Hybrid models of the cell cycle molecular machinery

Piecewise smooth hybrid systems, involving continuous and discrete variables, are suitable models for describing the multiscale regulatory machinery of the biological cells. In hybrid models, the discrete variables can switch on and off some molecular interactions, simulating cell progression through a series of functioning modes. The advancement through the cell cycle is the archetype of such an organized sequence of events. We present an approach, inspired from tropical geometry ideas, allowing to reduce, hybridize and analyse cell cycle models consisting of polynomial or rational ordinary differential equations.

💡 Research Summary

The paper proposes a systematic framework for constructing hybrid models of the cell‑cycle regulatory machinery by exploiting ideas from tropical geometry. Hybrid systems combine continuous dynamics (e.g., concentrations of biochemical species) with discrete switches that turn molecular interactions on or off. This combination is particularly suitable for biological processes that involve multiple time and spatial scales, such as the progression through the cell‑cycle, which consists of a well‑ordered sequence of phases separated by checkpoint transitions.

The authors first review the general concept of piecewise‑smooth hybrid systems, distinguishing several subclasses (switched systems, multivalued differential automata, etc.) and noting that stochastic extensions are also possible. They argue that hybrid modeling offers a compromise between biological realism and analytical tractability, enabling reachability analysis, temporal‑logic verification, and model reduction.

The core methodological contribution is a two‑step tropicalization procedure that converts a system of polynomial or rational ordinary differential equations (ODEs) into a piecewise‑smooth hybrid system. The ODEs are assumed to have the generic form

 dx_i/dt = F_i(x) = P_i(x)/Q_i(x),

where P_i and Q_i are multivariate polynomials (or Laurent polynomials with positive coefficients). In many biochemical networks, the monomials of these polynomials differ by orders of magnitude because of widely separated kinetic constants. The tropical approach replaces each polynomial by its dominant monomial, i.e., the term with the largest absolute value for a given state vector. Two variants are defined:

1. Complete tropicalization – every polynomial (both numerator and denominator) is replaced by its dominant monomial, yielding the system

 dx_i/dt = Dom P_i(x) / Dom Q_i(x).

2. Two‑term tropicalization – applicable when the ODEs are expressed as the difference of two positive Laurent polynomials (production minus degradation). Each of the two sums is reduced to its dominant term, giving

 dx_i/dt = Dom P⁺_i(x) – Dom P⁻_i(x).

The set of points where two or more monomials attain the same maximal value defines a tropical manifold (or tropical variety). These manifolds partition the state space into sectors; within each sector a single monomial dominates, and the dynamics are smooth. Crossing a tropical manifold corresponds to a discrete switch, thus providing a natural hybrid representation.

To justify the approximation, the authors introduce the ecological notion of permanency: a system is permanent if all its variables remain bounded away from zero and infinity after some finite transient, uniformly with respect to a small parameter ε that scales the kinetic constants. Under permanency, the difference between the original ODE solution and the tropicalized solution can be bounded by

 |y(t)| ≤ C₁ ε^γ exp(b t),

where C₁, γ, b are constants independent of ε. This result follows from Grönwall’s lemma. Moreover, if the original system is structurally stable on a compact domain, the tropical system inherits permanency and there exists an ε‑dependent homeomorphism h_ε close to the identity that maps trajectories of one system to the other.

A key condition for permanency is the tropical equilibration: each ODE must contain at least two dominant monomials of opposite signs (one production, one degradation). This mirrors the classical Newton polytope condition and ensures that the dominant balance can sustain a bounded steady state. The authors provide an algorithmic scheme: select, for each equation, a pair of opposite‑sign monomials, solve a linear system for the scaling exponents a_i, and verify a set of linear inequalities. The combinatorial explosion of possible pairs is acknowledged, but the method yields explicit scaling exponents and a reduced “tropically truncated” system.

The dynamical analysis of the truncated system reveals a hierarchy of time scales μ₁ < μ₂ < … < μ_n. The fastest variable (μ₁ = 0) evolves on the slow manifold defined by the remaining variables, leading to a chain‑like relaxation toward a global attractor. Under generic conditions (μ_i distinct), the system either converges to a globally attracting equilibrium or, if the last two equations admit a hyperbolic limit cycle, to a globally attracting periodic orbit. Theorem 2.7 formalizes this result, showing that the original system inherits the same qualitative attractor for sufficiently small ε.

The paper then applies the framework to a well‑known cell‑cycle model by Tyson. This model captures the interaction between cyclin, Cdc2, and the maturation‑promoting factor (MPF) complex, exhibiting three regimes: oscillatory (embryonic rapid divisions), steady high‑MPF (metaphase arrest), and excitable switch (somatic growth). By tropicalizing the Tyson ODEs, the authors identify the dominant monomials, construct the corresponding hybrid system, and verify tropical equilibration and permanency at each stage. The resulting hybrid model reproduces the original dynamics (oscillation, steady state, switch) while drastically reducing the number of active terms and simplifying the phase‑space geometry. Moreover, the tropical manifolds correspond to biologically meaningful thresholds (e.g., checkpoint activation), providing an intuitive interpretation of the discrete switches.

In conclusion, the authors demonstrate that tropical geometry offers a principled way to derive hybrid models from detailed biochemical ODEs. The approach yields mathematically rigorous error bounds, clarifies the role of dominant reaction pathways, and enables model reduction via invariant manifolds. It is particularly well‑suited for cell‑cycle networks, where multiple time scales and checkpoint‑driven mode changes are intrinsic. Future directions suggested include extensions to stochastic hybrid systems, incorporation of spatial diffusion, and data‑driven parameter identification within the tropical‑hybrid framework.