Artificial neural network based optimization of sunflower oil supplementation in polyhydroxyalkanoates production by Cupriavidus necator

Abstract

The assessment of feeding strategy and availability of alternate carbon source in vegetable oil-supplemented microbial polyhydroxyalkanoates (PHA) production remains largely qualitative. This study aims to model the impact of sunflower oil to sucrose ratio, and feeding strategy on PHA production by Cupriavidus necator. Sets containing oil (20 ml l⁻¹), and oil-sucrose mixture (10 ml l⁻¹, 10 g l⁻¹) resulted in 3.907 g l⁻¹ and 3.342 g l⁻¹ of PHA, much higher than in sucrose-only (2.081 g l⁻¹) set. Shifting sucrose (20 g l⁻¹) pre-grown biomass on into oil (20 ml l⁻¹) containing medium resulted in a substantial increase in polymer than in sets either amended with oil or having an equal proportion sucrose-oil blend. Hyperparameter optimized double-layer (6-3) ANN model for PHA production, developed from factorial design experiment with different sucrose-oil ratios (0.053–19 w/v), inoculum dose (0.853–1.219 g l⁻¹), and time-of-shift (24–72 h), offered reasonably good fitness both on training (R²: 0.960) and cross-validation (R²_CV: 0.593). Optimization of the ANN model projected 10.411 g l⁻¹ of PHA from a sucrose-oil ratio of 1.729 (w/v), inoculum dose (0.854 g l⁻¹), and sucrose-to-oil shift after 44 h of incubation.