nature reviews neuroscience https://doi.org/10.

1038/s41583-023-00752-3

Review article Check for updates

Towards a biologically annotated

brain connectome
Vincent Bazinet, Justine Y. Hansen & Bratislav Misic
Abstract Sections

The brain is a network of interleaved neural circuits. In modern Introduction

connectomics, brain connectivity is typically encoded as a network Embracing regional

of nodes and edges, abstracting away the rich biological detail of local heterogeneity

neuronal populations. Yet biological annotations for network nodes — Annotations and neural wiring
such as gene expression, cytoarchitecture, neurotransmitter receptors Annotations and network
or intrinsic dynamics — can be readily measured and overlaid on architecture

network models. Here we review how connectomes can be represented Annotation-enhanced models
of the brain
and analysed as annotated networks. Annotated connectomes allow
us to reconceptualize architectural features of networks and to relate Geometry, topology and
annotations
the connection patterns of brain regions to their underlying biology.
Emerging work demonstrates that annotated connectomes help to make Annotation similarity networks

more veridical models of brain network formation, neural dynamics and Concluding remarks
disease propagation. Finally, annotations can be used to infer entirely
new inter-regional relationships and to construct new types of network
that complement existing connectome representations. In summary,
biologically annotated connectomes offer a compelling way to study
neural wiring in concert with local biological features.

Montréal Neurological Institute, McGill University, Montréal, Quebec, Canada. e-mail: bratislav.misic@mcgill.ca

