Inteligência artificial aplicada a simulação cinética de processos químicos e metabólicos

Abstract

The kinetic simulation of chemical and metabolic processes can contribute for the understanding of several reactions that are present in all chemical and biochemical processes and hence, one can propose the corresponding mechanisms of such systems. All these processes can contribute, in principle, in determining modifications due to different compounds that also have biological activity, for instance, drugs or toxic substances. Accordingly, the present thesis will deal with the development and applications of computational artificial intelligence (AI) techniques such as genetic algorithms (GA) and fuzzy logic (FL) both coupled to Petri Nets. The proposed approach will be tested to study chemical and metabolic processes. In order to analyze this AI methodology for studying metabolic systems two different applications will be used. The first case corresponds to the determination of Arrhenius parameters (activation energy and pre exponential factor) for the processof semiconductors systems. Thermogravimetric experimental data will be used as the input data and TiO2-SnO2 system was used. The results demonstrated an efficient approach to determine the kinetic behavior of the whole chemical process and this produces a procedure to obtain the Arrhenius parameters as a function of temperature. The second application corresponds to the determination and parametrization of metabolic routes using as the input data either experimental or modeling data of species concentrations as a function of time. As demonstrated the Petri Nets are directly correlated to the time evolution of all simultaneous reactions of a specified system. In the present case two biological systems were used namely, 1,2-diacilglycerol, in which there are 4 coupled reactions, and another one more complex, the citric acid cycle, that there are 21 coupled reactions. The results in this particular case (constant temperature) were: all routes of both biochemistry processes, the final product concentrations of all species and the corresponding kinetic constants. The relative average error of both cases is of the order of 1%.