Artificial neural network hybridized with a genetic algorithm for optimization of lipase production from Penicillium roqueforti ATCC 10110 in solid-state fermentation

In the present work, an artificial neural network hybridized with a genetic algorithm (ANN-GA) has been applied to optimize Penicillium roqueforti ATCC 10110 lipase production in solid-state fermentation (SSF). For such a purpose, a feed-forward ANN with polynomial configuration 3-49-1 (i.e. 3 neurons in the input layer, 49 neurons in the hidden layer and 1 neuron in the output layer) was used to computationally model the experiment and a GA was used to optimize lipase production through the ANN model. The input variables optimized by the ANN-GA were fermentation time (1 day), incubation temperature (31.2 degrees C) and percentage moisture content (78%). Validation was performed by considering the optimal and central point conditions, thus obtaining a lipase activity value of 48.00 U g(-1), which is three times greater than by using other methodologies. Furthermore, the ANN model was obtained using 28 essays (small dataset) with interpolation and generalization capability based on a significant and precise data choice and justified by mean square error and determination coefficient values. A total of 5.0 x 10(7) artificial tests were simulated from the small dataset of 28 experiments.

Automated summary

A hybrid approach of artificial neural network with genetic algorithm (ANN-GA) was applied to optimize lipase production of Penicillium roqueforti ATCC 10110 in solid-state fermentation. The optimized input variables were fermentation time, incubation temperature, and moisture content, resulting in a lipase activity value of 48.00 U g(-1), three times greater than other methodologies. The ANN model was obtained from a small dataset of 28 experiments with interpolation and generalization capability.