Network Neuroscience

Community detection provides a principled lens on mesoscale organization in functional brain networks, yet many widely used methods presume assortative structure and depend on arbitrary thresholding, which complicates the selection of the community count K. We conducted a systematic benchmark of three assumption lean approaches that operate directly on weighted functional connectivity matrices: the Weighted Stochastic Block Model, Spectral Clustering, and K-means. Performance was assessed on synthetic networks with known ground truth and on three neuroimaging cohorts spanning development, namely the Human Connectome Project, Washington University 120, and the Baby Connectome Project. We compared strategies for choosing K, including post hoc indices such as silhouette, Calinski-Harabasz, C index, modularity, variation of information, Normalized Mutual Information, and zRand, together with a likelihood-based criterion for the Weighted Stochastic Block Model that uses bootstrap confidence intervals for differences in log likelihood between successive values of K. In simulations all methods recovered stable partitions, but the post hoc indices favored incorrect values of K under weak signal and nonassortative mixing. In adult datasets the indices do not yield a unique optimum, whereas the likelihood-based criterion selects a parsimonious range centered near K = 11, which is consistent with established sensory and association systems. In infants and toddlers, the same procedure supports a larger K around 15 and reveals developmentally distinct mesoscale architecture, including anterior and posterior subdivisions within default mode and fronto parietal systems. A consensus relabeling scheme based on Hungarian matching with Hamming distance further stabilizes solutions across runs and across values of K. Overall, threshold free weighted methods mitigate assortative bias and the likelihood-based comparison provides a reproducible path to selecting K.

The brains ability to transition between functional states while maintaining both flexibility and stability is shaped by its structural connectivity. Understanding the relationship between brain structure and neural dynamics is a central challenge in neuroscience. Prior studies link neural dynamics to local noisy activity and mesoscale coupling mechanisms, but causal links at the whole-brain scale remain elusive. This study investigates how the balance between integration and segregation in brain networks influences their dynamical properties, focusing on multistability (switching between stable states) and metastability (transient stability over time). We analyzed a spectrum of network models, from highly segregated to highly integrated, using structural metrics like modularity, efficiency, and small-worldness. Simulating neural activity with a neural mass model and analyzing Functional Connectivity Dynamics (FCD), we found that segregated networks sustain dynamic synchronization patterns, while small-world networks, which balance local clustering and global efficiency, exhibit the richest dynamical behavior. Networks with intermediate small-worldness ({omega}) values showed peak dynamical richness, measured by variance in FCD and metastability. Using Mutual Information (MI), we quantified the structure-dynamics relationship, revealing that modularity is the strongest predictor of network dynamics, as modular architectures support transitions between dynamical states. These findings underscore the importance of the small-world architecture in brain networks, where the balance between local specialization and global integration fosters the dynamic complexity necessary for cognitive functions. By emphasizing the role of modularity, this study enhances understanding of how structural features shape neural dynamics and offers insights into disruptions linked to neurological disorders.

The Allen Mouse Brain Connectivity Atlas (AMBCA) is widely used to represent structural connectivity in the mouse brain. The AMBCA consists of tracer injection experiments where neuronal projections axonally connected to the initial injection site are labelled. The resulting whole-brain structural connectomes, derived from a subset of these experiments in C57BL/6 mice, have been used in several studies of connectomic architectures. However, through close inspection of n=437 distinct experiments used in a publicly-available connectome (Knox et al., 2018), we observed experiments with off-target injections, diffuse projections, unrealistically small injections and projections, and anatomical misalignments, affecting the accuracy and applicability of these connectivity experiments. We applied a combined automated and manual quality control (QC) and identified n=56 ([~]13% of the original n=437) experiments representing a wide variety of injection and projection failures across the brain. Automated QC was used to detect extreme injection and projection sizes and misalignments, while manual QC was used to detect subtle off-target tracer spreading. Using the remaining n=381 experiments, we rebuilt two different connectomes using previously-published methods; specifically: the regionalized voxel model from Knox et al. (2018), and the homogeneous model from Oh et al. (2014). Our rebuilt connectomes show strong losses in connectivity between regions with limited evidence of structural connectivity by other methods (e.g. hippocampus-medulla, cerebellum-isocortex) and gains in connectivity between regions with strong connectivity evidence (hypothalamus-cerebellum, hypothalamus-isocortex). Finally, we analyzed the rich club and community organization to demonstrate the potential downstream impacts on the representation of the overall structural connectome architectures of our QCd connectomes and observed subtle whole-brain organizational changes. We present our rebuilt connectomes, and particularly highlight the regionalized voxel model, as more accurate representations of structural connectivity derived from the AMBCA.

