Maximizing mRNA vaccine production with Bayesian optimization.

Area of Science:

• Biotechnology
• Vaccine Development
• Process Optimization

Background:

• Messenger RNA (mRNA) vaccines represent a novel vaccine platform crucial for infectious disease control.
• mRNA vaccines are manufactured via in vitro transcription (IVT) reactions, requiring optimization for efficient, large-scale production.
• Traditional optimization methods for IVT are time-consuming and prone to bias.

Purpose of the Study:

• To apply Machine Learning (ML) for data-driven optimization of mRNA IVT reactions.
• To automate experiment design and establish a feedback loop for efficient process optimization.
• To improve both the yield and speed of mRNA IVT processes.

Main Methods:

• Utilized Bayesian optimization and model interpretability techniques for automated experiment design.
• Implemented a data-driven feedback loop to guide the optimization process.
• Conducted optimization runs to identify ideal IVT reaction conditions.

Main Results:

• Achieved a yield of 12 g/L in just 2 hours under optimized conditions.
• Outperformed published industry standards in both reaction yield and production time.
• Demonstrated the effectiveness of ML in significantly enhancing mRNA IVT efficiency.

Conclusions:

• Bayesian optimization offers a powerful, cost-effective tool for optimizing (bio)chemical processes like mRNA IVT.
• The developed ML approach significantly accelerates mRNA vaccine manufacturing.
• This methodology holds broad potential for advancing biopharmaceutical production.