Somatic evolution is increasingly being recognized as the main driver not only in cancer progression but importantly in the emergence of resistance to existing treatments. Much of the focus in the research of evolution in cancer has been devoted to leveraging existing -omics tools to evaluate how tumors change but Darwinian evolution requires us to understand tumor population heterogeneity and the environmental role in selecting between available phenotypes. Studying selection in cancer evolution requires new tools that integrate clinical data, experimental approaches and mathematical modeling. In this talk I will describe some of our recent work where we use computational and experimental approaches to capture tumor-stromal interactions in multiple myeloma and in breast cancer. Our work shows how environmental selection can shape intra tumor heterogeneity and explain the emergence of resistance as well as suggest ways in which treatment could be applied differently to minimize this risk.