Cancer occurs through the stepwise acquisition of somatic genetic alterations that lead to the deregulation of critical cellular phenotypes such as growth, survival, motility, and genome stability. The identification of the oncogene and tumor suppressor gene targets of such somatic alterations has been an area of active investigation over the last three decades, resulting in tremendous advances in both the understanding of cancer pathogenesis and the identification of novel therapeutic targets.

Somatic copy number alteration (SCNA) represents a common mechanism by which developing tumors alter gene expression and function. Due to their large size compared to other classes of cancer-causing somatic alteration, SCNA has been particularly amenable to systematic identification and characterization through cytogenic and, more recently, array-based measuring technologies. However, a critical challenge in the genome-wide analysis of SCNAs is distinguishing the alterations that drive cancer growth from the numerous, apparently random alterations that accumulate during tumorigenesis.

A common approach to this problem is to study large collections of cancer samples, on the notion that the causal alterations should recur across independent samples and therefore be identified at a significantly greater frequency than the functionally irrelevant 'passenger' alterations. While a number of statistical methods have recently been developed to facilitate the detection of genomic regions undergoing SCNA at significant frequencies, these methods have generally been limited in their ability to confidently pinpoint the genes being targeted by such events. Moreover, while numerous large studies of individual cancer types have recently been performed, few studies have attempted to systematically compare the pattern of SCNA across multiple cancer types, making it impossible to quantify the extent to which biologically significant SCNAs are restricted to individual cancer types or shared across multiple types.

In this thesis, we describe the development of significantly improved statistical methods for the identification of genes being targeted by focal SCNA across human cancers and the results of applying these methods to large datasets derived from multiple cancer subtypes. These tools and analyses shine new light on the shared patterns of SCNA occurring across cancer and have enabled the discovery of several validated and candidate human oncogenes. In particular, we present a novel lineage-specific role for SOX2 in the initiation of lung and esophageal squamous cell cancers and demonstrate that SCNA targeting anti-apoptotic BCL2 family members is a common yet previously unappreciated mechanism by which cancers of diverse tissue origins evade apoptosis.