Optimized phenol degradation and lipid production by Rhodosporidium toruloides using response surface methodology and genetic algorithm-optimized artificial neural network

Abstract

Oleaginous yeast can produce lipids while degrading phenol in wastewater treatment. In this study, a Plackett-Burman Design (PBD) was adopted to identify key factors of phenol degradation and lipid production using R toruloides 9564T. While temperature, inoculum size, and agitation were significant for both the processes (p < 0.05), pH and incubation were significant for lipid production, and phenol removal, respectively. Results from four factors (pH, temperature, inoculum size, and incubation period) central composite design (CCD) experiment were used to formulate quadratic and genetic algorithm-optimized ANN models. The reduced quadratic model for phenol degradation (R2: 0.993) and lipid production (R2: 0.958) were marginally inferior to ANN models (R2: 0.999, 0.982, respectively) on training sets. Multi-objective optimization with equal importance suggests phenol degradation between 106.4 and 108.76%, and lipid production of 0.864–0.903 g/L, by polynomial and ANN models. Complete phenol degradation (100%) and 3.35-fold increment (0.918 g/L) in lipid production were obtained at pH 6.07, inoculum size 14.68% v/v, at 29.5 °C in 92.17 h experimentally.