Evolution of a population of random Boolean networks

We investigate the evolution of populations of random Boolean networks under selection for robustness of the dynamics with respect to the perturbation of the state of a node. The fitness landscape contains a huge plateau of maximum fitness that spans the entire network space. When selection is so strong that it dominates over drift, the evolutionary process is accompanied by a slow increase in the mean connectivity and a slow decrease in the mean fitness. Populations evolved with higher mutation rates show a higher robustness under mutations. This means that even though all the evolved populations exist close to the plateau of maximum fitness, they end up in different regions of network space.

💡 Research Summary

The paper investigates how populations of random Boolean networks (RBNs) evolve when the sole selective pressure is robustness of the network dynamics to a perturbation of a single node’s state. The authors define fitness as the probability that, after flipping the state of a randomly chosen node, the network returns to its original attractor. This fitness measure directly captures dynamical resilience rather than static properties such as average connectivity or Boolean function bias.

Using a standard evolutionary algorithm, each generation consists of copying the current population, applying mutations (rewiring connections or changing Boolean functions) at a configurable mutation rate μ, and then selecting the next generation based on fitness. Two regimes of selection strength are explored: strong selection, where fitness differences dominate drift, and weak selection, where stochastic drift can outweigh selection. Network size is fixed at N = 100, while the average in‑degree K is initially drawn from a uniform distribution between 1 and 5, and Boolean update functions are chosen uniformly.

A key finding is that the fitness landscape contains a vast plateau of maximal fitness. Within this plateau, networks with widely varying connectivities and Boolean functions all achieve the same highest fitness value. Consequently, evolutionary dynamics are governed by the interplay between drift and selection rather than by a gradient toward a single optimum.

When selection is strong, the population drifts across the plateau in a way that slowly raises the mean connectivity ⟨K⟩ while causing a subtle decline in mean fitness ⟨F⟩. The increase in ⟨K⟩ reflects a trend toward more densely connected networks, which paradoxically makes the system slightly more vulnerable to node perturbations because each node’s state influences a larger portion of the network. Nonetheless, the fitness loss is modest because the population remains on the plateau.

Mutation rate exerts a pronounced effect on mutational robustness. Populations evolved under higher μ (e.g., 0.05–0.1) develop networks that retain higher fitness after an additional random mutation compared with low‑μ populations. This indicates that high‑mutation environments push the population toward regions of the plateau that are more “neutral” with respect to mutations—i.e., where many mutational neighbours share the same fitness. Thus, even though all evolved networks lie on the same maximal‑fitness plateau, they occupy distinct sub‑regions characterized by different structural properties.

The authors interpret these results in the context of neutral evolution and the concept of a fitness plateau. The coexistence of many genotypes with identical fitness allows drift to explore genotype space, while selection shapes the statistical properties of the explored region (e.g., average connectivity). The observed trade‑off between connectivity and robustness mirrors findings in biological gene‑regulatory networks, where higher connectivity can increase functional integration but also amplify the impact of perturbations.

Implications for synthetic biology and engineered complex systems are highlighted. Introducing a relatively high mutation rate during the design or training phase may deliberately steer artificial networks toward mutation‑robust configurations, which could be advantageous for hardware implementations subject to faults or for adaptive algorithms operating in noisy environments.

In summary, the study demonstrates that random Boolean network populations evolve on a broad maximal‑fitness plateau, with strong selection driving a gradual increase in connectivity and a slight fitness erosion, while higher mutation rates foster greater mutational robustness. These dynamics illustrate how complex systems can simultaneously achieve high functional performance and resilience, and they provide a concrete computational model for exploring the balance between robustness, connectivity, and evolutionary forces.