Leveraging knowledge engineering and machine learning for microbial bio-manufacturing.

Area of Science:

• Synthetic biology
• Metabolic engineering
• Computational biology

Background:

• Genome scale modeling (GSM) is crucial for predicting microbial performance and identifying gene targets.
• Current GSMs struggle with uncertainties from pathway regulations, metabolic burdens, and bioreactor variations, limiting computational strain design (CSD).

Purpose of the Study:

• To explore the integration of machine learning (ML) with GSM for enhanced strain development.
• To highlight how data-driven frameworks and knowledge engineering can improve CSD and cellular process understanding.

Main Methods:

• Leveraging knowledge engineering to standardize information from literature (genomics, phenomics, synthetic biology, bioprocesses).
• Utilizing data-driven frameworks to provide constraints for mechanistic models.
• Applying machine learning techniques, including deep learning and transfer learning.

Main Results:

• Data-driven frameworks can offer new constraints for mechanistic models to describe cellular regulations.
• ML can aid in pathway design, gene target identification, and prediction of fermentation metrics (titer, rate, yield).
• Integration facilitates a "Learn and Design" cycle for microbial strain improvement.

Conclusions:

• Machine learning presents a viable complementary approach to GSM for advancing strain design and deciphering cellular processes.
• Effective information collection and database construction are key to developing robust ML models for biotechnology.
• This integrated approach promises to enhance the efficiency and predictability of microbial strain development.