Machine Learning of Designed Translational Control Allows Predictive Pathway Optimization in Escherichia coli

Area of Science:

• Synthetic biology
• Metabolic engineering
• Systems biology

Background:

• Microbial cell factories are crucial for producing valuable compounds.
• Optimizing gene expression is key for efficient production, with ribosome binding sites (RBSs) controlling translation initiation.
• Predictive tools for RBS design are needed for complex multigene pathways.

Purpose of the Study:

• To implement machine learning algorithms for predicting optimal RBS sequences in bacterial production systems.
• To enable accurate prediction of high-producing microbial strains through RBS library screening.
• To develop efficient screening methods for large combinatorial RBS libraries.

Main Methods:

• Machine learning models were trained on RBS sequence-phenotype data from combinatorial libraries.
• Predictive models were applied to identify optimal RBS combinations for a target pathway.
• A multiwell plate fermentation procedure was developed for high-throughput screening.
• The methodology was tested on a recombinant monoterpenoid production pathway in Escherichia coli.

Main Results:

• Machine learning accurately predicted optimal RBS sequences, identifying high-producing strains.
• Recombinant monoterpenoid production titers were increased by over 60% using the predictive approach.
• Screening efficiency was improved, requiring analysis of less than 3% of a library.
• A high-throughput screening method allowed discrimination between high and low producers.

Conclusions:

• Machine learning provides a powerful tool for the predictive design of bacterial production chassis.
• The developed methodology enables efficient optimization of gene expression in synthetic biology.
• This approach accelerates the development of microbial cell factories for various applications.
• The method is adaptable for optimizing any biochemical pathway in bacteria.