Author ORCID Identifier

0000-0003-2241-5066

Date of Graduation

12-2019

Document Type

Dissertation (PhD)

Program Affiliation

Biostatistics, Bioinformatics and Systems Biology

Degree Name

Doctor of Philosophy (PhD)

Advisor/Committee Chair

Erik Sulman

Committee Member

Jason Huse

Committee Member

Krishna Bhat

Committee Member

Nicholas Navin

Committee Member

Arvind Rao

Committee Member

Ann Klopp

Abstract

Molecular classification based on mutations, expression subtypes, and copy number variants has improved diagnosis and treatment decision-making for patients with brain tumors, particularly malignant gliomas. However, the association between epigenetic signature and genetic alterations is poorly understood. For example, mutation of isocitrate dehydrogenase (IDH) is associated with genome-wide hypermethylation of CpG islands in gliomas. But other subtype-associated alterations, including telomerase reverse transcriptase (TERT) promoter mutation, alpha thalassemia/mental retardation syndrome X-linked (ATRX) mutation, chromosome 1p19q co-deletion (chr1p19q codel), and gene expression subtypes, have yet to be associated with any epigenetic signature. Therefore, we hypothesized that DNA methylation signatures can classify gliomas based on these alterations and give insight into subgroup characteristics. Machine learning models, including elastic net and random forest, were used to predict somatic mutations of IDH, TERTp, and ATRX, chr1p19q codel, and gene expression subtype of gliomas. Data from the NOA-04 randomized phase III trial were used for external validation. In total, 926 cases from The Cancer Genome Atlas were included in this study. Prediction accuracies for IDH, TERTp, and ATRX mutations, and chr1p19q codel were 100%, 98.3%, 90.48%, and 99.21%, respectively in test set. Accuracy for gene expression subtype prediction was 72.2%. The methylation-based prediction models for both ATRX and chr1p19q codel statuses proved superior to conventional assays for these biomarkers. Similarly, characteristic alterations associated with gene expression subtypes were better discriminated using methylation compared to transcriptome-based classification. DNA methylation signatures accurately predicted somatic alterations and improved over existing classifiers. The established Unified Diagnostic Pipeline (UniD) is a rapid and cost-effective diagnostic platform of genomic alterations and gene expression subtypes at initial clinical diagnosis and improves over individual assays currently in clinical use. The significant relationship between genetic alterations and epigenetic signatures indicates the broad applicability of our approach to other malignancies.

Keywords

gliomas, DNA methylation, genomic alterations, pipeline

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.