Tomaž Curk (2007) Computational approaches for gene network discovery. PhD thesis.
This dissertation proposes a set of computational methods for inference of gene networks from heterogeneous data sources. These methods address problems of function prediction using different computational phenotypes, methods for the analysis of gene regulatory regions, and methods for decomposition of gene expression signature profiles. The main contribution of this dissertation is a method that relies on a new machine learning approach called rule-based clustering. The method can combine regulatory DNA sequence and phenotype data to infer rules that describe clusters of genes with similar phenotype and regulatory structure. We propose a set of visualizations to aid in the presentation and interpretation of inferred rules. We successfully applied the proposed methods to answer some important biological questions about the regulation of gene expression. The method for the decomposition of a gene expression signature profile can be used to place a (new) DNA microarray experiment into a biological context, which proved useful when inferring pathways and functions of genes.
Actions (login required)