Bayesian Optimization for Design of Multiscale Biological Circuits.

Area of Science:

• Synthetic Biology
• Computational Biology
• Biotechnology

Background:

• Synthetic biology enables complex molecular circuit construction across cellular scales.
• Current computational optimization methods struggle with multiscale biological systems due to simulation stiffness.
• Efficient design tools are needed for complex gene regulation, signaling, and metabolic pathways.

Purpose of the Study:

• To develop an efficient machine learning method for optimizing synthetic biological circuits across multiple scales.
• To address the limitations of current computational methods in handling multiscale biological systems.
• To enable joint optimization of circuit architecture and parameters for improved biological system design.

Main Methods:

• Utilized Bayesian optimization, a machine learning technique for optimizing deep neural networks.
• Applied the method to gene circuits controlling biosynthetic pathways with nonlinearities and multiple scales.
• Developed a strategy for navigating nonconvex optimization problems in mixed-integer input spaces.

Main Results:

• Demonstrated efficient optimization of biological circuits across multiple temporal and concentration scales.
• Successfully handled large, multiscale biological system design problems.
• Enabled parametric sweeps for assessing circuit robustness to perturbations via in silico screening.

Conclusions:

• The developed machine learning method offers a feasible approach for designing complex synthetic biology circuits.
• This strategy efficiently optimizes multiscale biological systems, overcoming limitations of traditional simulation methods.
• The approach serves as an effective in silico screening tool, accelerating experimental implementation of synthetic biology designs.