Machine learning assisted media optimization for enhanced insulin production in Pseudomonas fluorescens cell factory and scale-up studies.

Diabetes mellitus, a chronic metabolic disorder, is characterized by high blood glucose levels. External insulin administration, along with diet and exercise, is recommended to the patients. The increase in the number of diabetics and the adoption of oral insulin uptake (higher dosage) has led to an enhancement in insulin demand. Although insulin has been expressed in Pseudomonas fluorescens-based expression system, medium optimization, a crucial step in bioprocess for product titer and yield improvement, has not been reported. In this study, medium components were screened using the Placket-Burman design (PBD) approach, which showed that glucose, sodium chloride, ammonium chloride, and magnesium sulphate have a significant effect on GST-proinsulin (GPI) fusion protein expression. The Box-Behnken design (BBD) and the Artificial Neural Network-Genetic Algorithm (ANN-GA) were implemented to determine the optimized level of the screened significant medium components. The optimized medium was used to validate the result in the shake flask, and the scale-up potential was studied in the batch reactor. The kinetic parameters estimated from the batch study were used to perform the fed-batch and continuous bioreactor experiments. The fed-batch reactor led to a maximum expression of 1145 mg/L proinsulin fusion protein, a 19-fold higher titer in the optimized medium. This study establishes the P. fluorescens expression system for high titer production of recombinant proteins and the importance of machine learning-assisted optimization over traditional statistical optimization techniques.