Genetic algorithm-optimized artificial neural network for multi-objective optimization of biomass and exopolysaccharide production by Haloferax mediterranei.

Abstract

Microbial production of industrially important exopolysaccharide (EPS) from extremophiles has several advantages. In this study, key media components (i.e., sucrose, yeast extract, and urea) were optimized for biomass growth and extracellular EPS production in Haloferax mediterranei DSM 1411 using Box-Behnken design. In a multi-objective optimization framework, response surface methodology (RSM) and genetic algorithm (GA)-optimized artificial neural network (ANN) were used to minimize biomass growth while increasing EPS production. The performance of the selected ANN model for the prediction of biomass and EPS (R²: 0.964 and 0.975, respectively) was found to be better than that of the multiple regression model (R²: 0.818, 0.963, respectively). The main effect of sucrose and its interaction with urea appears to have a significant effect on both responses. The ANN model projects an increase in EPS production from 4.49 to 18.2 g l^-1 while shifting the priority from biomass to biopolymer. The optimized condition predicted a maximum biomass and EPS production of 17.27 g l^-1 and 17.80 g l^-1, respectively, at concentrations of sucrose (19.98 g l^-1), yeast extract (1.97 g l^-1), and urea (1.99 g l^-1). Based on multi-objective optimization, the GA-ANN model predicted an increase in the EPS to biomass ratio for increasing the EPS and associated biomass production. The extracted EPS, identified as Gellan gum through NMR spectroscopy, was further characterized for surface and elemental composition using SEM-EDX analysis.