Epigenetic modifications, such as DNA methylation, play a significant role in human disease. DNA methylation is altered in cancer, and DNA methylation-induced silencing of promoters of tumor-suppressors is thought to play a key role in tumorigenesis. Though there is an increasing number of methylation sequencing datasets generated from cancer, they are confounded by tumor heterogeneity. Understanding the subclonal architecture underlying heterogeneity is important because individual cancer subclones impact cancer growth, treatment resistance, and metastasis. Increased ?epigenetic heterogeneity,? or variation in DNA methylation patterns, was determined to be correlated with worse prognosis in patients with diffuse large B-cell lymphoma (DLBCL) and acute myeloid leukemia (AML). However, these studies only profiled epigenetic heterogeneity at a global scale and did not analyze methylation of individual subclones. Many computational methods have been developed for this purpose in epigenome-wide association studies (EWAS), but these methods are not applicable for methylation sequencing data. As such, there is a need to develop new computational methods to study subclonal methylation profiles in cancer. I will develop a novel computational method (DXM) that deconvolves DNA methylation of heterogeneous clinical samples into their major subpopulations, their prevalence, and their respective DNA methylation profiles (Aim 1). DXM will be developed on simulations generated from data available from both the Roadmap Epigenomics Project and the Blueprint Epigenome Project and will be validated with methylation sequencing on DLBCL samples. I will then leverage DXM to study subclonal methylation profiles in AML (Aim 2). I hypothesize that methylation can confer fitness advantages in epigenetic subclones of AML, leading to clonal expansion following therapeutic intervention and subsequent relapse. Using DXM, I will first identify epigenetic subclones from whole genome bisulfite sequencing data from AML patients with well-characterized genetic subclonal architecture and compare the epigenetic subclonal architecture to that of the genetic subclones. In a second dataset containing methylation sequencing data from paired diagnosis-relapse samples, I will apply DXM, identify subclonal methylation profiles, and characterize all genes impacted by subclonal methylation to determine if they are expected to confer fitness advantages. Finally, I will test if mutations known to impact DNA methylation in AML (DNMT3A, IDH1/2) alter epigenetic subclonal expansion. Taken together, this proposal will result in a new computational approach to interpret epigenetic data from heterogeneous clinical samples as well as identification of DNA methylation changes in epigenetic subclones of AML that contribute to disease progression and correlate with prognosis.
Studies that characterize which variations in cancer relate to worse outcome may do analysis on the entire tumor sample, which is composed of many different types of cells, each with distinct biology. The goal of this proposal is to develop a novel computational method to analyze variations in cancer at the level of cell-types. These results can be used to improve patient prognostics and care in multiple cancers.