Extensive preclinical experimentation has conceptually changed the way we perceive breast cancer, with the wide spectrum of genomic alterations governing its malignant progression now being recognized. Functional genomics has helped us identify important genetic defects that can be pharmaceutically targeted in the setting of metastatic disease. Rationally chosen combination regimens are now under clinical investigation. Recent data underline the functional importance of the tumour-associated stroma, with several candidate molecular targets now emerging. Data elucidating a cellular hierarchy within the breast cancer cellular compartment support the existence of a therapy-resistant subpopulation of breast cancer stem cells. Identification of the developmental pathways that dictate their malignant phenotype and use of high-throughput screening techniques are leading to new therapeutic avenues. In this Review, we present the biological rationale for the clinical development of more than 15 different classes of targeted agents in breast cancer, along with evidence supporting rational combinations. However, metastatic breast cancer resembles a Darwinian evolutionary system, with 'driver' mutations and epigenetic changes determining clonal selection according to branching trajectories. This evolution is reflected in the molecular heterogeneity of the disease and poses severe impediments to the successful clinical development of emerging targeted agents.