Quasi-robust control of biochemical reaction networks via stochastic morphing

Tomislav Plesa, Guy‐Bart Stan, Thomas E. Ouldridge, Wooli BaeJournal of The Royal Society Interface 2021 · DOI

TL;DR

This paper introduces "stochastic morphing," a novel control strategy for biochemical reaction networks that goes beyond simply controlling the average behavior (mean) or variability (variance) of molecules. Instead, it allows for precise manipulation of the entire probability distribution of molecular species, enabling the creation of complex multi-modal behaviors and arbitrary distributions, with potential for robust experimental implementation.

What Problem Does This Paper Address?

Synthetic biology aims to engineer biological systems with predictable behaviors. A key challenge is controlling biochemical reaction networks, especially within the noisy environment of small compartments like cells. Existing controllers primarily focus on manipulating the average concentration (mean) or reducing fluctuations (variance) of molecules. However, many crucial biological processes, like cellular differentiation or antibiotic resistance, rely on higher-order statistical properties of molecular distributions, such as the number and positions of peaks (modes). This paper addresses the gap by developing a controller that can "morph" the probability distribution of a target molecule into a desired shape, including complex multi-modal distributions, even when the underlying network is partially unknown.

Methods

The core innovation is the "stochastic morpher" controller. It operates by introducing a faster "interface" sub-network that overrides the input network, coupled with a slower "core" sub-network. This architecture, leveraging principles of noise-induced mixing and timescale separation, allows for gradual transformation of the input probability distribution into a target one.

Two main versions are presented:

  • Lower-resolution morpher: Achieves multi-modal distributions by using simple production and degradation reactions for the target species. This allows for control over the number and positions of modes, as well as the timing of transitions between them.
  • Higher-resolution morpher: Employs more complex reactions, including catalytic ones, to achieve arbitrary probability distributions, specifically demonstrating the ability to generate Kronecker-delta distributions (highly localized peaks).

The authors rigorously establish the theoretical underpinnings of these controllers, proving their convergence and quasi-robustness (accuracy despite unknown initial conditions or rate coefficients). They also propose a blueprint for experimental implementation using DNA strand-displacement nanotechnology.

Datasets

Not discussed in the paper. The paper focuses on theoretical modeling and simulation of biochemical reaction networks.

Key Insights and Ideas

  • Beyond Mean and Variance: The most significant insight is the shift from controlling only the first two moments (mean and variance) to controlling the entire probability distribution. This unlocks the ability to engineer complex biological functions that depend on higher-order statistics.
  • Stochasticity as a Tool: Instead of solely fighting noise, the stochastic morpher leverages intrinsic molecular noise as a mechanism to achieve desired distributions, particularly for multi-modality.
  • Timescale Separation is Key: The controller's effectiveness relies on creating a significant difference in timescales between the interface and core sub-networks, and between the controller and the input network.
  • Quasi-Robustness: The controller is designed to be quasi-robust, meaning it can achieve the desired dynamics even with incomplete knowledge of the input network's parameters and initial conditions, a crucial feature for biological applications.
  • Experimental Feasibility: The proposed DNA nanotechnology implementation demonstrates a concrete path towards realizing these sophisticated controllers in synthetic cells.

Key Results

  • Multi-modal Control: The lower-resolution morpher successfully transformed uni-modal input distributions into bi-modal and even tri-modal distributions, with control over the positions and relative heights of the modes.
  • Arbitrary Distribution Generation: The higher-resolution morpher demonstrated the ability to generate Kronecker-delta distributions, representing highly precise molecular localization.
  • Implicit Control: The paper showed that explicit control over one species (e.g., mRNA) can implicitly influence the distribution of another species (e.g., protein) without direct interaction, enabling indirect control strategies.
  • Convergence and Robustness: Theoretical analysis confirmed that the stochastic morpher controllers converge exponentially fast to the target distributions with an error that decreases linearly with the controller's speed, and are robust to initial conditions.

Limitations and Future Work

  • Timescale Separation Requirements: While quasi-robustness is achieved, strong control relies on significant timescale separations, which might be challenging to engineer perfectly in all experimental scenarios. The paper notes that even with weaker separations, partial control can be achieved.
  • Experimental Realization Complexity: While a blueprint is provided, the practical implementation of the DNA-based controller, particularly eliminating unwanted reactions and ensuring precise control over rates, will require significant engineering effort.
  • Spatial Heterogeneity: The current models assume well-mixed environments. Future work could explore how spatial heterogeneity within cells affects controller performance and how to account for it.
  • Broader Applicability: While demonstrated on specific network types, extending the stochastic morpher to more complex and diverse biological systems, including those with feedback loops and non-linear dynamics, remains an important avenue. The authors also suggest exploring its application to gene expression networks and other biological phenomena.

Summary updated

Abstract

One of the main objectives of synthetic biology is the development of molecular controllers that can manipulate the dynamics of a given biochemical network that is at most partially known. When integrated into smaller compartments, such as living or synthetic cells, controllers have to be calibrated to factor in the intrinsic noise. In this context, biochemical controllers put forward in the literature have focused on manipulating the mean (first moment) and reducing the variance (second moment) of the target molecular species. However, many critical biochemical processes are realized via higher-order moments, particularly the number and configuration of the probability distribution modes (maxima). To bridge the gap, we put forward the <i>stochastic morpher</i> controller that can, under suitable timescale separations, morph the probability distribution of the target molecular species into a predefined form. The morphing can be performed at a lower-resolution, allowing one to achieve desired multi-modality/multi-stability, and at a higher-resolution, allowing one to achieve arbitrary probability distributions. Properties of the controller, such as robustness and convergence, are rigorously established, and demonstrated on various examples. Also proposed is a blueprint for an experimental implementation of stochastic morpher.
BibTeX
@article{plesa2021quasi,
  author = {Tomislav Plesa and Guy‐Bart Stan and Thomas E. Ouldridge and Wooli Bae},
  title = {{Quasi-robust control of biochemical reaction networks via stochastic morphing}},
  journal = {Journal of The Royal Society Interface},
  year = {2021},
  doi = {10.1098/rsif.2020.0985},
}
Figure