Prodigiosin, a red pigment with diverse biotechnological applications, is produced as a secondary metabolite by Gram-negative bacilli Serratia marcescens. In this study, we implemented an AI-guided hybrid optimization framework combining Response Surface Methodology (RSM) using a Circumscribed Central Composite Design (CCCD) and Artificial Neural Network (ANN) modeling to enhance prodigiosin pigment production. Across 34 experimental runs, we optimized sucrose and peptone concentrations along with inoculum size. The RSM-derived model exhibited a strong correlation (R2 = 0.953), while the ANN, trained using a backpropagation algorithm, demonstrated superior predictive power (R2 = 0.998; MSE = 0.000414), underscoring the potential of artificial intelligence in modeling complex bioprocesses. Beyond statistical optimization, an induction strategy using 1% of various natural additives (vegetable oils and egg components) identified egg white, rich in albumin, as the most effective enhancer, tripling prodigiosin yield. Further investigation revealed that a 2% egg white concentration maximized production to 1070 mg L−1, a substantial increase compared to the optimized yield of 359.2 ± 12 mg L−1 and predicted value of 391.86 mg L−1. These results highlight the value of integrating machine learning with experimental design and protein-rich inducers to strengthen sustainable microbial pigment production in a cost-effective and scalable manner.
Breig et al. (Tue,) studied this question.