(642b) Modeling and Optimization of Cholesterol Oxidase Production By Streptomyces Olivaceus Mtcc 6820 Using Response Surface Methodology Coupled with Artificial Neural Network-Genetic Algorithm

Cholesterol oxidase (EC 1.1.3.6) is a bifunctional FAD-dependent oxidoreductase enzyme, which catalyzes the oxidation of cholesterol to an intermediate 5-cholesten-3-one, with the concurrent isomerization of 5-cholesten-3-one to form 4-cholesten-3-one. The wide ranging applications of cholesterol oxidase in various fields has increased its demand in recent years. In clinical chemistry, it is used for the detection and quantification of cholesterol in blood serum and other clinical samples, while in pharmaceutical industry, cholesterol oxidase is used for the bioconversion of cholesterol to 4-cholesten-3-one, 4-andrastene-dione etc., other pharmaceutical steroids and hormones etc. In food industry, cholesterol oxidase is used for the detection of cholesterol in food samples, whereas it also has an agricultural importance due to its insecticidal and larvicidal properties. Despite of its increasing demand, on one hand, production and availability of cholesterol oxidase is restricted only to the microbial fermentation process and no other sources (animal or plant) has been reported till date. While on the other hand, the production of cholesterol oxidase by several microorganisms exhibit inducible expression pattern, some are intracellular producers while some other microbes are pathogenic in nature, which not only limit its production but also makes it an expensive enzyme for industrial as well as clinical applications. Streptomyces sp. has been reported to produce high levels of extracellular cholesterol oxidase. Over-production of cholesterol oxidase by applying various approaches has been the matter of current interest amongst the researchers worldwide.

The production of metabolites through microbial strains are largely affected by the process parameters. Fermentation processes are multivariable and optimization of bioprocesses play effective role in enhancing production of metabolites, though a cumbersome task. In the present work, the limitations of conventional one factor at a time (OFAT) method of optimization was overcome by the use of statistical models and mathematical designs in order to reduce the number of experiments, to increase the precision of results and to reach the true optimum by studying the complex interactions among the variables. This was achieved with the help of a combined Response Surface Methodology-Artificial Neural Network-Genetic Algorithm (RSM-ANN-GA) approach for modeling and optimization of microbial production of cholesterol oxidase. RSM is based on design of experiments (DOE) comprising a combination of mathematical and statistical techniques, generally used for the development of models on multivariable systems, estimation of model coefficients and prediction of response for optimum conditions. ANNs are complex mathematical models that successfully mimic biological neural networks and are used to optimize and model highly nonlinear and complex biological processes. Mathematical model generated by RSM or ANNs can be optimized more precisely by using mathematical tools like GA.

The feasibility of statistical versus artificial intelligence techniques such as RSM, ANN and GA have been tested to optimize the culture conditions for production of cholesterol oxidase. A mathematical model was developed for the production of cholesterol oxidase by Streptomyces olivaceus MTCC 6820 using RSM and optimization of culture parameters was done by applying ANN coupled with GA. Based on the predicted cholesterol oxidase concentration, the ANN model was found to be superior to the model developed with RSM, and ANN was found to be the better predictor than RSM. The maximum cholesterol oxidase production obtained by combined RSM-ANN-GA approach was 4.2 U/ml which was nearly 2.1 times higher than that of the unoptimized culture conditions. Both the RSM and ANN models were compared in terms of coefficient of determination R² (99.99_ANN>97.09_RSM), predicted distribution coefficient (0.9573_ANN>0.8986_RSM) and absolute average deviation AAD (2.1156%_ANN<2.228%_RSM). Moreover, development of simple kinetic models were attempted using Logistic equation for cell growth and Luedekig Piret equation for cholesterol oxidase production.