The social environment has a critical influence on human development, cognition, and health. By using network approaches to map and analyze the connectivity between all pairs of brain regions simultaneously, we can clarify how relationships between brain regions (e.g. connectivity) change as a function of social relationships. Here we apply multilayer modeling and modularity maximization-both established tools in network neuroscience-to jointly cluster patterns of brain-behavior associations for seven social support measures. Our analyses build on both neuroecological findings and network neuroscientific approaches. In particular we find that subcortical and control systems are especially sensitive to different constructs of perceived social support. Network nodes in these systems are highly flexible; their community affiliations, which reflect groups of nodes with similar patterns of brain behavior associations, differ across social support measures. The multilayer approach used here enables direct comparison of the roles of all regions of the brain across all social support measures included. Additionally, our application of multilayer modeling to patterns of brainbehavior correlations, as opposed to just functional connectivity, represents an innovation in how multilayer models are used in. More than that, it offers a generalizable technique for studying the stability brain-behavior correlations.

It has been proposed that the estimation of the normalized graph Laplacian over a brain networks spectral decomposition can reveal the connectome harmonics (eigenvectors) corresponding to certain frequencies (eigenvalues). Here, I used test-retest dMRI data from the Human Connectome Project to explore the repeatability, and the influence of graph construction schemes on a) graph Laplacian spectrum, b) topological properties, c) high-order interactions (3,4-motifs,odd-cycles), and d) their associations on structural brain networks (SBN). Additionally, I investigated the performance of subjects identification accuracy (brain fingerprinting) of the graph Laplacian spectrum, the topological properties, and the high-order interactions. Normalized Laplacian eigenvalues were found to be subject-specific and repeatable across the five graph construction schemes. The repeatability of connectome harmonics is lower than that of the Laplacian eigenvalues and shows a heavy dependency on the graph construction scheme. A repeatable relationship between specific topological properties of the SBN with the Laplacian spectrum was also revealed. The identification accuracy of normalized Laplacian eigenvalues was absolute (100%) across the graph construction schemes, while a similar performance was observed for a combination of topological properties of SBN (communities,3,4-motifs, odd-cycles) only for the 9m-OMST. Collectively, Laplacian spectrum, topological properties, and high-order interactions characterized uniquely SBN.

The functional interactions between regions of the human brain can be viewed as a network, empowering neuroscientists to leverage tools such as graph theory to obtain insight about brain function. However, obtaining a brain network from functional neuroimaging data inevitably involves multiple steps of data manipulation, which can affect the organisation (topology) of the resulting network and its properties. Test-retest reliability is a gold standard for both basic research and clinical use: a suitable data-processing pipeline for brain networks should recover the same network topology across repeated scan sessions of the same individual. Analyzing resting-state functional Magnetic Resonance Imaging (rs-fMRI) recordings from two test-retest studies across short (45 minutes), medium (2-4 weeks) and long term delays (5-16 months), we investigated the reliability of network topologies constructed by applying 576 unique pipelines to the same fMRI data, obtained from considering combinations of atlas type and size, edge definition and thresholding, and use of global signal regression. We adopted the portrait divergence, an information-theoretic criterion to measure differences in network topology across all scales, enabling us to quantify the influence of different pipelines on the overall organisation of the resulting network. Remarkably, our findings reveal that the choice of pipeline plays a fundamental role in determining how reproducible an individuals brain network topology will be across different scans: there is large and systematic variability across pipelines, such that an inappropriate choice of pipeline can distort the resulting network more than an interval of several months between scans. Across datasets and time-spans, we also identify specific combinations of data-processing steps that consistently yield networks with reproducible topology, enabling us to make recommendations about best practices to ensure high-quality brain networks.

