Identification of gene sets that predict Acute Myeloid Leukemia prognosis using integrative gene network analysis
Date
2018-08
Authors
Samimi, Hanie
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Orthogonal data types can potentially provide new opportunities to pinpoint the
underlying molecular mechanisms of diseases. However, currently-available
techniques to capitalize on information from different data types suffer from a
substantial loss of statistical power. Therefore, there is urgent need to develop
algorithms to integrate data types. In this thesis, I have developed a data
integration approach based on multi-view clustering. I demonstrate the
usefulness of my approach in prognostication of Acute Myeloid Leukemia (AML),
a particular type of blood cancer. AML accounts for 1.2% of cancer deaths per
year in the USA. AML patients are categorized into low, medium and high-risk
groups. The variable survival rate for medium-risk patients leads to difficulties in
deciding on the appropriate treatment for these patients. Current methods of
prognostication of AML use only gene expression, mutations and molecular
cytogenetic abnormalities. However, the DNA methylation data, which have
valuable information that would be useful for prognostication, have not yet been
effectively used in the existing clinical tests. In this project, I have used The
Cancer Genome Atlas (TCGA) dataset and developed a method that analyzes
both gene expression and DNA methylation data in a single model using network
analysis. The model based on this methodology correctly classified 13 out of 90
patients as high-risk, whereas they were previously labeled as medium-risk using
current clinical methods. All 13 of these cases died within two years after
diagnosis. To validate these results, I tested the method using an independent
dataset. The model labeled 11 out of 228 patients as high-risk, whereas they
were previously labeled as medium-risk based on the European Leukemia
Net (ELN) 2010 criteria. All 11 patients died within two years of diagnosis, and
their risk group is not predictable with other currently used methods.
Description
Keywords
Network analysis, Gene expression, DNA methylation
Citation
Samimi, H. (2018). <i>Identification of gene sets that predict Acute Myeloid Leukemia prognosis using integrative gene network analysis</i> (Unpublished thesis). Texas State University, San Marcos, Texas.