A discrete inhomogeneous model for the yeast cell cycle

We study the robustness and stability of the yeast cell regulatory network by using a general inhomogeneous discrete model. We find that inhomogeneity, on average, enhances the stability of the biggest attractor of the dynamics and that the large size of the basin of attraction is robust against changes in the parameters of inhomogeneity. We find that the most frequent orbit, which represents the cell-cycle pathway, has a better biological meaning than the one exhibited by the homogeneous model.

💡 Research Summary

The paper extends the classic Boolean‐type model of the budding yeast (Saccharomyces cerevisiae) cell‑cycle regulatory network by allowing each regulatory interaction to have its own strength rather than assuming uniform coupling. The authors construct a discrete‑time, synchronous update system that includes eleven key genes (Cln3, MBF, SBF, Cln1/2, Clb5/6, Cdc20, Cdh1, etc.) and thirteen directed regulatory links. For every link a real‑valued weight is drawn from a chosen distribution; the mean (μ) and variance (σ) of this distribution control the degree of inhomogeneity. The state of each node at the next time step is determined by the sign of the weighted sum of its inputs, preserving the Boolean nature of the original model while introducing quantitative heterogeneity.

A massive computational experiment was performed: one million random initial configurations were evolved for many different (μ, σ) pairs. For each parameter set the authors catalogued all attractors (fixed points or limit cycles) and measured the size of their basins of attraction. The central findings are threefold. First, increasing σ – i.e., making the interaction strengths more heterogeneous – raises the probability that the largest attractor corresponds to the biologically relevant G1 fixed point. In other words, heterogeneity tends to push the system toward the quiescent state that precedes the cell‑cycle start. Second, the basin of this dominant attractor remains remarkably robust: its relative size changes little across a wide range of σ, and even shows a modest increase for moderate heterogeneity. This demonstrates that the network’s global stability is not fragile to variations in individual interaction strengths. Third, the most frequently observed cyclic attractor in the inhomogeneous regime has a period of 13 and reproduces the canonical G1 → S → G2 → M progression of yeast cells. By contrast, the homogeneous model often yields a 12‑step cycle that includes biologically implausible transitions (e.g., a direct G1→G2 jump). Thus, the heterogeneous model provides a more faithful representation of the actual cell‑cycle pathway.

The authors also identify “weak points” in the parameter space: certain combinations of low inhibitory weights can fragment the large basin, giving rise to many small attractors and destabilizing the cycle. This sensitivity suggests that while overall heterogeneity is stabilizing, specific alterations in key regulatory links can have disproportionate effects—a finding with potential implications for drug targeting or synthetic biology interventions aimed at modulating cell‑cycle dynamics.

In the discussion, the paper argues that robustness of the yeast cell‑cycle network derives not only from its wiring diagram but also from the distribution of interaction strengths. The results support the view that biological networks are evolutionarily tuned to tolerate fluctuations in protein concentrations, binding affinities, and post‑translational modifications. The authors propose future work that would integrate experimentally measured kinetic parameters (e.g., binding constants, phosphorylation rates) to replace the random weight sampling, thereby bridging the gap between abstract Boolean models and detailed continuous differential‑equation frameworks. Overall, the study demonstrates that incorporating realistic inhomogeneity enhances both the stability and biological relevance of discrete models of cellular regulatory circuits.