Machine learning approach for the effectual production of a novel esterase and its application in bioremediation of dairy effluent

Abstract

The present work intends to maximize the production of an industrially important novel esterase from Trichoderma harzianum using both machine learning and statistical approaches and its application in the bioremediation of dairy effluent. Significant medium compounds, such as almond oil, casein, NaH2PO4, and trace elements, were screened from a pool of carbon, nitrogen, oil and surfactant sources along with other micronutrients using a first order statistical design. The optimum value of screened parameters was found using response surface methodology that resulted in the maximum esterase production of 2985.31U/L. Artificial neural network trained with the experimental response of RSM experiments coupled with genetic algorithm optimization technique has resulted in the maximum esterase production of 3256.73U/L, which is 150-folds higher as compared to the unoptimized condition and also found to be the higher than the outcome of statistical optimization. Dairy effluent was used for the production of esterase as well as for its bioremediation. At 50% load, the maximum reduction in chemical oxygen demand and the production of esterase were observed to be 46.84% and 34.12U/L, respectively. The COD reduction of diary effluent was modelled using Michaelis-Menten kinetics and the kinetic parameters, half saturation constant (Km) and maximum reaction rate (rmax) value were evaluated as 286.88mg/L and 26.04mg/L/d, respectively.