Computational methods for analyzing complex high-throughput data from cancers
Year: 2014
Type: dissertation
Abstract: Cancers are a heterogeneous group of diseases that cause 7.6 million deaths yearly worldwide. At the cellular level, cancer is characterized by increased proliferation and invasion of tissue. These phenotypes are caused by environmental or inherited factors that increase the mutability of the genome, leading to dysregulation of a number of cellular processes. Identifying the genotypic changes and their phenotypic consequences is key to accurate diagnosis and prognosis, as well as improved treatment regimens. Cancer cells can be investigated at a genome-wide scale using high-throughput measurement techniques such as DNA sequencing and microarrays. These rapidly evolving technologies provide experimental data that have two challenging characteristics: the volume of data is large and data are structurally complex. These data need to be analyzed in an accurate and scalable manner to arrive at biomedically relevant conclusions. I have developed three computational methods for analyzing high-throughput genomic data, and applied the methods to experimental data from three cancers. The first computational method is an extensible workflow framework, Anduril, for organizing the overall software structure of an analysis in a scalable manner. The second method, SPINLONG, is a flexible algorithm for analyzing chromatin immunoprecipitation followed by deep sequencing (ChIP-seq) data from complex experimental designs, such as time series measurements of multiple markers. The third method, GROK, is used for preprocessing deep sequencing data. Its design is based on a mathematical formalism that provides a succinct language for these operations. The experimental part studies gene regulation and expression in glioblastoma multiforme, and breast and prostate cancer. The results demonstrate the applicability of the developed methods to cancer research and provide insights into the dysregulation of gene expression in cancer. All three studies use both cell line and clinical material to connect the molecular and disease outcome aspects of cancer. These experiments yield results at two conceptual levels. At the holistic level, lists of significant genes or genomic regions provide a genome-wide view into genomic alterations in cancer. At the specific level, we focus on one or a few central genes, which are experimentally validated, to provide an accessible starting point for understanding the results. Together, the thesis focuses on understanding the complexity of cancer and managing the complexity of genome-wide data.
Source: Työväentutkimus Vuosikirja
Author: Kristian Ovaska
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W1482958535