[Paper Review] Dynamics of Boolean Networks with Scale-Free Topology

Maximino Aldana|arXiv (Cornell University)|Sep 25, 2002

Gene Regulatory Network Analysis21 citations

TL;DR

This paper investigates Boolean networks with scale-free topology, showing analytically that a phase transition occurs at a scale-free exponent γ ∈ (2, 2.5), eliminating the need for fine-tuning parameters p and K to achieve stable dynamics. The study demonstrates that scale-free input and output topologies are equivalent, and the critical phase emerges naturally without requiring low connectivity or extreme bias in update functions.

ABSTRACT

The dynamics of Boolean networks (the N-K model) with scale-free topology are studied here. The existence of a phase transition governed by the value of the scale-free exponent of the network is shown analytically by analyzing the overlap between two distinct trajectories. The phase diagram shows that the phase transition occurs for values of the scale-free exponent in the open interval (2,2.5). Since the Boolean networks under study are directed graphs, the scale-free topology of the input connections and that of the output connections are studied separately. Ultimately these two topologies are shown to be equivalent. An important result of this work is that the fine-tuning usually required to achieve stability in Boolean networks with a totally random topology is no longer necessary when the network topology is scale-free.

Motivation & Objective

• To investigate whether scale-free topology in Boolean networks eliminates the need for fine-tuning parameters p and K to achieve stable dynamics.
• To determine how the scale-free exponent γ governs the dynamical phase transition in directed Boolean networks.
• To compare the statistical properties of input (in-degree) and output (out-degree) topologies in directed scale-free networks.
• To assess whether real genetic networks, with heterogeneous connectivity, can support stable, ordered dynamics without parameter fine-tuning.

Proposed method

• Analyzes the overlap between two trajectories in the network to detect phase transitions.
• Uses the Derrida analysis framework to compute the stability of perturbations in the network dynamics.
• Derives the input degree distribution P_I(k) from the output degree distribution P_o(l) under random wiring assumptions.
• Applies the configuration model to show that the first moment of input and output degree distributions are equal in the N→∞ limit.
• Solves for the critical scale-free exponent γ_c(p) using the condition R = 1, where R is the average number of distinct neighbors reached by a perturbation.
• Constructs a phase diagram plotting γ_c(p) for different values of p, showing γ_c ∈ [2, 2.5] with maximum at p=0.5.

Experimental results

Research questions

• RQ1 Does a scale-free topology in Boolean networks lead to a phase transition governed by the scale-free exponent γ?
• RQ2 Can the dynamical stability of Boolean networks be achieved without fine-tuning p and K, as required in the standard N-K model?
• RQ3 Are the input and output degree distributions in directed scale-free networks equivalent in the thermodynamic limit?
• RQ4 What is the critical value of the scale-free exponent γ_c for which the system transitions from ordered to chaotic dynamics?
• RQ5 Does the phase transition window γ ∈ (2, 2.5) allow for realistic connectivity distributions observed in real genetic networks?

Key findings

• The phase transition in Boolean networks with scale-free topology occurs for scale-free exponents in the open interval (2, 2.5), with γ_c reaching a maximum of approximately 2.47875 at p = 0.5.
• For γ > γ_c ≈ 2.47875, the system remains in the ordered phase regardless of p, indicating robust stability.
• The input degree distribution P_I(k) converges to a Poisson distribution in the N→∞ limit when output degrees l_i follow a scale-free distribution P_o(l) ∝ l^{-γ}.
• The first moment of the input distribution ⟨k⟩_I equals the first moment of the output distribution ⟨l⟩_o, ensuring equivalence between input and output topologies in the thermodynamic limit.
• The requirement for fine-tuning p to achieve stability in the standard N-K model is eliminated when the network topology is scale-free.
• The model supports a wide range of connectivities—consistent with real genetic networks—without requiring extreme values of p or low K, making it more biologically plausible.