Abstract
Recent advances in synthetic biology have enabled the construction of molecular circuits that operate across multiple scales of cellular organization, for example by interfacing gene regulation with signalling or metabolic pathways. Computational methods can effectively aid and accelerate the design process, but current methods are generally unsuited for systems with multiple temporal or concentration scales, as these are challenging to simulate due to their numerical stiffness. Here, we present a machine learning method for the efficient optimization of biological circuits across scales. We employ a Bayesian optimization approach and nonparametric statistical models to learn the shape of a performance landscape and iteratively navigate the design space towards an optimal design. This strategy allows the joint optimization of both circuit architecture and parameters, and hence provides a feasible approach to solve a highly non-convex optimization problem in a mixed-integer input space. We illustrate the applicability of the method on gene circuits designed to control biosynthetic pathways, as these display strong nonlinearities and have molecular components that evolve in different timescales and different scales of molecular concentrations. We test the method on various models of dynamic production pathways previously built in the literature, and highlight its ability to optimize large multiscale models with more than 20 species and circuit architectures, as well as large parametric sweeps that are useful for assessing the robustness of optimal designs to perturbations. The method can serve as an efficient in silico screening method for circuit architectures prior to experimental testing.
Competing Interest Statement
The authors have declared no competing interest.