Optimization for isoflavone production in Maackia amurensis suspension cells based on neural networks and accelerating genetic algorithm / 生物工程学报

ABSTRACT

The medium for isoflavone production in Maackia amurensis suspension cells has been optiwised through the artificial neural networks (ANNs) and the real coding based accelerating genetic algorithm (RAGA). Among the ingredients of the medium, nitrogen sources and plant growth regulators were found to be the main factors affecting the production of isoflavone genistein. (NH4)2SO4, KNO3, 2,4-D and 6-BA, 100 approximately 800 mg/L, 1500 approximately 3000 mg/L, 0 approximately 3 mg/L and 0 approximately 1 mg/L respectively, significantly increased genistein yield, in the ranges of effective concentrations. The random ten combinations of these four components generated by RAGA as input data and the genistein yields of ten combinations as output data were used for ANNs-RAGA (the artificial neural networks associated with the accelerating genetic algorithm) modeling. The resultant model showed a high fit between the experimental data and calculating values by ANNs-RAGA. Based on the prediction of the model, the optimum combination of four factors for genistein production was determined on 149.68 mg/L for (NH4)2SO4, 2936.10 mg/L KNO3, 0.01 mg/L 2,4-D and 0.19 mg/L 6-BA. When cells were cultured in the optimized medium, their capability of genistein production was remarkably enhanced to 14.13 mg/L, which was about 19 times higher than that in the original medium. The relative discrepancy between the experimental value and the predictive value of genistein yield from the optimized medium was 7.38%.