Date of Award

Spring 5-15-2021

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



ABSTRACT OF THE DISSERTATIONIdentifying Subpopulation-level Methylation Patterns in Hematological Malignancies by Jerry Fong Doctor of Philosophy in Biology and Biomedical Sciences Computational and Systems Biology Washington University in St. Louis, 2019 Associate Professor John Edwards, Chair

DNA methylation is an epigenetic modification that has been implicated in X-inactivation, retrotransposon silencing, and genomic imprinting. DNA methylation patterns help govern aspects of normal cell maturation and development, and they are typically altered in cancer. In principle, an epigenetic change such as DNA methylation could contribute to subclone expansion in cancer, but identifying such subclones remains challenging, in large part because of tumor heterogeneity. There are many causes of heterogeneity within a cancer sample, including spatial organization within a tumor, the presence of stromal and immune cells, and competition among multiple cancer subclones. Each of these causes may have distinct biological relevance. Though newer experimental methods such as single-cell genome-wide methylation profiling may help address heterogeneity, newer techniques may not be available for the many already completed genomic sequencing studies that have been conducted on heterogeneous samples. Additionally, genome-wide methylation profiling (e.g. bisulfite sequencing) is also becoming commonplace for studies that seek to study epigenetic changes in a sample. As such, there is a need for computational methods to analyze subpopulation-level methylation changes from heterogeneous samples. I have developed a novel method, DXM, to deconvolve the bisulfite sequencing data from a heterogeneous sample into its main subpopulations, their relative prevalence, and their respective methylation profiles. DXM does not require prior knowledge of the number of subpopulations or types of cells to expect. I benchmarked the performance of DXM across many mixture conditions and demonstrated vast improvement over existing deconvolution methods. In this process, I also identified a set of methylation differences that could be found even in sorted cell subpopulations, which may reflect how there can be additional relevant biology beyond what is normally ascribed to different cell subpopulations by surface markers. I then further validated DXM predictions for subpopulation-methylation profiles in four primary DLBCL samples using FACS-sorted CD4+ and CD19+ cells from each sample. This degree of validation offers a strong degree of confidence for using DXM to analyze subpopulation level methylation events of any given heterogeneous sample, and I expect DXM to be broadly useful in identifying subclonal methylation events that contribute to cancer progression. I then sought to analyze how subpopulation-methylation in both DLBCL and AML related to relapse and to known genetic subclones. As proof-of-concept using a small cohort of 18 patients, I found that patients presenting with fewer DXM-identified intrasample differentially methylated regions (i-DMR) were more likely to relapse. I then utilized DXM to a cohort of 138 paired diagnosis-relapse AML samples to identify potential subclones whose methylation profiles improved their fitness. Next, given the importance of IDH1 mutations in AML and the availability of AML31, a well-characterized AML samples with a high variant allele frequency for IDH1 mutation (~60%), I sought to utilize DXM to relate subclonal methylation changes to known genetic mutations such as IDH1. I validated a previously identified “CpG-island methylator phenotype” for IDH1-mutations in AML31, but this did not hold for a larger cohort of 119 AML samples. As such, I proposed a new framework for identifying subclonal methylation changes associated with genetic subclones. Taken together, my work highlights the importance of subclonal methylation events in AML and offers a foundation to systematically analyze the potential epigenetic subclones of any cancer.


English (en)

Chair and Committee

John R. Edwards Daniel Link

Committee Members

Obi L. Griffith, David H. Spencer, Jacqueline E. Payton,

Available for download on Tuesday, May 21, 2041

Included in

Biostatistics Commons