| As we enter the post-genome era with readily available technologies for global profiling of gene expression, such as microarrays and Serial Analysis of Gene Expression (SAGE), there is great potential for discovering underlying gene networks that could not be possible with single gene analyses. One bottleneck in the investigation of rich datasets of gene expression is the lack of adequate tools for analyzing and interpreting the high dimensional transcriptomic data. The overarching goal of this dissertation is to explore the chromosomal organization of gene expression as a framework for analyzing transcriptomes of human tumors. Cancers are characterized by regional genetic damage (e.g. chromosomal imbalances, amplifications, and deletions) that potentially affect the expression of many genes.; To achieve the goal of this dissertation, three specific projects were undertaken. First, a model-based scan statistic method was developed to define regions of increased and decreased gene expression in tumors. Second, the model-based scan statistic was applied to an ovarian cancer gene expression data set, and the predictive capacity of this method to identify regions of amplification based on increased chromosomal regions of gene expression was assessed. Finally, the model-based scan statistic was used to define regions of amplification and deletion based on genome-wide copy number data from a subset of those ovarian tumors analyzed in the previous project. These results were then compared to those obtained from the expression data to assess the regional effect of copy number alteration on regional alterations in gene expression.; Overall, this dissertation demonstrates the importance of and the additional information gained when moving beyond a single gene view and incorporating the chromosomal organization of genes in the analysis of gene expression data obtained from cancer studies. With methods of genomic analysis similar to the one developed and applied in this dissertation, we are now able to identify, characterize, and predict genomic aberrations based on gene expression changes, which may lead to practical advances in the medical treatment of cancers. |
