Model-based gene expression analysis
Abstract
Systems biology is a rapidly emerging branch of modern biology, aiming at elucidating the complex regulatory mechanisms using an ever-growing amount of molecular data and advanced mathematical, computational tools. To understand the genome-wide complexity of transcription regulation, my research objective was to develop a model-based approach to predict critical transcription-factor binding motifs (TFBMs) using microarray-derived mRNA expression levels and genomic DNA sequences. By defining an activation level of TFBMs as a unique state variable, the mathematical model was built to establish a quantitative relationship between the observed mRNA expression level and frequencies of TFBMs in regulatory DNA regions. Identification of the critical set of TFBMs was formulated as a combinatorial optimization problem using three biological systems including the shear-stress responses in synovial cells, the interleukin-1 responses in chondrocyte cells, and the differentiation processes in human chondrogenesis. Mathematical manipulations such as singular value decomposition, genetic algorithm, particle swarm optimization, and ant algorithm were employed to predict the critical set of TFBMs, whose number was estimated from Akaike information criterion. First, the results show that the described model is useful to predict and evaluate the critical set of TFBMs from high-throughput gene expression data and regulatory DNA sequences. Second, a genome-wide transcription network can be built by evaluating generality and specificity of the critical TFBMs in different biological model systems. Lastly, it is important to validate the predicted TFBMs using experimental assays because of uncertainties in existing knowledge, particularly definition of regulatory DNA regions. In conclusion, the described mathematical and computational approach can help biologists to raise testable hypotheses in transcription regulation. Combined with the biochemical assays such as a promoter competition assay, the model-based approach is expected to contribute to the understanding of system-wide regulatory processes and elucidating functional significance of TFBMs.
Degree
Ph.D.
Advisors
Doerschuk, Purdue University.
Subject Area
Biomedical research|Genetics
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.