Source-linked AI summary
The Diverse Club: The Integrative Core of Complex Networks
M. A. Bertolero, B. T. T. Yeo, M. D'Esposito
TL;DR
The paper asks whether high-strength rich-club nodes are the main integrative core of complex networks or whether nodes with diverse community connections better support integration. It compares these clubs across networks and models their possible evolutionary origins, finding that the diverse club is more interconnected, more critical for efficient communication, and selectively produced by a modularity–efficiency model.
Problem
Rich-club nodes are commonly assumed to support global network integration, but nodes with connections distributed across communities may provide a more advantageous integrative structure.
Method
The study compares high-strength rich clubs with high-participation-coefficient diverse clubs across biological and engineered networks and uses a generative model balancing modularity and efficiency.
Results
Across analyzed networks, the diverse club was more interconnected and its edges were more critical for efficient global communication than rich-club edges; the model produced a diverse club but not a rich club.
Takeaways & Limitations
The diverse club is proposed as the integrative core of complex networks, warranting separate analysis of diverse and rich clubs.
Takeaways & Limitations
Community-detection algorithms have resolution limits, and networks may contain multiple valid scales of community structure; weighted-network thresholding is also arbitrary.
Abstract
from arXiv · showhide
A complex system can be represented and analyzed as a network, where nodes represent the units of the network and edges represent connections between those units. For example, a brain network represents neurons as nodes and axons between neurons as edges. In many networks, some nodes have a disproportionately high number of edges. These nodes also have many edges between each other, and are referred to as the rich club. In many different networks, the nodes of this club are assumed to support global network integration. However, another set of nodes potentially exhibits a connectivity structure that is more advantageous to global network integration. Here, in a myriad of different biological and man-made networks, we discover the diverse club--a set of nodes that have edges diversely distributed across the network. The diverse club exhibits, to a greater extent than the rich club, properties consistent with an integrative network function--these nodes are more highly interconnected and their edges are more critical for efficient global integration. Moreover, we present a generative evolutionary network model that produces networks with a diverse club but not a rich club, thus demonstrating that these two clubs potentially evolved via distinct selection pressures. Given the variety of different networks that we analyzed--the c. elegans, the macaque brain, the human brain, the United States power grid, and global air traffic--the diverse club appears to be ubiquitous in complex networks. These results warrant the distinction and analysis of two critical clubs of nodes in all complex systems.
Results
Across biological and engineered networks, the diverse club—nodes with high participation coefficients—was more interconnected and more centrally distributed across communities than the rich club. Its edges were more critical for efficient global communication, and a model selected for modularity and efficiency produced a diverse club without a rich club.
- Clubness: Across networks, the diverse club was typically more interconnected than the rich club, with clubness equal to or higher at high participation-coefficient ranks.This pattern remained similar after standard-deviation normalization and edge-weight shuffling.
- Club topology: The diverse and rich clubs were mostly distinct: human networks had no more than 23% node overlap, while diverse-club nodes appeared in at least as many communities as rich-club nodes.The distinction was also visible anatomically and topologically, with diverse-club nodes near network centers and rich-club nodes toward peripheries.
- Network integration: Edges within the diverse club had higher edge betweenness than rich-club edges across almost all networks, indicating greater involvement in shortest-path communication.Node-level betweenness centrality itself was not consistently higher for either club.
- Network integration: Removing diverse-club edges increased the sum of shortest paths more than removing rich-club edges in every network, indicating greater disruption of efficient global communication.The analysis randomly removed 50–90% of intra-club edges across 10,000 iterations while avoiding graph disconnection.
- Cognitive activity: Diverse-club activity increased as cognitive tasks engaged more communities or components, whereas rich-club activity decreased.For the diverse club, mean correlations were r=0.45 with communities and r=0.395 with cognitive components.
- Generative model: A generative model balancing modularity and efficiency produced a highly interconnected diverse club but not a rich club, supporting distinct selection pressures.At a 0.75 modularity-to-efficiency weighting ratio, the model matched several network properties and exceeded random models in diverse-club clubness.
Discussion
Across the analyzed networks, the diverse club appears to be a more integrative core than the rich club, while the two clubs remain largely distinct and may reflect different evolutionary pressures. The findings motivate analyzing both clubs because the rich club may instead support network stability.
- Comparative network properties: The diverse club is more interconnected than the rich club in every analyzed network, reaching up to four times the rich club’s interconnectedness in the human brain.Few nodes belong to both clubs across the examined networks.
- Comparative network properties: Diverse-club connectivity spans more communities, has equal betweenness centrality, and higher edge betweenness than rich-club connectivity.This connectivity pattern is distributed across the network and provides economical routes between nodes, properties associated with integration across communities.
- Comparative network properties: Removing diverse-club edges increases the sum of shortest paths more than removing rich-club edges, indicating greater importance for efficient global communication.The same pattern was observed across all analyzed networks.
- Evolutionary origins: A selection model balancing modularity and efficient integration produces a highly interconnected diverse club but not a rich club.The evolutionary model also produces significantly higher diverse-club clubness than a random null model.
- Functional interpretation: The authors propose that the diverse club is the true integrative core, whereas the rich club may instead help stabilize network dynamics through slower processing.The proposed distinction remains open to further investigation, particularly for the rich club’s potential stability function.
Supplementary Methods
The methods detect communities with multiple algorithms, quantify participation and strength, and define normalized rich and diverse clubness against randomized networks. They evaluate network efficiency and apply these procedures across biological and man-made networks.
- Community detection: Community detection algorithms group nodes according to different definitions, including modularity, random walks, local neighbor structure, edge betweenness, or spin-glass energy.Louvain and spectral methods maximize Q, whereas other algorithms use alternative structural or dynamical principles.
- Community detection: Resolution limits can obscure smaller communities, so Louvain stability and Walktrap N were used to vary community scale without arbitrarily removing edges.The analyses varied density, Louvain resolution, and Walktrap community number across networks.
- Participation coefficients: Participation coefficient measures how evenly a node’s edge weights are distributed across communities, reaching its maximum when weights are equal across all communities.It is zero when all of a node’s edges connect to a single community.
- Clubs and clubness: Clubs are formed by rank-ordering nodes by strength or participation coefficient, then measuring within-club edge-weight density as clubness.The normalized clubness coefficient compares observed clubness with the mean from randomized networks preserving degree and strength distributions.
- Efficiency and datasets: Network efficiency is based on inverse shortest-path sums, with binary edges used for shortest-path calculations across structural and functional datasets.The study analyzes C. elegans, macaque, and human neural networks, the United States power grid, and global air traffic.