In humans, the adaptive immune system uses the exchange of information between cells to detect and eliminate foreign or damaged cells; however, the removal of unwanted cells does not always require an adaptive immune system1,2. For example, cell selection in Drosophila uses a cell selection mechanism based on 'fitness fingerprints', which allow it to delay ageing3, prevent developmental malformations3,4 and replace old tissues during regeneration5. At the molecular level, these fitness fingerprints consist of combinations of Flower membrane proteins3,4,6. Proteins that indicate reduced fitness are called Flower-Lose, because they are expressed in cells marked to be eliminated6. However, the presence of Flower-Lose isoforms at a cell's membrane does not always lead to elimination, because if neighbouring cells have similar levels of Lose proteins, the cell will not be killed4,6,7. Humans could benefit from the capability to recognize unfit cells, because accumulation of damaged but viable cells during development and ageing causes organ dysfunction and disease8-17. However, in Drosophila this mechanism is hijacked by premalignant cells to gain a competitive growth advantage18. This would be undesirable for humans because it might make tumours more aggressive19-21. It is unknown whether a similar mechanism of cell-fitness comparison is present in humans. Here we show that two human Flower isoforms (hFWE1 and hFWE3) behave as Flower-Lose proteins, whereas the other two isoforms (hFWE2 and hFWE4) behave as Flower-Win proteins. The latter give cells a competitive advantage over cells expressing Lose isoforms, but Lose-expressing cells are not eliminated if their neighbours express similar levels of Lose isoforms; these proteins therefore act as fitness fingerprints. Moreover, human cancer cells show increased Win isoform expression and proliferate in the presence of Lose-expressing stroma, which confers a competitive growth advantage on the cancer cells. Inhibition of the expression of Flower proteins reduces tumour growth and metastasis, and induces sensitivity to chemotherapy. Our results show that ancient mechanisms of cell recognition and selection are active in humans and affect oncogenic growth.