Iterative design of a NAND hybrid riboswitch by deep batch Bayesian optimization.

Area of Science:

• Synthetic Biology
• Genetic Engineering
• Computational Biology

Background:

• Designing complex genetic circuits requires efficient regulatory devices.
• Hybrid riboswitches are synthetic RNA elements (<100 nucleotides) that can perform Boolean logic operations.
• These devices regulate gene expression by binding specific molecules without auxiliary factors.

Purpose of the Study:

• To design hybrid riboswitches that emulate Boolean NAND logic in yeast.
• To develop a machine learning-based framework for optimizing genetic regulatory devices.
• To achieve precise gene regulation with minimal metabolic burden on host cells.

Main Methods:

• A novel machine learning framework combining high-throughput in vivo screening and deep Bayesian optimization was proposed.
• Initial screening identified a hybrid riboswitch exhibiting NAND logic behavior.
• Batch Bayesian optimization with an ensemble neural network surrogate was used to enhance NAND functionality.

Main Results:

• A hybrid riboswitch with initial NAND logic behavior was discovered.
• The NAND functionality was significantly improved using Bayesian optimization, achieving near-digital performance.
• The developed framework allows fine-grained adaptation of genetic construct functionality.

Conclusions:

• The proposed machine learning framework effectively designs and optimizes hybrid riboswitches for complex logic operations.
• This approach enables the creation of sophisticated genetic regulatory devices for synthetic biology applications.
• The method complements experimental approaches by facilitating precise functional adaptation, even for single nucleotide changes.