Nature Reviews Neuroscience

Review article

Introduction architectural features suggests conserved and universal wiring

Advances in imaging and tracing technology are yielding increasingly principles of brain connectomes.
detailed wiring diagrams of neural systems1. These wiring diagrams, However, connectomes are typically represented and ana-
termed connectomes, encode connection patterns among neural lysed as graphs with identical nodes3, abstracting away important
elements2. Over the past 15 years, the dominant paradigm has been micro-architectural features (including cytoarchitecture, neurotrans-
to represent connectomes as graphs, in which nodes represent single mitter receptor profiles and laminar differentiation) that may drive
neurons or neuronal populations and edges represent anatomical macroscale structure–function coupling30–32. How does macroscale
projections or functional interactions3. wiring between neuronal populations relate to their microscale fea-
The graph model has become the lingua franca of network neuro- tures? Concomitant advances in imaging technology and data science
science, helping to quantify architectural features of brain networks4,5. allow us to generate and access high-resolution maps of the brain
These include regionally specific connection profiles6–8, a tendency for across multiple modalities33–35. By superimposing these maps of bio-
functionally similar areas to form spatially contiguous communities9 logical annotations onto connectomes, we can generate a new type of
and a small set of disproportionately well-connected hub nodes that data object in neuroscience: annotated connectomes.
promote signal integration10–13. These features have been observed Here we review the nascent field of annotated connectomics. We
across multiple spatial scales and reconstruction techniques and in argue that connectome representations enriched with biological fea-
multiple model organisms14,15, including invertebrate16–19, avian20, tures open fundamentally new avenues for quantifying and articulating
rodent21–23, feline24–26 and primate species27–29. The ubiquity of these brain network organization. We review new findings, as well as classical
and emerging theories, of how local annotations relate to neural wiring.
We then explore methodological and conceptual advances, including
Box 1 adapting network theoretic measures for annotated networks, incorpo-
rating annotations in dynamic models of network formation, network
function and disease propagation, and disentangling the relationships
Annotating network edges between annotations, brain network topology and geometry. Finally,
we outline how entirely new classes of connectivity modes can be
Throughout this Review, we focus primarily on annotating the reconstructed using biological annotations. Throughout, we focus
nodes of the network. However, brain connectomes can also on a singular theme: what new questions about brain organization
have annotated edges. Edges capture pairwise relationships can be addressed by jointly considering connectome architecture and
between brain regions and are usually weighted by the strength biological annotations?
of the relationship between the two brain regions. For structural
connectomes, edges are generally weighted by the number or Embracing regional heterogeneity
density of streamlines between brain regions. The conventional workflow in connectomics is to parcellate the brain
Similar to how nodes can be annotated by local biological into a set of discrete regions and to focus on their connection patterns.
or micro-architectural properties, edges too can be annotated Multiple subfields have emerged in loose-knit tandem, focusing on
with features besides connection strength215,216. For example, quantifying architectural features using network theoretic methods36,
it is possible to annotate edges based on neurite density217, identifying individuals using connection patterns37–39, relating connec-
axon diameter distribution218,219, myelin content220 or the g-ratio tion patterns to cognition and behaviour40,41, isolating connectivity
(ratio of inner and outer diameters of myelinated axons)221,222. markers of disease42 and mapping connection patterns across phylog-
These techniques theoretically reflect the velocity and fidelity with eny and ontogeny8,43. Implicit in all those approaches is the idea that
which electrical impulses propagate via anatomical projections, brain regions are identical, insofar as none of these methods explicitly
although validation efforts are ongoing223. In some instances, takes differences in node biology into account.
physical projection lengths may also serve as edge annotations, Yet, high-resolution maps of the brain can be readily generated
such as when modelling the speed of diffusion of pathogenic using standard protocols in a wide array of histological, imaging and
molecules in neurodegenerative diseases178. In model organisms, recording technologies, such as microarray and single-cell sequenc-
tools such as radioactive tracers, genetic labelling and electron ing, microscopy, autoradiography, magnetic resonance imaging
microscopy can be used to comprehensively reconstruct finer (MRI), electroencephalography (EEG) and magnetoencephalography
properties of individual axonal projections and synaptic contacts224. (MEG), and positron-emission tomography (PET). These maps chart the
Finally, edges can be annotated with directionality, such that each spatial distribution of multiple structural and functional attributes,
edge represents a direction of information flow. Directionality, including gene transcription44,45, neurotransmitter receptors and
sometimes referred to as ‘effective connectivity’, can be inferred transporters46–51, cell density52, cell types53,54 and morphology55, laminar
using model-based techniques225, empirical estimation of temporal differentiation56,57, myelination58–61, grey matter morphometry62,63,
precedence226,227 and asymmetry of network-theoretic statistics228. evolutionary and developmental expansion64–67, metabolism68,69,
The concepts outlined in the present Review can be readily and intrinsic electromagnetic dynamics and haemodynamics70–75.
extended to annotated edges. Edge annotations — whether Network nodes can also be annotated using meta-analytic features,
biological, physical or directional — can enrich network such as propensity to activate during particular tasks76–78 or disease
reconstructions, leading to more biologically refined estimates of vulnerability79–81. In network neuroscience, nodes are also traditionally
network features such as hubs and modules (Fig. 2). Connectomes annotated according to their affiliations with larger discrete classes
with annotated edges can also be seamlessly integrated into (for example, intrinsic networks82 or cortical types83–85). Finally, we
models of communication and disease propagation (Fig. 3). note that network edges can analogously be annotated with biological
metadata (Box 1). Such maps are increasingly shared via open-access

Nature Reviews Neuroscience

Review article

Cell types Gene transcription Metabolism Receptors

for example, for example, for example, for example,
astrocytes PVALB CRMO2 5HT1A

Synapse density Neural dynamics Myelination Laminar thickness

for example, for example, for example, for example,
SV2A MEG alpha power T1w/T2w ratio layer IV

