Empirical predictive modelling of poly-ɛ-lysine biosynthesis in resting cells of Streptomyces noursei

Abstract

Poly-ɛ-l-lysine (ɛ-PL) biosynthesis was investigated using the resting cell culture technique and nutritional parameters were optimized with response surface methodology (RSM) and an artificial neural network (ANN). ɛ-PL production in resting cell cultures of Streptomyces noursei NRRL 5126 was compared using RSM and ANN optimization techniques. The predicted ɛ-PL yield of 924.65 mg/L using ANN simulation was in better agreement with validation experimental results of 918.35±7.56 mg/L than RSM simulation results of 966.24 mg/L. The optimized medium consisted of 3% glucose, 1% ammonium sulphate, and 5 mM citric acid in both a shake flask and a 5 L bioreactor. The shake flask ɛ-PL production as 1.0 g/L and bioreactor production as 2.36 g/ L was observed. The ANN predictive model was better than the RSM predictive model during nonlinear behavior of the system.