Optimisation of pesticide crystal protein production from Bacillus thuringiensis employing artificial intelligence techniques
Mixtures containing spores of the bacterium Bacillus thuringiensis and its δ-endotoxins – referred to as pesticide crystal protein (PCP) – are very well known to be effective against several insects and pests. In this work, three factors, namely medium pH, inoculum size and sugar concentration from molasses, that were found to be highly significant for the production of PCP were optimised of their levels using a combination of artificial intelligence techniques – artificial neural network (ANN) and genetic algorithm (GA). Earlier results, expressed in terms of the culture optical density at 600 nm wavelength (OD600), were first modelled by ANN based on back propagation algorithm, which was highly accurate in predicting the system with coefficient of determination (R²) value greater than 0.99 in both training and validation of the network. Optimum values of 3.65 for pH, 6.009% for inoculum size and 1.61 g/L for sugar concentration were obtained using GA based on the developed ANN model. At these optimised settings of the factors, a predicted maximum (OD600) value of 0.5764 was achieved, which was 9.17% more than the previously obtained maximum experimental value.
Keywords: artificial neural networks, ANNs, genetic algorithms, GAs, pesticide crystal protein, PCP, Bacillus thuringiensis, optimisation, back propagation, artificial intelligence, medium pH, inoculum size, sugar concentration, parasporal crystals