequencing (NGS) technology becomes more and more affordable, sequencing-based methylation measurements would become popular in epigenetic studies. But it also presents challenge in data analysis, including co-methylation network analysis. In Chapter 3, we showed that while standard sequencing-based methylation measurement provides an unbiased estimate for the methylation level, it leads to a biased estimate for correlation between methylation sites. In Chapter 4, we also showed that sequencing-based methylation measure also leads to biased estimates for linear regression coefficients. We proposed a new method to obtain unbiased estimate for methylation correlation and linear coefficients based on bisulfite sequencing data. We demonstrated its performance using various simulation settings as well as real data generated using bisulfite sequencing technique.