The brains intrinsic organization into resting-state networks has long been suggested to be fundamental for the offline support of mental processes. Extensive task-based evidence support the relevance of the crosstalk between network segregation, supporting systems specialization, and network integration, allowing to flexibly implement complex behavior. However, only scarce evidence focusing on few behavioral measures directly link changes in these network properties at rest with interindividual differences in behavior. In this work, we investigated whether the maintenance of behavior is associated with a segregated intrinsic resting-state networks organization. Using a comprehensive set of behavioral measures spanning cognition, emotion, and personality together with resting-state functional magnetic resonance imaging from the human connectome project, we performed functional connectivity prediction of behavior combined with model interpretability and latent connectivity-behavior factors extraction. We then assessed whether connectivity-behavior patterns were associated with changes in segregation or integration based on GINNA, a 33 resting-state-networks atlas with cognitive characterization, providing opportunities for comparison of the involved cognitive processes. We found that connectivity relevant for behavior organizes into 3 main latent dimensions, summarizing Cognition, Positive Affect and Negative Affect. Crucially, we show that only Cognition, but not Affect, was associated with global network segregation and reduced network integration, suggesting that Cognition is supported by an intrinsic segregated network architecture, necessary for modular specialization, while Affect may rely on distributed mechanisms across intrinsic brain networks. We further reveal differential resting-state-networks involvements, with Cognition associated with the segregation of higher-level resting-state-networks, and the integration of lower-level, visual networks. All in all, the present results reinforce the view that cognition rests upon a segregated intrinsic brain architecture, fostering the maintenance of specialized cognitive modules.

In this work, we propose a new class of graph measures for weighted connectivity information in the human brain based on node relative strengths: relative strength variability (RSV), measuring susceptibility to targeted attacks, and hierarchical RSV (hRSV), a first weighted statistical complexity measure for networks. Using six different network weights for structural connectomes from the UK Biobank, we conduct comprehensive analyses to explore relationships between the RSV and hRSV, and (i) other known network measures, (ii) general cognitive function ( g). Both measures exhibit low correlations with other graph measures across all connectivity weightings indicating that they capture new information of the brain connectome. We found higher g was associated with lower RSV and lower hRSV. That is, higher g was associated with higher resistance to targeted attack and lower statistical complexity. Moreover, the proposed measures had consistently stronger associations with g than other widely used graph measures including clustering coefficient and global efficiency and were incrementally significant for predicting g above other measures for five of the six network weights. Overall, we present a new class of weighted network measures based on variations of relative node strengths which significantly improved prediction of general cognition from traditional weighted structural connectomes.

Brain network modeling uses computer simulations to infer about latent neural properties at micro- and mesoscales by fitting brain dynamic models to empirical data of individual subjects or groups. However, computational costs of (individualized) model fitting is a major bottleneck, limiting the practical feasibility of this approach to larger cohorts and more complex models, and highlighting the need for scalable simulation implementations. Here, we introduce cuBNM, a Python package which leverages parallel processing of graphics processing units to massively accelerate simulations of brain network models. We show running simulations on graphics processing units is several hundred times faster compared to central processing units. We demonstrate the usage of cuBNM by running optimization of group-level and individualized low- and high-dimensional models. As examples of the utility of individualized models, we investigated test-retest reliability and heritability of simulated and empirical measures in the Human Connectome Project dataset, and found simulated features to be fairly reliable and heritable. Overall, cuBNM provides an effcient framework for large-scale simulations of brain network models, facilitating investigations of latent neural processes across larger cohorts, denser networks, and higher-dimensional models, which were previously less feasible in practice.

The hierarchical organization of the brains distributed network has received growing interest from the neuroscientific community, largely because of its potential to enhance our understanding of human cognition and behavior, in health and disease. This interest is motivated by the hypothesis that near-critical brain dynamics enable multiscale integration and segregation of neural dynamics. While most multiscale connectivity analyses focus on structural and functional networks, characterizing the effective connectome across multiple scales has been somewhat overlooked--primarily for computational reasons. The difficulty of estimating large cyclic causal models, together with the scarcity of theoretical frameworks for systematically moving between scales, has hindered progress in this direction. This technical note introduces a top-down multiscale parcellation scheme for dynamic causal models, with application to neuroimaging data. The method is based on Bayesian model comparison, as a generalization of the well-known {Delta}BIC method. To facilitate computation, recent developments in linear dynamic causal modeling (DCM) and Bayesian model reduction (BMR) are deployed. Specifically, a naive version of BMR is introduced, enabling the parcellation scheme to scale to hundreds or thousands of regions. Notably, the derivations reveal an analytical relationship between reduced model evidence and minimum cut problem in graph theory. This duality puts the tools of graph theory at the service of model evidence optimization and significance testing. The proposed method was applied to simulated and empirical causal models to establish face and construct validity. Consequently, the large empirical causal network, inferred from a neuroimaging dataset, exhibited log-log scaling trends, suggestive of scale invariance in multiple dynamical measures. Future generalizations of this technique and its potential applications in systems and clinical neuroscience are discussed.

