Cancer is a particularly daunting disease to treat because, in its most insidious forms, cells are disseminated throughout the body, rendering surgery and local therapies ineffective. Hence, systemic disease must be targeted by its metabolic, physiological or molecular phenotype rather than by its location. Both tumour metabolism and physiology have been exploited to target therapies with some success. There is every reason to expect that these approaches will continue to advance therapies that discriminate tumour metabolism from that of normal tissues. As an alternative, molecular phenotyping (pharmacogenomics) is a relatively new science, and holds great promise for development of novel therapies and approaches. The discipline of pharmacogenomics has, to date, involved segmentation of patients according to their protein expression patterns in order to direct existing therapies to those populations who stand to benefit the most. In this communication, the authors propose a further application of this technology to develop agents that are reverse-engineered to explicitly target a patient's expressed protein patterns.