Improving odorant chemical class prediction with multi-layer perceptrons using temporal odorant spike responses from drosophila melanogaster olfactory receptor neurons.

Abstract

In this work, we examine the possibility of improving the prediction performance of an olfactory biosensor through the use of temporal spiking data. We present an Artificial Neural Network (ANN), in the form of an optimal hybrid Multi-Layer Perceptron (MLP) system for the classification of chemical odorants from olfactory receptor neuron spike responses of the Drosophila melanogaster fruit fly (DmOrs). The data used in this study contains the responses to 34 odorants from 6 individual DmOrs, of which we exploit the temporal spiking responses of a 500ms odorant stimulus window. We report, for the first time, the difference between the classification performance of the temporal spiking data to an equivalent spontaneous scalar dataset that we have reported previously. We demonstrate that a higher prediction (%) was obtained when using the temporal data, in which a greater number of validation odorants are identified to their correct chemical class. This work presents a novel technique to improve the classification performance of an olfactory biosensor, whilst maintaining a limited sensory array of 6 DmOr receptors.