Chemistry central journal - Essay Example

A prediction was made successfully five out of seven times (Schmuker, de Bruyne, Hähnel, and Schneider, 2007) on an odorant’s activity. Presently the field is not error proof. Although the study of ORNs and SARs between odorant and activated receptor has grown in recent years, much is left to learn. This study could only predict tested: models by recording in vivo receptor neuron responses to a new set of odorants and successfully predicted the responses of five out of seven receptor neurons.

Correlation coefficients ranged from 0.66 to 0.85, demonstrating the applicability of our approach for the analysis of olfactory receptor activation data. (Schmuker, de Bruyne, Hähnel, and Schneider, 2007) ORNs to 47 ORNs in response to stimulation with odorant molecules” (Schmuker, de Bruyne, Hähnel, and Schneider, 2007). The responses of the Drosophila ORNs to forty-seven odorants were measured by electophysiological in vivo recordings from de Bruyne, Foster, and Carlson 2001 study “Odor coding in the Drosophila antenna.

” In this 2001 study, the activity in a given compound was classified as “active’, ‘inactive’, or ‘uncertain’, depending on the spike rate it elicits in the ORN” (de Bruyne, Foster, and Carlson, 2001). The ‘uncertain’ data was not used when training the ANNs for specific ORNs for this study. Then Schmuker, de Bruyne, Hähnel, and Schneider (2007): trained 30,000 ANN models per ORN, selected those with the highest predictive power, and used them to predict ORN responses to 21 compounds, which were subsequently tested in vivo (in the following referred to as "test data").

We also assayed ten compounds that had already been tested in the previous study. The MATLAB Neural Network Toolbox was used for ANN modeling, employing backpropagation training with a gradient descent algorithm as implemented in MATLABs traingdx function. (Hertz, Palmer, and Krogh, 1991) A well-trained, well-generalizing model will have a high