Background: Prostate cancer is the most common and leading cause of cancer death among U.S. males. Age is the strongest risk factor, followed by ethnicity since the rate among African Americans exceeds rates among whites, Asians, and American Indians. Prostate cancer mortality is strongly associated with the development of clinical metastatic disease. Metastatic progression in patients can develop slowly over a number of years or rapidly over a period of months and can vary widely in its clinical outcome. A reliable indicator of clinical metastatic disease is a high Gleason score (>7) in tissue during radical retropubic prostatectomy (RPP) and short doubling time (<10 months) for serum prostate specific antigen (PSA). Gene and protein expression studies suggest that fibroblast growth factors (FGFs) and the receptors FGFR1 and FGFR2 are linked to cellular proliferation, benign prostatic hypertrophy, aging prostate tissue, and survival of prostate cancer cell lines. Recently we observed that, in FGFR1- and FGFR2-inducible mice, (a) chemically-induced dimerization (CID) of FGFR1 stimulation leads to proliferation of prostate epithelial cells, (b) FGFR2 receptor stimulation does not lead to proliferation of prostate epithelial cells, and (c) prolonged FGFR1 stimulation leads to neo-vascularization and CID-independent growth. In addition, in transgenic adenocarcinoma of mouse prostate (TRAMP) mice, in which prostate epithelium p53 and Rb pathways are abrogated, there is spontaneous prostate cancer complete with distant site metastasis with progression to androgen independent disease. Given these observations, it is our intent to develop microarray-based models of gene regulatory networks in FGFR1- and FGFR2-inducible mice, TRAMP mice, and in human prostate tumors for recurrent patients with early and late PSA recurrence times.
Specific aim one : identify putative targets of the FGFR1 and FGFR2 pathways in mouse prostate tissue. We plan to induce FGFR1 and FGFR2 in inbred mice and generate microarray-based transcriptional profiles of prostate tissue using the Baylor 15k-gene spotted mouse arrays. Genes that are differentially expressed in treated and normal mouse prostate tissue will be selected as candidate targets of the FGFR1 and FGFR2 pathways. Confirmation of microarray results for the selected subset of putative targets of the FGFR1 and FGFR2 pathways will be done by northern blot analysis. Protein expression will be evaluated by western blot or immunohistochemical stains. Microarray data analysis will include fold-change analysis, cluster analysis, and principal component analysis to identify (a) unique patterns of expression (strongly up or downregulated), and (b) genes that are co-expressed with candidate genes.
Specific aim two : develop a promoter model to infer gene regulatory networks for FGFR1 and FGFR2 pathways as they relate to prostate cancer progression. Bioinformatic data analysis will include exon mapping of cDNA sequences and extraction of upstream promoter regions for genes co-expressed with FGFR1, FGFR2, other growth factors and candidate genes (e.g., KGF, EGF, VEGF, caveolin, p27, p53, Rb, cyclin D, cathepsin, c-jun, c-fos, c-myc, jun-B, etc.), construction of promoter models based on shared transcription binding sites for co-expressed genes, and performing multiple DNA sequence alignments of promoter regions of co-expressed genes. We also propose to improve the windows-based CLUSFAVOR computer program for cluster and principal component analysis of DNA microarray data, and reprogram it for use in UNIX, Linux, and Mac operating systems.
Specific aim three : determine clinicopathological significance of expression patterns of genes regulated by FGFR1 and FGFR2 pathways in prostate tumors of retrospective recurrent patients with early and late PSA recurrence times. The clinicopathological significance of microarray-based promoter models for the FGFR1 and FGFR2 pathways in Aim 1 will be compared with microarray-based promoter models for co-expressed genes in prostate tumors from retrospective recurrent patients in the prostate cancer SPORE database having the least and greatest PSA recurrence times. Prospectively, microarray-based expression of genes strongly predictive of rapid metastatic progression could be profiled in biopsy tissues of new patients for early detection and prediction of PSA recurrence in order to optimize treatment and maximize efficiency of health care utilization. The long-term goals of this study are to identify regulatory gene networks that control co-expression of target genes responsible for rapid progression of metastases. Characterization of the FGFR1 and FGFR2 pathways in prostate cancer will contribute to the molecular understanding and early detection of patients that may experience rapid progression of metastases.
| Peterson, Leif E; Coleman, Matthew A (2008) Machine learning-based receiver operating characteristic (ROC) curves for crisp and fuzzy classification of DNA microarrays in cancer research. Int J Approx Reason 47:17-36 |
