Aneuploidy is a common feature of cancer and several lines of evidence suggest that cytogenetic aberrations can significantly influence cancer diagnosis, prognosis, and treatment. While molecular genetic based methods, such as comparative genomic hybridization (CGH), have traditionally been used to determine cell karyotypes, recent transcriptional profiling studies have suggested that it is possible to predict cytogenetic changes from microarray gene expression data. A technique we term comparative genomic microarray analysis (CGMA) is based on the observation that gene expression values show expression biases, either increased or decreased, in regions of chromosomal gain and loss, respectively. In tumor samples, CGMA predictions are made by mapping gene expression values to the public human genome assembly and scanning for genomic regions that contain a statistically significant upwards or downwards gene expression bias. While the first-generation of algorithms that identify gene expression biases produce reasonably good cytogenetic predictions, it is likely that more sophisticated algorithms could produce better results. The R21 phase of this proposal focuses on implementing and testing a set of refined CGMA algorithms to make more accurate and higher resolution cytogenetic predictions. Cytogenetic and transcriptional profiling data obtained from a small set of colon tumors will be used for algorithm testing. The R21 milestone is to establish a CGMA prediction method that matches CGH determinations with high accuracy. The R33 phase will focus on developing algorithms to make CGMA predictions across multiple samples and will test if frequently changed regions identified by CGMA match those regions previously identified by CGH. For the R33 phase, hepatocellular carcinoma (HCC) and renal cell carcinoma (RCC) will serve as models because large sets of gene expression and cytogenetic profiling data are currently available. Historically, candidate genes have been identified by determining if a gene located within a region of frequent cytogenetic change is either mutated or misregulated. In this proposal, candidate genes will be identified from the HCC and RCC gene expression profiles by first using CGMA to locate frequently changed genomic regions and then by using traditional gene expression analysis to identify abnormally expressed genes located within these regions of frequent cytogenetic change.
