Abstract
Simplified mechanistic models of gene regulation are fundamental to systems biology and essential for synthetic biology. However, conventional simplified models typically have outputs that are not directly measurable and are based on assumptions that do not often hold under experimental conditions. To resolve these issues, we propose a 'model reduction' methodology and simplified kinetic models of total mRNA and total protein concentration, which link measurements, models and biochemical mechanisms. The proposed approach is based on assumptions that hold generally and include typical cases in systems and synthetic biology where conventional models do not hold. We use novel assumptions regarding the 'speed of reactions', which are required for the methodology to be consistent with experimental data. We also apply the methodology to propose simplified models of gene regulation in the presence of multiple protein binding sites, providing both biological insights and an illustration of the generality of the methodology. Lastly, we show that modelling total protein concentration allows us to address key questions on gene regulation, such as efficiency, burden, competition and modularity.
Introduction: The Gap in Conventional Modeling
In synthetic and systems biology, researchers frequently rely on simplified models, such as Hill functions, to describe the dynamics of gene regulatory networks. While mathematically convenient, these models suffer from two significant limitations. First, they often depend on variables that are impossible to measure directly, such as the concentration of a "free monomeric transcription factor." In practice, laboratory equipment typically measures total protein concentration (e.g., via fluorescent tags). Second, traditional simplifications assume that molecular binding occurs much faster than transcription or translation. In reality, these "fast" reactions can be slower than mRNA degradation, leading to substantial modeling errors and poor predictive power for complex synthetic circuits.
Methodology: Modeling Total Concentrations
To bridge this gap, this research introduces a novel model reduction methodology that shifts the focus to total mRNA ($m_T$) and total protein ($x_T$) concentrations. Instead of assuming binding is infinitely fast, the framework assumes that the lifetimes of proteins and mRNA provide the natural time scale for the system's dynamics.
Key technical components of this approach include:
- Total Concentration Variables: Grounding the system in measurable outputs.
- Multimerization Efficiency ($\eta_m$): A novel term that approximates the fraction of protein in its active multimeric form without requiring complex implicit equations.
- Case-Based Regulation: Splitting models into "multimer-dominant" and "monomer-dominant" regimes based on a biological parameter ($\epsilon$), allowing for simpler explicit equations.
Addressing Sequestering and Competition
A major advantage of this methodology is its ability to account for "sequestering" logic—where "decoy" binding sites compete for transcription factors. This is a common occurrence in real cells that traditional models often ignore or oversimplify.
By modeling total protein concentration, the researchers naturally resolve issues of "retroactivity," where downstream modules affect upstream ones. If degradation is uniform, adding a downstream binding site does not change the upstream total protein dynamics, providing a more modular framework for synthetic design.
Key Results and Validation
The proposed reduced models were validated against "gold standard" full mechanistic simulations using well-known synthetic benchmarks:
- Superior Accuracy: In comparisons of protein expression levels, the reduced model showed a near-perfect match with full mechanistic simulations, whereas standard Hill functions introduced significant quantitative errors.
- Qualitative Success in Oscillators: For the Repressilator, traditional Hill functions often incorrectly predicted stable behavior or the wrong frequency. The reduced model accurately captured the oscillations of the full mechanistic model.
- Bistability Prediction: In a Toggle Switch with additional binding sites, the reduced model correctly predicted that sequestering sites could induce bistability, a feature missed by simpler models.
Conclusion
By redefining the variables of interest to align with experimental measurability and grounding assumptions in realistic reaction speeds, this methodology provides a robust framework for both analyzing natural systems and designing synthetic ones. It addresses critical questions regarding cellular burden, competition, and modularity, paving the way for more predictable and efficient biological engineering.
Summary updated