ИСТИНА |
Войти в систему Регистрация |
|
ИПМех РАН |
||
Commercial value of any biosensor is determined by its possibilities to analyze real samples, which as a rule are complex multicomponent mixtures. Quantitative analysis of cholinesterase inhibitors is usually restricted by a measurement of an integral response of a sensor. To reveal contributions of single components to the integral response discriminative approaches are required to be used. Several biosensor systems are presented to discriminate the groups of inhibitors. A number of esterases with different inhibitor specificity were utilized (butyrylcholinesterase/acetylcholinesterase, butyrylcholinesterase/fish cholinesterase). Organophosphate hydrolase was additionally used for selective hydrolysis of organophosphorus compounds. Low selective effectors of inhibition (pH, ionic strength, additives of organic solvents, etc.) were applied to arrange the esterase panel according to a simple principle: single enzyme in a several different conditions is equivalent to several enzymes with different inhibitor specificities. The inhibition of butyrylcholinesterase with the mixtures of organophosphates (diisopropylphosphofluoride, paraoxon) and carbamates (carbaryl, carbofuran) was examined under a number of experimental conditions. The analysis of multidimensional vector by methods of formal kinetics, artificial neural networks, and fussy logic has been done. In the frames of this analysis, the inhibitor potency of the mixtures serves as the inputs of the system while inhibitor concentrations are the outputs. The theoretical modeling of the experiment for two-, three and four-component mixtures of inhibitors with the randomly generated experimental data set has been carried out in order to select the best algorithm and the optimum size of training and test data set. The effects of the working range of inhibitor concentrations and the inputs/outputs ratio on the relative error of the determination of the concentration of inhibitors in the mixtures and the number of correct predictions were examined. The results of the theoretical modeling are in a good agreement with the data on two- and three-component mixture analysis with a real experimental data set. Similar approaches are developed for the quantitative analysis of the mixtures of enzymes (esterases) that is important for the evaluation of the "esterase state" of the human body for rapid and specific detection of human exposure to chemical agents and accurate diagnosis of chemically induced diseases.