A machine learning-based approach for improving plasmid DNA production in Escherichia coli fed-batch fermentations

Area of Science:

• Biotechnology and Fermentation Science
• Industrial Microbiology
• Artificial Intelligence in Bioprocessing

Background:

• Traditional plasmid fermentation suffers from unstable yields due to subjective process control.
• Optimizing fermentation requires understanding complex metabolic dynamics and environmental factors.

Purpose of the Study:

• To develop an intelligent, self-optimizing method for fermentation processes using AI.
• To improve plasmid yield and productivity through data-driven process control.

Main Methods:

• Multi-parameter correlation analysis to identify key factors influencing plasmid yield.
• Development of kinetic models for biomass, plasmid, and substrate concentrations.
• Machine learning (Random Forest) model to predict optimal heating strategies.

Main Results:

• Identified heating rate and timing as critical for balanced cell growth and plasmid production.
• Achieved 1167.74 mg L⁻¹ plasmid yield and 8.87 mg L⁻¹/OD₆₀₀ specific productivity using the AI-predicted strategy.
• Demonstrated a 71% increase in plasmid yield and 21% increase in specific productivity compared to traditional methods.

Conclusions:

• The AI-driven approach transforms empirical fermentation optimization into an efficient, rational self-optimization method.
• The methodology is applicable to other fermentation products, advancing intelligent automation.
• Machine learning provides a powerful tool for dynamic regulation of fermentation processes.