Network control theory (NCT) is a simple and powerful tool for studying how network topology informs and constrains dynamics. Compared to other structure-function coupling approaches, the strength of NCT lies in its capacity to predict the patterns of external control signals that may alter dynamics in a desired way. We have extensively developed and validated the application of NCT to the human structural connectome. Through these efforts, we have studied (i) how different aspects of connectome topology affect neural dynamics, (ii) whether NCT outputs cohere with empirical data on brain function and stimulation, and (iii) how NCT outputs vary across development and correlate with behavior and mental health symptoms. In this protocol, we introduce a framework for applying NCT to structural connectomes following two main pathways. Our primary pathway focuses on computing the control energy associated with transitioning between specific neural activity states. Our second pathway focuses on computing average controllability, which indexes nodes general capacity to control dynamics. We also provide recommendations for comparing NCT outputs against null network models. Finally, we support this protocol with a Python-based software package called network control theory for python (nctpy).

Functional connectivity (FC) is frequently operationalized as a correlation. Many studies have examined changes in correlation networks across time, claiming to link time-varying fluctuations to ongoing mental operations and physiological processes. Other studies, however, have called these results into question, noting that statistically indistinguishable patterns of time-varying fluctuations can be obtained by windowing synthetic time series generated from ground-truth stationary correlation structure. Recently, we developed a technique for tracking rapid (framewise) fluctuations in network connectivity over time. Here, we show that these "edge time series" are mathematically equivalent to interaction terms in a specific family of general linear models. We exploit this fact to further demonstrate that time-varying connectivity carries explanatory power above and beyond brain activations. This observation suggests that time-varying connectivity is likely more than a statistical artifact. SUMMARYBrain activity and connectivity have been linked to ongoing behavior and mentation but usually in isolation and almost never in the same model. Here, we show that "edge time series" - a recently proposed method for tracking moment-to-moment connectivity changes - are equivalent to an interaction term in a linear model. By including terms for activations in the same model, it provides an elegant framework for assessing the relative explanatory power of edges and activations. In our work, we use this modeling framework to study time-varying behavior in zebrafish, worms, and humans. We find that connectivity contains unique explanatory power above and beyond activity.

The quality of a childs social and physical environment is a key influence on brain development, educational attainment and mental wellbeing. However, there still remains a mechanistic gap in our understanding of how environmental influences converge on changes in the brains developmental trajectory. In a sample of 145 children with structural diffusion tensor imaging data, we used generative network modelling to simulate the emergence of whole brain network organisation. We then applied data-driven clustering to stratify the sample according to socio-economic disadvantage, with one of the resulting clusters containing mostly children living below the poverty line. A formal comparison of the simulated networks from the generative model revealed that the computational principles governing network formation were subtly different for children experiencing socio-economic disadvantage, and that this resulted in significantly altered developmental timing of network modularity emergence. Children in the low socio-economic status (SES) group had a significantly slower time to peak modularity, relative to the higher SES group (t(69) = 3.02, P = 3.50 x 10-4, d = 0.491). In a subsequent simulation we showed that the alteration in generative properties increases the variability in wiring probabilities during network formation (KS test: D = 0.012, P < 0.001). One possibility is that multiple environmental influences such as stress, diet and environmental stimulation impact both the systematic coordination of neuronal activity and biological resource constraints, converging on a shift in the economic conditions under which networks form. Alternatively, it is possible that this stochasticity reflects an adaptive mechanism that creates "resilient" networks better suited to unpredictable environments. Author SummaryWe used generative network models to simulate macroscopic brain network development in a sample of 145 children. Within these models, network connections form probabilistically depending on the estimated "cost" of forming a connection, versus topological "value" that the connection would confer. Tracking the formation of the network across the simulation, we could establish the changes in global brain organisation measures such as integration and segregation. Simulations for children experiencing socio-economic disadvantage were associated with a shift in emergence of a topologically valuable network property, namely modularity.

