Systems approaches to identify molecular signatures from high-throughput expression data: towards next generation patient diagnostics

Translational Bioinformatics;Systems Biology;Transcriptomics;Classification

Sung, Jaeyun

关键词: Translational Bioinformatics;    Systems Biology;    Transcriptomics;    Classification;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/34469/Sung_Jaeyun.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

The advent of high-throughput (so called “omics”) technologies for the comprehensive and rapid measurement of virtually all molecular components within human cells, tissues, organs, and serum has led to the generation of a tremendous amount of raw information.However, converting large-scale data to essential knowledge on human health and disease mechanisms has been a significant challenge thus far to the scientific and medical community.To this end, we call for the systems biology approach to medicine (Systems Medicine), in which disease is viewed as a result of one or more disease-perturbed biomolecular networks caused by DNA mutations, pathogenic microorganisms, or environmental toxins.These perturbations lead to alterations in the abundance of intra/extracellular biomolecules, which offer diagnostic clues to the presence, as well as progression of disease.In this dissertation, I describe my investigations of developing computational systems approaches that aim to identify robust molecular diagnostic signatures from omics data, and to thereby advance personalized medicine and blood diagnostics.Specifically, this work makes three novel contributions based on the analysis of publicly archived high-throughput expression data: (1) Development of two classification algorithms based on relative expression reversals of biological features, which demonstrate robust phenotype distinction in binary and multi-category scenarios; (2) Discovery of organ-level diagnostic signatures and addressing of batch effects through multi-study integration of brain cancer transcriptomes; and (3) Identification of conserved expression patterns in mRNA and protein profiles from human cancers for prediction of relative feature abundances across heterogeneous data types for in vivo monitoring of disease-perturbed networks.We hope the work presented in this dissertation will play a significant role in bringing omics-based technologies into clinical practice, and lead to innovative medical applications that enhance our understanding of human health and disease.