Tumors are characterized by numerous genetic alterations ranging from point mutations in critical genes to a multitude of major alterations in the structure and number of chromosomes. Translocations, deletions, duplications and alterations in ploidy occur in almost all tumors and alter the expression of thousands of genes. The large number of alterations in the genome of a tumor cell can make it extremely hard to identify the specific genes whose change in expression is causal in the neoplastic process. Yet this knowledge is important since the identification of the causal genetic changes common to specific types of tumors will eventually lead to therapies based on a molecular profile. We propose to develop an automated method to generate a genome wide molecular profile of mouse and human tumors in a single experiment. This will enable the whole genome the whole genome to be analyzed in hundreds of tumors, which should facilitate the identification of common causal genetic changes. The technology to perform this analysis is based on comparative genomic hybridization (CGH) of fluorescently labeled DNA. We are applying CGH to high density microarrays of MAC clones of genomic DNA which have been covalently coupled on a glass microscope slide. To facilitate this we have developed a novel DNA-glass coupling chemistry which provides exquisite sensitivity. These arrays are co- hybridized with normal DNA and tumor DNA labeled with different colors and a laser scanner with photomultiplier detection is used to scan the array and convert the different colors and intensities of the fluorescence signals into an intensity ratio histogram. This provides a genome wide molecular profile of the tumor with respect to regions of the genome which are deleted or amplified compared with the normal genome at the resolution of a BAC (200kb). The """"""""array analysis"""""""" of the genome of multiple tumor samples can be rapidly accomplished by this method. This technology will initially e developed for the mouse but will subsequently be developed for the human and applied to prostate cancer. Computational tools will be developed to analyze this data. This will enable the identification of alterations in key regions of the genome specific to certain tumor types. This can be correlated with knowledge of the response of the tumor to certain therapies to provide prognostic outcomes based on a molecular profile. This should also enable the specific genes which are causal in the neoplastic process to be identified.