Fig. 1 | Constructing an annotated connectome. Top row: annotated (for example, PET-derived cerebral metabolic rate for oxygen (CMRO2)68),
connectomes are built by integrating a brain map representing some biological neurotransmitter receptors (for example, PET-derived 5HT1A receptor density47),
feature (‘annotation’, depicted by the surface map), and a brain network synapse density (for example, PET-derived synaptic vesicle glycoprotein 2A density
representing brain regions and their connection patterns (depicted by the grey (SV2A)212), neural dynamics (for example, magnetoencephalography (MEG)-
nodes and edges; top left panel). Vertex or voxel values in the biological map are derived 8–12 Hz alpha band power spectral density), myelination (for example,
parcellated according to the same atlas used to define the nodes of the connectome. MRI-derived T1w/T2w ratio) and laminar thickness (for example, layer IV thickness,
Annotation values for each region are estimated by calculating a summary measure derived by segmenting cortical layers in the Merker-stained BigBrain histological
(for example, mean, median or first principal component; top centre panel). The atlas52,56). Supplementary Table 1 lists resources for accessing annotations. The
resulting object is a connectome in which nodes are enriched with weights that figure was generated using open neuroimaging data in the neuromaps toolbox
reflect a specific biological phenomenon — the annotated connectome (top right and open gene expression data from the Allen Human Brain Atlas35,44. Data from
panel). Centre and bottom rows: brain networks can be annotated with a wide neuromaps were downloaded, parcellated to 800 cortical regions according to
range of biological metadata capturing, for instance, regional variations in cell the Schaefer atlas213 and then plotted. Gene expression data from the Allen Human
types (for example, astrocyte-specific gene expression54), gene transcription (for Brain Atlas were processed using default parameters in the abagen toolbox88,
example, PVALB, a proxy for parvalbumin-expressing interneurons), metabolism parcellated to the same 800 cortical regions and plotted.

repositories and toolboxes, with a concerted effort to standardize subdivided into parcels according to the same atlas that was used to
coordinate systems35,86,87 and workflows88–91. reconstruct the connectome (Fig. 1). The most common approach is
How do we use these datasets to build an annotated connec- to take the mean value, but alternative methods, such as the median
tome? Once you have a biological map, annotating a connectome or the first principal component, are also possible. If annotations
is straightforward. Vertex or voxel values in the biological map are are not densely sampled across the surface, such as in the case of

Nature Reviews Neuroscience

 Review article

a Homophily b Hubs c Communities d Paths e Motifs

Annotation values

Annotation values

Annotation values

Annotation values

Frequency
...

Annotation values Centrality Communities Hops Motifs

Fig. 2 | Architectural features of annotated connectomes. Biological features values for the nodes in each community, and are coloured according to the
add further depth to conventional graph theoretic measures. In all panels, nodes mean annotation value in each community. d, Top: annotating communication
represent brain regions, edges represent connections between regions and pathways in brain networks sheds light on how signals traverse diverse neuronal
colours represent the value of a biological annotation for each node. a, Top: the populations and how they are transformed en route from source to target. The
network displays homophily, whereby nodes with similar annotations (colours) figure highlights two paths in the network and the annotations (colours) of
have a higher tendency to be connected. Bottom: homophilic attachment the nodes in each path. Bottom: variations in annotation values are displayed as
can be estimated using assortativity — the correlation between annotation a function of hops (steps) along each path. Insets show the isolated paths from
values of connected nodes. b, Top: relating node centrality (how connected the network that correspond to each line. e, Top: annotated networks can be
or influential a node is) with biological features provides insight into the used to investigate how network motifs (for example, closed triangles versus
neural circuit attributes that support information integration. Bottom: node open triangles, highlighted in the figure) are supported by different types of
centrality is correlated with annotation values, suggesting that well-connected biological annotations (colours). Bottom: frequency of occurrence (count)
hub nodes have different biological features compared with non-hub nodes. of different motifs stratified by the annotation values of their constituent
c, Top: the network has prominent communities with distinct biological nodes. Scatter plots were generated using synthetic data. All other plots
annotations (colours). Bottom: boxplots show the distributions of annotation are conceptual illustrations.

