Abstract
Current electronic noses, or e-noses, that employ insect odorant receptors (Ors) as their sensory front end are potentially limited by the fact that the Ors come from a single species. In addition, a realistic e-nose also demands low numbers of Ors at its sensory front end due to the difficulties of receptor/sensor integration and functionalisation. In this work, we report the first investigations of a `Super E-Nose' that incorporates Ors from both the model organism Drosophila melanogaster fruit fly (DmOr) and the mosquito, Anopheles gambiae (AgOr). Furthermore, we report how an Artificial Neural Network (ANN), in the form of a hybrid double hidden layer Multi-Layer Perceptron (MLP), can be used to determine the optimal Ors that provide the best prediction performance in the classification of unknown odorants into their respective chemical class. Our findings demonstrate how 3-Or arrays consisting of DmOr only, AgOr only, or cross-species DmOr-AgOr combinations correctly classified all unknown odorants of the validation set. In addition, we report that all 3-Or combinations perform equally well as the complete 74 DmOr-AgOr array. Thus, the results of this work support further investigation into cross-species `Super E-noses' coupled with hybrid MLPs for the classification of unknown odorants.
