The impact of graph construction scheme and community detection algorithm on the repeatability of community and hub identification in structural brain networks

Community detection in the human connectome: Method types, differences and their impact on inference

Community structure is a fundamental topological characteristic of optimally organized brain networks. Currently, there is no clear standard or systematic approach for selecting the most appropriate community detection method. Furthermore, the impact of method choice on the accuracy and robustness of estimated communities (and network modularity), as well as method-dependent relationships between network communities and cognitive and other individual measures, are not well understood. This study analyzed large datasets of real brain networks (estimated from resting-state fMRI fromn $$ n $$ = 5251 pre/early adolescents in the adolescent brain cognitive development [ABCD] study), andn $$ n $$ = 5338 synthetic networks with heterogeneous, data-inspired topologies, with the goal to investigate and compare three classes of community detection methods: (i) modularity maximization-based (Newman and Louvain), (ii) probabilistic (Bayesian inference within the framework of stochastic block modeling (SBM)), and (iii) geometric (based on graph Ricci flow). Extensive comparisons between methods and their individual accuracy (relative to the ground truth in synthetic networks), and reliability (when applied to multiple fMRI runs from the same brains) suggest that the underlying brain network topology plays a critical role in the accuracy, reliability and agreement of community detection methods. Consistent method (dis)similarities, and their correlations with topological properties, were estimated across fMRI runs. Based on synthetic graphs, most methods performed similarly and had comparable high accuracy only in some topological regimes, specifically those corresponding to developed connectomes with at least quasi-optimal community organization. In contrast, in densely and/or weakly connected networks with difficult to detect communities, the methods yielded highly dissimilar results, with Bayesian inference within SBM having significantly higher accuracy compared to all others. Associations between method-specific modularity and demographic, anthropometric, physiological and cognitive parameters showed mostly method invariance but some method dependence as well. Although method sensitivity to different levels of community structure may in part explain method-dependent associations between modularity estimates and parameters of interest, method dependence also highlights potential issues of reliability and reproducibility. These findings suggest that a probabilistic approach, such as Bayesian inference in the framework of SBM, may provide consistently reliable estimates of community structure across network topologies. In addition, to maximize robustness of biological inferences, identified network communities and their cognitive, behavioral and other correlates should be confirmed with multiple reliable detection methods.

Test-retest reliability of modular-relevant analysis in brain functional network

Introduction

The human brain could be modeled as a complex network via functional magnetic resonance imaging (fMRI), and the architecture of these brain functional networks can be studied from multiple spatial scales with different graph theory tools. Detecting modules is an important mesoscale network measuring approach that has provided crucial insights for uncovering how brain organizes itself among different functional subsystems. Despite its successful application in a wide range of brain network studies, the lack of comprehensive reliability assessment prevents its potential extension to clinical trials.

Methods

To fill this gap, this paper, using resting-state test-retest fMRI data, systematically explored the reliabilities of five popular network metrics derived from modular structure. Considering the repeatability of network partition depends heavily on network size and module detection algorithm, we constructed three types of brain functional networks for each subject by using a set of coarse-to-fine brain atlases and adopted four methods for single-subject module detection and twelve methods for group-level module detection.

Results

The results reported moderate-to-good reliability in modularity, intra- and inter-modular functional connectivities, within-modular degree and participation coefficient at both individual and group levels, indicating modular-relevant network metrics can provide robust evaluation results. Further analysis identified the significant influence of module detection algorithm and node definition approach on reliabilities of network partitions and its derived network analysis results.

Discussion

This paper provides important guidance for choosing reliable modular-relevant network metrics and analysis strategies in future studies.

The impact of genetic risk for Alzheimer's disease on the structural brain networks of young adults

Introduction

We investigated the structural brain networks of 562 young adults in relation to polygenic risk for Alzheimer's disease, using magnetic resonance imaging (MRI) and genotype data from the Avon Longitudinal Study of Parents and Children.

Methods

Diffusion MRI data were used to perform whole-brain tractography and generate structural brain networks for the whole-brain connectome, and for the default mode, limbic and visual subnetworks. The mean clustering coefficient, mean betweenness centrality, characteristic path length, global efficiency and mean nodal strength were calculated for these networks, for each participant. The connectivity of the rich-club, feeder and local connections was also calculated. Polygenic risk scores (PRS), estimating each participant's genetic risk, were calculated at genome-wide level and for nine specific disease pathways. Correlations were calculated between the PRS and (a) the graph theoretical metrics of the structural networks and (b) the rich-club, feeder and local connectivity of the whole-brain networks.

Results

In the visual subnetwork, the mean nodal strength was negatively correlated with the genome-wide PRS (r = -0.19, p = 1.4 × 10-3), the mean betweenness centrality was positively correlated with the plasma lipoprotein particle assembly PRS (r = 0.16, p = 5.5 × 10-3), and the mean clustering coefficient was negatively correlated with the tau-protein binding PRS (r = -0.16, p = 0.016). In the default mode network, the mean nodal strength was negatively correlated with the genome-wide PRS (r = -0.14, p = 0.044). The rich-club and feeder connectivities were negatively correlated with the genome-wide PRS (r = -0.16, p = 0.035; r = -0.15, p = 0.036).

Discussion

We identified small reductions in brain connectivity in young adults at risk of developing Alzheimer's disease in later life.

The impact of graph construction scheme and community detection algorithm on the repeatability of community and hub identification in structural brain networks

A critical question in network neuroscience is how nodes cluster together to form communities, to form the mesoscale organisation of the brain. Various algorithms have been proposed for identifying such communities, each identifying different communities within the same network. Here, (using test–retest data from the Human Connectome Project), the repeatability of thirty‐three community detection algorithms, each paired with seven different graph construction schemes were assessed. Repeatability of community partition depended heavily on both the community detection algorithm and graph construction scheme. Hard community detection algorithms (in which each node is assigned to only one community) outperformed soft ones (in which each node can belong to more than one community). The highest repeatability was observed for the fast multi‐scale community detection algorithm paired with a graph construction scheme that combines nine white matter metrics. This pair also gave the highest similarity between representative group community affiliation and individual community affiliation. Connector hubs had higher repeatability than provincial hubs. Our results provide a workflow for repeatable identification of structural brain networks communities, based on the optimal pairing of community detection algorithm and graph construction scheme.