microarray probes for gene transcription44 or intracranial electrodes nodes108 (Fig. 2a). Although assortativity has been primarily used to
in electrocorticography73, it is necessary to ensure that discrete sam- assess the tendency for hub nodes to be connected with each other
ples are matched to the appropriate parcel88,92. Ideally, the parcel (degree assortativity)109, the concept is more general and can be used
boundaries should be aligned with spatial variation in annotations to to quantify the homophily of any biological annotation110.
optimize within-parcel homogeneity and between-parcel differentia- The relationship between annotations and connectivity is conven-
tion. By ‘colouring’ network nodes with biological annotations, we now tionally conceptualized as homophilic mixing. An underexplored direc-
have an enriched representation of the brain that opens fundamentally tion is to consider heterophilic mixing, that is, whether regions with
new opportunities for discovery93,94. different classes or types of annotations tend to be connected110. For
example, do populations enriched in a particular cell type have a higher
Annotations and neural wiring tendency to be connected to neural populations enriched in a different
Once a biological feature map has been parcellated and used to anno- cell type, or could populations that express a particular neurotransmit-
tate a brain network, it is natural to ask how the annotations of two ter receptor connect to populations that express a different neuro-
brain regions are related to the connectivity between them. Numerous transmitter receptor? Indeed, connectivity is highly structured across
studies have explored whether brain regions exhibit homophilic mix- cortical layers, with heterogeneous patterns of connectivity between
ing: areas that are connected being more similar than areas that are not brain regions with specific laminar and cytoarchitectonic features111–114.
connected. This simple wiring principle appears to be ubiquitous across Similarly, interactions between slow-acting and fast-acting neuro-
a wide range of attributes including gene expression95–97, cytoarchitec- transmitter systems coordinate functions across timescales, from
ture (for example, laminar differentiation, neuron density and morphol- arousal to goal-directed behaviour115. However, connectivity-mediated
ogy, intracortical myelin)57,61,98–100, morphometry in healthy and clinical mixing among receptor types is not fully understood. Quantitative
populations101–103, neurotransmitter receptors49,104 and local neural investigation of heterophilic mixing between biological annotations
dynamics71,74,105. Collectively, these results may reflect the influence of is an exciting new frontier for understanding complex interactions
molecular and genomic gradients that guide axon formation during between cytoarchitecture, chemoarchitecture and cortico-cortical
development, resulting in connectivity among neural populations with connectivity110.
similar attributes32,106,107. In practice, homophilic mixing can be readily One salient example of how neural wiring can demonstrate both
quantified using the network-theoretic statistic of assortativity: the homophilic and heterophilic mixing comes from neuroanatomical
correlation between the annotation values across pairs of connected tract-tracing studies in non-human primates. These studies have

