[Paper Review] Roles of hubs in Boolean networks

Chikoo Oosawa|ArXiv.org|Mar 16, 2007

Gene Regulatory Network Analysis2 references4 citations

TL;DR

This study investigates the role of hubs in Boolean networks with power-law outdegree distributions, showing that hubs with higher outdegrees reduce path length and amplify signal propagation. The type of Boolean function at hubs (XOR > AND-OR > CONSTANT) significantly influences entropy and mutual information, indicating hubs critically shape network coherence and dynamics.

ABSTRACT

We examined the effects of inhomogeneity on the dynamics and structural properties using Boolean networks. Two different power-law rank outdegree distributions were embedded to determine the role of hubs. The degree of randomness and coherence of the binary sequence in the networks were measured by entropy and mutual information, depending on the number of outdegrees and types of Boolean functions for the hub. With a large number of outdegrees, the path length from the hub reduces as well as the effects of Boolean function on the hub are more prominent. These results indicate that the hubs play important roles in networks' dynamics and structural properties. By comparing the effect of the skewness of the two different power-law rank distributions, we found that networks with more uniform distribution exhibit shorter average path length and higher event probability of coherence but lower degree of coherence. Networks with more skewed rank distribution have complementary properties. These results indicate that highly connected hubs provide an effective route for propagating their signals to the entire network.

Motivation & Objective

• To understand how inhomogeneous connectivity, particularly hubs, affects dynamical and structural properties in Boolean networks.
• To investigate the influence of hub outdegree and Boolean function type on entropy and mutual information as measures of randomness and coherence.
• To compare two power-law outdegree distributions (skewed vs. uniform) in terms of path length, coherence, and attractor dynamics.
• To determine whether hubs act as effective signal propagators due to structural and functional properties.

Proposed method

• Constructed 10,ing 4 Boolean networks with fixed size (256 nodes, 512 edges) and K_in = 2 using two power-law outdegree distributions (γ ≈ 0.8 and 0.5).
• Assigned 16 Boolean functions (XOR, AND-OR, CONSTANT) to nodes with equal probability, focusing on hubs with varying outdegrees.
• Measured entropy (randomness) and mutual information (coherence) from attractor dynamics across 2×10^4 initial states per network.
• Calculated path length from hubs using the formula i = log_{K_in}( (N-1)(K_in-1)/X + 1 ), where X is hub outdegree.
• Analyzed dependence of entropy and mutual information on hub function type and outdegree rank across both distribution types.
• Used attractor statistics (size, number, event probability) to compare dynamical behavior between type I and type II distributions.

Experimental results

Research questions

• RQ1 How does the number of outdegrees from a hub affect path length and signal propagation in Boolean networks?
• RQ2 To what extent does the type of Boolean function at a hub influence entropy and mutual information in the network?
• RQ3 How do differences in outdegree distribution skewness (type I vs. type II) affect average path length, coherence, and attractor dynamics?
• RQ4 What is the relationship between hub rank, outdegree, and the emergence of global coherence in state variables?
• RQ5 Does structural connectivity via hubs alone ensure coherence, or is the functional role of the Boolean function at the hub critical?

Key findings

• Hubs with higher outdegrees significantly reduce path length to downstream nodes, with path length decreasing as outdegree increases (e.g., X=1 → i=8 steps).
• The type of Boolean function at hubs strongly influences mutual information: XOR > AND-OR > CONSTANT, indicating functional role in coherence.
• Networks with more skewed outdegree distributions (type I, γ≈0.8) exhibit shorter average path lengths and stronger dependence on hub function type.
• Networks with more uniform distributions (type II, γ≈0.5) show higher event probability of positive entropy and mutual information, despite longer path lengths.
• Type I networks have higher mutual information and entropy magnitudes but lower event probability of coherence compared to type II.
• The number of attractors is significantly higher in type II networks (459,240) than in type I (137,254), indicating greater dynamical diversity in less skewed topologies.