Artificial intelligence-driven dynamic regulation for high-efficiency gentamicin C1a production

Abstract

The industrial biosynthesis of complex antibiotics such as gentamicin C1a requires precise, dynamic control of microbial metabolism. In this study, an artificial intelligence (AI)-driven control framework was developed that integrated data-driven decision-making with real-time sensing to optimize the green production of gentamicin C1a. The system consists of four tightly coupled modules encompassing backpropagation neural network (BPNN)-based kinetic modeling, multi-objective optimization (NSGA-II), dual-spectroscopy monitoring (near-infrared and Raman), and closed-loop feedback control. The BPNN model accurately captured nonlinear correlations between specific substrate consumption rates, specific growth rates, and specific gentamicin C1a production rates, with R² values of 0.9631, 0.9578, and 0.9689, respectively. By resolving phase-specific trade-offs in metabolic demands, the AI-driven dynamic regulation enabled real-time coordination between carbon, nitrogen, and oxygen supplementation as well as cellular requirements. Through the implementation of an AI-driven dynamic regulation system, gentamicin C1a production was significantly enhanced, achieving a titer of 430.5 mg L⁻¹, a 75.7% improvement over traditional fed-batch fermentation. Notably, the gentamicin C1a yield reached 10.3 mg g⁻¹ and the specific productivity attained 0.079 mg gDCW⁻¹ h⁻¹, both of which represented the highest levels reported to date. During the late fermentation phase, integrated metabolomics and metabolic flux analyses revealed a dynamic reorganization of the metabolic network, characterized by increased flux through the pentose phosphate pathway, enhanced NADPH generation and consumption, and improved carbon–nitrogen allocation favoring gentamicin C1a biosynthesis. Finally, the integrated techno-economic analysis and life cycle assessment confirm the commercial feasibility and greenhouse gas mitigation potential of gentamicin C1a production via AI-enhanced fermentation. These results demonstrate the power of AI-enhanced bioprocesses for intelligent fermentation control and establish a scalable, mechanistically informed strategy for the green industrial production of secondary metabolites.