Study of Relationship Between Surface Anaesthesia and Chromatographic Properties of Alkoxy Esters of Phenylcarbamic Acid by Neural Network Method. Part I.
The advanced mathematical method of neural network was employed for studying of surface anaesthetical activity of three homologous series of 2-, 3- and 4-alkoxy substituted morpholinoethyl, piperidinoethyl and azepanylethyl esters of phenylcarbamic acids. RP-HPLC capacity factors were used for the characterization of the lipophility of tested drugs and they were included to the neural network. The three-layer perceptron, that is trained by the back propagation of errors was successfully used for the supplementing of the incomplete original data matrix and also for the smoothing of the noisy biological data. The dependencies between the surface anaesthesia and the number of C atoms in the side alkoxy chain presented the peak character which is in agreement with the theoretical assumptions. Surface anaesthesia grew up in order from para to ortho position of alkoxy chain (para < meta < ortho) in individual homologous series. The azepanyl derivatives presented in average the highest surface anaesthesia. This fact is in agreement with the lowest polarity of azepanyl substituent.