The heterogeneous population of cells conforming malignant tumors is the result of Darwinian evolution and selection. There is yet much to unveil about these processes, which are key to understanding tumor aggressiveness, response to therapy and the emergence of resistance. We study evolutionary dynamics from a twofold perspective: theoretical and practical. For the first, we work with discrete and continuous phenotypically structured mathematical models to describe phenotypic instability and evolvability. For the latter, we use medical imaging and clinical data to study metabolic scaling laws arising in human cancers.