Understanding brain functionality and predicting human behavior based on functional brain activity is a major goal of neuroscience. Numerous studies have been conducted to investigate the relationship between functional brain activity and attention, subject characteristics, autism, psychiatric disorders, and more. By modeling brain activity data as networks, researchers can leverage the mathematical tools of graph and network theory to probe these relationships. In their landmark study, Smith et al. (2015) analyzed the relationship of young adult connectomes and subject measures, using data from the Human Connectome Project (HCP). Using canonical correlation analysis (CCA), Smith et al. found that there was a single prominent CCA mode which explained a statistically significant percentage of the observed variance in connectomes and subject measures. They also found a strong positive correlation of 0.87 between the primary CCA mode connectome and subject measure weights. In this study, we computationally replicate the findings of the original study in both the HCP 500 and HCP 1200 subject releases. The exact computational replication in the HCP 500 dataset was a success, validating our analysis pipeline for extension studies. The extended replication in the larger HCP 1200 dataset was partially successful and demonstrated a dominant primary mode.

General intelligence, referred to as g, is hypothesized to emerge from the capacity to dynamically and adaptively reorganize macroscale brain connectivity. Temporal reconfiguration can be assessed using dynamic functional connectivity (dFC), which captures the propensity of brain connectivity to transition between a recurring repertoire of distinct states. Conventional dFC metrics commonly focus on categorical state switching frequencies which do not fully assess individual variation in continuous connectivity reconfiguration. Here, we supplement frequency measures by quantifying within-state connectivity consistency, dissimilarity between connectivity across states, and conformity of individual connectivity to group-average state connectivity. We utilized resting-state fMRI data from the large-scale Human Connectome Project and applied data-driven multivariate Partial Least Squares Correlation to explore emergent associations between dynamic network properties and cognitive ability. Our findings reveal a positive association between g and the stable maintenance of states characterized by distinct connectivity between higher-order networks, efficient reconfiguration (i.e., minimal connectivity changes during transitions between similar states, large connectivity changes between dissimilar states), and ability to sustain connectivity close to group-average state connectivity. This hints at fundamental properties of brain-behavior organization, suggesting that general cognitive processing capacity is supported by the ability to efficiently reconfigure between stable and population-typical connectivity patterns. Impact StatementNovel evidence for an association between the stability, efficiency, and typicality of macro-scale dynamic functional connectivity patterns of the brain and higher general intelligence.

Interacting sets of nodes and fluctuations in their interaction are important properties of a dynamic network system. In some cases the edges reflecting these interactions are directly quantifiable from the data collected. However, in many cases (such as functional magnetic resonance imaging (fMRI) data), the edges must be inferred from statistical relations between the nodes. Here we present a new method, Temporal Communities through Trajectory Clustering (TCTC), that derives time-varying communities directly from time-series data collected from the nodes in a network. First, we verify TCTC on resting and task fMRI data by showing that time-averaged results correspond with expected static connectivity results. We then show that the time-varying communities correlate and predict single-trial behaviour. This new perspective on temporal community detection of node-collected data identifies robust communities revealing ongoing spatiotemporal community configurations during task performance.

Popular methods for analyzing the brains functional connectome examine statistical associations between pairs of atlas-defined brain regions, viewing the strength of these links as independent values. However, edges within a standard connectivity matrix, i.e., correlations between individual regions or nodes, are not independent. They are part of an interconnected system. Here, we propose that consideration of both independent, linear relationships (as in standard approaches such as linear kernel ridge regression and connectome-based predictive modeling) as well as higher order statistical associations - such as tertiary interactions between matrix components and global features of the matrix space - will enhance identification of meaningful individual differences. To test this, we adopt a geometrically grounded measure of similarity that accounts for higher-order local statistical relationships and global interactions, the Wasserstein metric. Results indicate that considering connectivity matrices as representations of their associated Gaussian distributions significantly improves both identification of individuals based on their connectivity matrices (aka, fingerprinting) and prediction of individual differences in phenotypes such as fluid intelligence and openness to experience. Thus, both pairwise local and global brain connectivity properties encode for meaningful individual differences that relate to phenotypic expressions and should be considered in brain-behavior predictive models.

