Olfactory receptors evolved to provide animals with ecologically and behaviourally relevant information. The resulting extreme sensitivity and discrimination has proven useful to humans, who have therefore co-opted some animals' sense of smell. One aim of machine olfaction research is to replace the use of animal noses and one avenue of such research aims to incorporate olfactory receptors into artificial noses. Here, we investigate how well the olfactory receptors of the fruit fly, Drosophila melanogaster, perform in classifying volatile odourants that they would not normally encounter. We collected a large number of in vivo recordings from individual Drosophila olfactory receptor neurons in response to an ecologically relevant set of 36 chemicals related to wine ('wine set') and an ecologically irrelevant set of 35 chemicals related to chemical hazards ('industrial set'), each chemical at a single concentration. Resampled response sets were used to classify the chemicals against all others within each set, using a standard linear support vector machine classifier and a wrapper approach. Drosophila receptors appear highly capable of distinguishing chemicals that they have not evolved to process. In contrast to previous work with metal oxide sensors, Drosophila receptors achieved the best recognition accuracy if the outputs of all 20 receptor types were used.