| Summary: | In this thesis, a number of techniques are developed for the integration of high-throughput genomic and clinical data. These techniques are motivated by, and demonstrated upon, a small scale study of advanced sporadic invasive epithelial ovarian cancer, CTCR-OV01. In the first part of this thesis, clinical data from the CTCR-OV01 study are introduced. A set of biologically motivated hypotheses on the CTCR-OV01 study, based on existing literature, is described. A novel approach to analysis of continuous mRNA expression in terms of hypotheses on discrete clinical sets is developed; this work extends conventional methods by allowing hypotheses that predict both similarities within and differences between sets of clinical sets. These methods are demonstrated on simulated data, following which tests on real data from the CTCR-OV01 study show low false discovery rates in assessing hypotheses on the data. Comparisons with alternative approaches show that the method is of value. An alterative approach to mRNA expression analysis, in which mRNA expression data is integrated with both continuous and discrete clinical data in a mixed-effects model is then presented. Methods of producing a continuous measure of response are discussed. A number of genes selected by the methods developed are validated by experiment. A set of novel statistical methods are developed for the analysis of array CGH data. Empirical Bayes techniques that are able to assess a number of hypotheses on array CGH data are established and tested on CTCR-OV01 data. Results from this analysis are encouraging from a biological standpoint and show some correlation with results acquired in mRNA expression analysis.
|
|---|