Reproducible functional connectivity-based biomarkers have remained elusive despite the promise of deeply phenotyped consortia. An important component of reproducibility is reliability over repeated measures, often measured by the intraclass correlation coefficient (ICC). Here, we test the use of functional connectome fingerprinting as a way to select pre- and post-processing parameters. We hypothesized that whichever parameters yielded the best fingerprint accuracies would also improve the ICC across scans. Using five datasets from the Consortium for Replicability and Reproducibility, we found that higher identification accuracies were achieved when using: (I) global signal regression; (II) finer brain parcellations; (III) cortical regions compared to subcortical and cerebellar structures; (IV) medial frontal and frontoparietal networks relative to the whole-brain; (V) discriminative edges; (VI) longer scan duration; and (VII) lower sample size. We observed that the ICC was consistently "poor" across the five datasets even with the application of two optimal fingerprint-informed pipelines. The fingerprint-informed pipelines may enable comparison, benchmarking, and adjudication of functional connectivity-based analysis pipelines or novel analytic approaches, as a means to enhance their reproducibility in heterogeneous populations. Key PointsO_LIConnectome-based fingerprinting can provide a useful testbed for reproducible functional connectivity analysis pipelines. C_LIO_LIFingerprint-informed pipelines offer an intuitive and less resource intensive way to select data pre-/post-processing parameters for improving the reproducibility of the functional connectome. C_LIO_LIConnectome-based fingerprinting offers an alternative approach to test-retest reliability. C_LI

Functional connectivity (FC) refers to the statistical dependencies between activity of distinct brain areas. To study temporal fluctuations in FC within the duration of a functional magnetic resonance imaging (fMRI) scanning session, researchers have proposed the computation of an edge time series (ETS) and their derivatives. Evidence suggests that FC is driven by a few time points of high-amplitude co-fluctuation (HACF) in the ETS, which may also contribute disproportionately to interindividual differences. However, it remains unclear to what degree different time points actually contribute to brain-behaviour associations. Here, we systematically evaluate this question by assessing the predictive utility of FC estimates at different levels of co-fluctuation using machine learning (ML) approaches. We demonstrate that time points of lower and intermediate co-fluctuation levels provide overall highest subject specificity as well as highest predictive capacity of individual-level phenotypes.

Neurological pathologies as e.g. Alzheimers Disease or Multiple Sclerosis are often associated to neurodegenerative processes affecting the strength and the transmission speed of long-range inter-regional fiber tracts. Such degradation of Structural Connectivity impacts on large-scale brain dynamics and the associated Functional Connectivity, eventually perturbing network computations and cognitive performance. Functional Connectivity however is not bound to merely mirror Structural Connectivity, but rather reflects the complex coordinated dynamics of many regions. Here, using analytical characterizations of toy models and computational simulations connectome-base whole-brain models, we predict that suitable modulations of regional dynamics could precisely compensate for the effects of structural degradation, as if the original Structural Connectivity strengths and speeds of conduction were effectively restored. The required dynamical changes are widespread and aspecific (i.e. they do not need to be restricted to specific regions) so that they could be potentially implemented via neuromodulation or pharmacological therapy, globally shifting regional excitability and/or excitation/inhibition balance. Computational modelling and theory thus suggest that, in the future therapeutic interventions could be designed to "repair brain dynamics" rather than structure to boost functional connectivity without having to block or revert neurodegenerative processes. AUTHOR SUMMARYNeurological disorders affect Structural Connectivity, i.e. the wiring infrastructure interlinking distributed brain regions. Here we propose that the resulting disruptions in Functional Connectivity, i.e. inter-regional coordination and information sharing, could be compensated by modifying local dynamics so to effectively emulate the restoration of Structural Connectivity (but through a suitable "software patch" rather than by repairing the "hardware"). For simple toy models involving a few regions we can achieve an analytical understanding of how structural and dynamical changes jointly control Functional Connectivity. We then show that the concept of "effective connectome change" via modulation of dynamics robustly extend also to simulation of large-scale models embedding realistic whole-brain connectivity. We thus forecast that novel therapeutic strategies could be devised, targeting dynamics rather than neurodegenerative mechanisms.