Network architecture is a brain-organizational motif present across spatial scales from cell assemblies to distributed systems. Structural pathology in some neurodegenerative disorders selectively afflicts a subset of functional networks, motivating the network degeneration hypothesis (NDH). Recent evidence suggests that structural pathology recapitulating physiology may be a general property of neuropsychiatric disorders. To test this possibility, we compared functional and structural network meta-analyses drawing upon the BrainMap database. The functional meta-analysis included results from >7,000 experiments of subjects performing >100 task paradigms; the structural meta-analysis included >2,000 experiments of patients with >40 brain disorders. Structure-function network concordance was high: 68% of networks matched (pFWE < 0.01), confirming the broader scope of NDH. This correspondence persisted across higher model orders. A positive linear association between disease and behavioral entropy (p = 0.0006;R2 = 0.53) suggests nodal stress as a common mechanism. Corroborating this interpretation with independent data, we show that metabolic ‘cost’ significantly differs along this transdiagnostic/multimodal gradient.
Introduction
Network architecture is a fundamental, multi-scale motif in brain organization, presumably reflecting evolutionary pressure for efficient information-processing. Network properties have been demonstrated across a wide range of spatial scales. The microscale contains connections between individual neurons, while the macroscale comprises distributed systems which encompass direct and indirect connections between more distant brain regions1. Progress in neuroscience over the past three decades has been extraordinary, much of which can be attributed to the development of high-resolution, whole-brain imaging methods, and advanced analytic approaches for network discovery. For human neuroscience, functional and structural magnetic resonance imaging (fMRI and sMRI) coupled with data-driven analytic methods applied at the systems level have been particularly impactful2,3,4.
System-level functional networks are defined by their functional connectivity, most often inferred by measuring temporal correlations of neuronal activity5. The reliability of functional connectivity as a biological construct has withstood rigorous examination: the functional organization of the brain is dominated by a core architecture shared between individuals (i.e., system-level networks), but stable connectivity features unique to individuals are also present6. Substantially less variability in functional brain connectivity is explained by day-to-day variability or even task-state6. Thus, functional connectivity is a robust metric to study behavior, cognition, and disease7. Fifteen to 20 functional networks are readily identifiable and can account for much of our understanding of brain–behavior ontology8,9,10. These functional circuits are considered to mediate susceptibility to dimensions of psychopathology rather than discrete disorders11, making them especially relevant for transdiagnostic investigation12.
The network denegation hypothesis (NDH) posits that disease-related structural alteration selectively occurs—and may even spread—within these system-level functional networks13. Just like neuronal activity, gray matter structural alteration (atrophy or hypertrophy) has shown to follow network-based principles14: structural alteration in one brain area is influenced by alteration in other brain areas15,16. This concept is hereon referred to as co-alteration structural connectivity (CA-SC). Previous work has linked four specific neurodegenerative disorders’ atrophy patterns to four corresponding functional circuits17, and other work has found shared CA-SC effects in a single functional circuit across six psychiatric diagnoses18. In recent years, some studies have even suggested that network-based pathology recapitulating physiology may be a general property of neuropsychiatric disorders that can occur in response to a combination of plausible disease mechanisms14. But this fundamental question of NDH, specifically the extent of structural and functional correspondence, remains unclear. Furthermore, toward a refined understanding of NDH, new evidence has even suggested that the variety and unpredictability of diseases that structurally affect a brain area may be associated with regions that are important for cognitive/integrative function19. Transdiagnostic disease vulnerability of brain networks has not yet been assessed and compared to functionally based, integrative indexes (i.e., behavioral specialization) in any formal capacity. Confirming this hypothesis would have important implications for further understanding NDH. Functional specialization is proposed to reflect unique metabolic brain characteristics20 that may underlie specific NDH mechanisms.
Meta-analytic network analysis, which draws upon decades of human neuroimaging research, has proven to be a powerful way to study brain organization and pathology21. Specifically, the BrainMap (www.brainmap.org) database project has involved the manual curation of standardized results (x–y–z brain coordinates) from thousands of whole-brain functional and structural neuroimaging experiments, along with a rich taxonomy of the relevant behavior (i.e., behavioral domain and task paradigm) and disease (i.e., ICD-10 diagnosis) metadata, respectively. In utilizing this dual-modality resource for network analysis and investigating NDH, a plethora of analytic methods are available. Independent component analysis (ICA), which requires perhaps the least assumptions of neuroimaging data as opposed to other analytic models2,22, can be applied to meta-analytic data from BrainMap as has been done previously10,23. ICA is a multivariate method that identifies a specified number of spatial networks by linearly unmixing whole-brain data into maximally independent sources24.
In the first part of this study, we test NDH’s broad-based proposition that network-based structural pathology adheres to the brain’s functional architecture when considering many neuropsychiatric disorders. We test this hypothesis in a data-driven manner by spatially comparing 20 transdiagnostic CA-SC networks to 20 task-activation functional connectivity (TA-FC) ICA networks; each network set was separately generated from their respective BrainMap modality (structure vs. function). We also examine higher model orders (d = 45, 70) to assess generalizability. In the second part of this work, utilizing a network-normalized entropy metric derived from metadata loadings (which captures both the diversity and non-specificity of metadata associations), we test the hypothesis that brain networks that are highly behavior entropic are also highly disease entropic. To further investigate the specific NDH mechanistic prediction of nodal stress (NS), we utilize a separate dataset25 that reported regional differences in the brain’s metabolic attributes among healthy individuals, and we test whether those markers associate with disease and behavior entropy.
