Using the telomerase knockout mouse model, we developed mouse cancer models with unstable genomes (which generates recurrent amplifications and deletions) and used these models as a platform for comparative oncogenomics profiling to identify new cancer genes. These studies demonstrated that humans and mice target many of the same genes in the development of cancer. In another series of studies, we have taken advantage of the large human cancer genome profiles to develop strategies to identify new cancer-specific vulnerabilities. These efforts led to two new concepts — collateral lethality and synthetic essentiality — for the identification of new cancer-specific targets.