Unravelling the Architecture of Duplications in Tumor Genomes

Tumor cells frequently exhibit large-scale changes in their genome, including genome rearrangements (e.g. chromosome inversions and translocations), duplications, and deletions. Such genomic changes are directly implicated in the pathogenesis of cancer and can alter gene structure and regulation. While many individual rearrangements and duplications in tumors have been identified, little is known about the detailed architecture of tumor genomes. A recently introduced technique called End Sequence Profiling (ESP) allows the identification of both rearrangements and duplications in tumor genomes at high resolution [5, 3]. In contrast, array-based techniques such as comparative genomic hybridization (CGH) are very useful in the identification of duplicated material in a tumor genome, but give little information about the organization and locations of duplicated material within the genome. For example, it is difficult to infer from CGH data whether two duplicated regions are located close together (or even on the same chromosome) in the tumor genome. In some tumors, duplicated material from several disparate regions of the human genome is co-localized [1, 6]. However, the precise architecture of these complex duplications and the mechanisms that explain their formation are not completely understood. We develop a computational method that allow us to unravel the architecture of complex duplications in a tumor genome using ESP data. The technique relies on a model of breakagefusion-bridge (BFB) cycles, a biological process that produces both rearrangements and duplications in tumor genomes. We apply our method to ESP data from the MCF7 breast tumor cell line. This work complements earlier analyses of genome rearrangements [3] and duplication by amplisome [4] using ESP data.