Risk-averse optimization of genetic circuits under uncertainty.

Area of Science:

• Synthetic biology
• Computational biology
• Bioengineering

Background:

• Engineering biological systems requires navigating complex design spaces, often limited by wet-lab experimentation.
• Mathematical modeling and computational optimization accelerate design but suffer from inherent model inaccuracies, leading to suboptimal in vivo performance.

Purpose of the Study:

• To develop a robust computational framework for designing functional biological circuits by addressing model uncertainty.
• To improve the prediction accuracy and in vivo performance of engineered biological systems.

Main Methods:

• Utilized Bayesian inference to estimate the distribution of model parameters from non-functional designs.
• Employed Thompson sampling and risk-averse optimization to select robust design parameters.
• Validated the approach on adaptation circuits and genetic oscillators using diverse model complexities and data types.

Main Results:

• The proposed method effectively estimates parameter distributions and identifies optimal, risk-averse designs.
• Demonstrated successful application in designing both adaptation circuits and genetic oscillators.
• Showcased the approach's versatility across various model complexities and data sources.

Conclusions:

• The integration of Bayesian inference, Thompson sampling, and risk management offers a powerful strategy for de-risking biological circuit design.
• This computational approach enhances the reliability and efficiency of engineering functional biological systems.
• The method provides a pathway to more predictable and successful synthetic biology applications.