Abstract
Cancer is a major source of morbidity and mortality, second only to heart disease as the leading cause of death in most developed countries. Our ability to treat the disease depends heavily on our understanding its aetiology. Different types of cancer often respond best to different courses of treatment, yet our understanding of cancer classification remains imperfect. Novel expression-monitoring technology provides unique opportunities to learn more about cancer at the molecular level, to improve classification methods, and to progress towards personalised cancer treatment. This review describes the techniques that make such advances possible, summarises recent work in the area, and discusses future developments needed to realise the potential of a thorough molecular classification of cancer.