Prediction of the bitterness of pharmaceutical drugs using a taste sensor with representative group data by the center of mass

The purpose of this study is to analyze the effectiveness of taste sensing system to predict the bitterness of six commercial pharmaceutical drugs including amoxicillin, paracetamol, phenylephrine, cetirizine, quinine, and ranitidine. The taste sensing system consists of Ag/AgCl as a reference electrode, platinum rod as a counter electrode and modified electrodes as working electrodes. The four different working electrodes had been fabricated by using CoPc, MnPc, FePc and ZnPc mixed with multi-wall carbon nanotube, respectively. The cyclic voltammogram of 1 mM of pharmaceutical samples was run at a scan rate of 0.09 V/s, with the potential range between-1 and 1 V. The wavelet analysis was used to reduce and extract the main feature of data. By using coordinates in the two-dimensional space of principal component analysis, the Euclidean distance of each sample was calculated with respect to coordinates of Acesulfame K. The coordinate of Acesulfame K was used as the point of reference. According to machine learning, the k-mean clustering and center of mass approach were used to locate the centroid of coordinates of the samples in the principal component analyses. The calculated Euclidean distances from the centroid point were directly associated with the intensity of bitterness of the samples. To further validated the ranking of bitterness drawn from taste sensing system, the data were correlated with the bitterness ranking drawn from gustatory sensation test. The good correlation was obtained between machine and human sensory ranking.