Limit cycles in piecewise-affine gene network models with multiple interaction loops

In this paper we consider piecewise affine differential equations modeling gene networks. We work with arbitrary decay rates, and under a local hypothesis expressed as an alignment condition of successive focal points. The interaction graph of the system may be rather complex (multiple intricate loops of any sign, multiple thresholds…). Our main result is an alternative theorem showing that, if a sequence of region is periodically visited by trajectories, then under our hypotheses, there exists either a unique stable periodic solution, or the origin attracts all trajectories in this sequence of regions. This result extends greatly our previous work on a single negative feedback loop. We give several examples and simulations illustrating different cases.

💡 Research Summary

The paper investigates the dynamics of piecewise‑affine (PWA) differential equations that are widely used to model gene regulatory networks. Unlike many previous studies, the authors allow each gene to have its own decay rate, thereby removing the restrictive assumption of identical degradation constants. Moreover, the interaction graph is permitted to be arbitrarily complex: multiple feedback loops of any sign, several thresholds per gene, and intricate interconnections are all admissible.

The central hypothesis is an “alignment condition” on successive focal points. In each linear region of the state space, the affine dynamics converge to a unique focal point (the equilibrium of that region). When a trajectory crosses a threshold and enters a new region, the new region’s focal point must lie on the same straight line as the focal point of the previous region. This geometric alignment guarantees that the trajectory does not experience a discontinuous jump in direction, making a periodic itinerary of regions dynamically feasible.

Under this hypothesis the authors prove an alternative theorem. Suppose a finite collection of regions is visited repeatedly by a trajectory (i.e., the trajectory follows a periodic sequence of region transitions). Then, within that collection, exactly one of the following two mutually exclusive scenarios occurs:

1. Unique stable periodic orbit – The Poincaré map associated with one full cycle of region transitions is a contraction. Consequently a single periodic solution exists, it is asymptotically stable, and every trajectory that respects the same region itinerary converges to it.

2. Global attraction to the origin – The same Poincaré map is a strict contraction toward the zero state. In this case all trajectories that follow the prescribed itinerary are drawn to the origin, meaning that all gene expression levels decay to zero regardless of initial conditions.

The proof proceeds by constructing the explicit linear‑affine transition map for each region, then chaining them together to obtain the global return map. The alignment condition forces the composite map to be either monotone decreasing or monotone increasing with a Lipschitz constant strictly less than one. Classical fixed‑point arguments then yield the dichotomy above. Importantly, the result holds irrespective of the sign pattern of the feedback loops, the number of thresholds, or the heterogeneity of decay rates. Hence, even in networks with several intertwined positive and negative loops, the long‑term dynamics are confined to the two simple possibilities described.

To illustrate the theory, three numerical examples are presented. The first example features two negative feedback loops that intersect; the alignment condition is satisfied and a stable two‑period orbit emerges. The second example mixes a positive and a negative loop; parameter choices lead to a contraction toward the origin, so all trajectories vanish. The third example uses asymmetric threshold placements; by varying a single parameter the system switches between the periodic regime and the origin‑attracting regime, demonstrating the sharpness of the theorem. Each simulation displays time series, region‑transition sequences, and the geometry of the focal‑point alignment, providing an intuitive picture of how the abstract conditions manifest in concrete gene‑network models.

Overall, the work extends earlier results that were limited to a single negative feedback loop. By introducing the focal‑point alignment condition, the authors obtain a robust, graph‑independent criterion that guarantees either a unique stable limit cycle or global extinction. This contributes a valuable theoretical tool for synthetic biology and systems biology, where designing gene circuits with predictable dynamical behavior is a central challenge.