Nature Reviews Neuroscience

Review article

examined the cellular architecture of source and target cortical regions enriched for micro-architectural features. The goal of these analy-
of neural projections to develop a theory of neural wiring called the ses is to identify micro-architectural features that support the func-
Structural Model32,112,116. Interestingly, they observed heterophilic tional characteristics of each module. For instance, intrinsic networks
mixing whereby feedback connections from poorly differentiated to tend to have distinct transcriptional137 and metabolic128 signatures.
highly differentiated regions originate in deep layers (layers V and VI) These differences manifest in the cytoarchitecture;138 for example,
and terminate in primarily upper layers (layer I). Likewise, feedforward unimodal networks have greater intracortical myelin than transmodal
connections from highly differentiated regions to poorly differenti- networks139–141, perhaps providing the capacity to transmit signals with
ated regions originate in upper layer III and terminate in middle layers greater speed and fidelity. An alternative approach is to directly use
(deep layer III to upper layer V)32. These studies have also observed biological annotations in conjunction with connectivity patterns to
homophilic mixing whereby regions with similar laminar differentia- guide the detection of communities. The conventional community
tion were connected by neural projections that originate and terminate detection methods can be extended to identify groups of nodes that
in all layers of the cortex117. In other words, neural projection patterns not only share dense connectivity but also have similar biological
depend on whether the connection is homophilic or heterophilic make-up. To our knowledge, this direction has been less explored in
(from the perspective of laminar architecture). Biologically annotated network neuroscience, but principled methods already exist in network
connectomes make it possible to extend these theories to the whole science and statistical physics, including annotated stochastic block
human brain, as well as to features other than lamination. models93,142 and multilayer community detection143.
We close this section by considering additional biologically mean-
Annotations and network architecture ingful architectural features of connectomes that could be enriched
Moving beyond statistical associations between wiring and annota- with biological annotations, including communication pathways,
tions, we next turn our attention to network features that have been motifs, cliques and cavities. Measures of network communication are
extensively studied in connectomics3,4,36. Hallmark features, such as frequently studied to infer node centrality, inter-regional relationships
the tendency for nodes to form specialized modules or communi- and the global capacity of brain networks to process information.
ties, the existence of highly interconnected hubs, and the commu- These methods typically involve tracing out sequences of connected
nication pathways that support integration, can be recast from the nodes along which signals propagate. How do signals traverse neural
perspective of biological annotations. Here we consider how concepts circuits and what types of neuronal populations do they encounter
from graph theory can be enriched by incorporating information about along the way (Fig. 2d)? By annotating the sequence of nodes in a
biological attributes. communication path using biological attributes, it is possible to infer
Traditionally, hubs are defined as brain regions that are dispro- how diverse neuronal populations exchange signals144. In a similar
portionately well connected with the rest of the brain10 (Fig. 2b). They way, network motifs — small 2-node, 3-node or n-node subgraphs
are considered to be functionally important and have occupied the that constitute the building blocks of a network145 — can potentially
attention of the field for more than a decade12. Hubs display charac- be annotated to understand whether specific circuit configurations
teristic activity patterns, dominated by slow fluctuation and high align with specific micro-architectural features (Fig. 2e). For instance,
autocorrelation105,118,119, consistent with the notion that regions with numerous studies have reported that chains of reciprocally connected
many inputs integrate information over longer timescales and across nodes are over-represented in brain networks across multiple species,
sensory modalities70–72. What are the biological hallmarks of these scales and reconstruction techniques29,146,147, potentially acting as a sub-
neural populations that make them suited for integrating and dis- strate for functional integration and recurrent processing. High-order
seminating signals? At the microscale, hub regions are composed of structures, such as cliques and cavities, are also thought to reflect the
neurons with larger dendritic trees and more dendritic spines, pre- capacity for neural circuits to engage in parallel versus serial informa-
sumably to enhance integrative capacity55,99,120. This is reflected in tion processing29,148,149. A salient question for future research is whether
their transcriptomic profile: hub regions show greater transcription the formation and arrangement of these higher-order network features
for genes related to dendrite and synapse development121. Hubs also is guided by their underlying microscale architecture.
exhibit more differentiated laminar organization and larger layer III
neurons, which are thought to support long-range cortico-cortical Annotation-enhanced models of the brain
communication122. The energetic cost of this complex architecture So far, we have focused on the static annotation of brain networks and
is well documented: hubs tend to be more metabolically active, with on comparing network organization with annotations. But annotations
greater transcription of metabolic-related genes96, greater cerebral can also be used to inform more dynamic models of brain network
blood flow123–125, greater glucose metabolism55,69 and greater aerobic formation, communication and disease propagation. As we outline
glycolysis126–128. Perhaps as a consequence, hub nodes also appear to below, a consistent thread in these studies is that imparting biological
be more vulnerable to particular neurological disorders, displaying heterogeneity on regional nodes yields more veridical models of brain
greater cortical abnormalities129,130, greater accumulation of misfolded structure and function.
proteins131 and greater expression of disease-specific risk genes132. A key question in systems neuroscience is how structural con-
Another well-studied feature of the connectome is community nectivity (white matter projections between neuronal populations)
structure: the tendency for regions to form strongly interconnected is related to functional connectivity (co-fluctuations in neural activ-
communities (or ‘modules’) that support specialized processing82,133–136 ity)30. Early studies focused on statistical and dynamical models to
(Fig. 2c). Modules are conventionally defined on the basis of connectiv- relate the two, but they typically assumed that structure–function
ity, but they can be readily re-imagined from the perspective of brain coupling was uniform across the brain. Yet recent reports consistently
annotations. One way is to use established modules (for example, find that structure–function coupling is regionally heterogeneous,
intrinsic resting-state networks) and superimpose maps of annotations with stronger coupling in unimodal cortex and weaker coupling in
to ask whether specific modules (for example, the visual network) are transmodal cortex150–154. In a similar vein, biophysical models in which

Nature Reviews Neuroscience

Review article

a Biophysical models
Homogeneous model Heterogeneous model

Predicted FC

Predicted FC
Empirical FC Empirical FC

b Generative models

Annotation-guided progressive placement of edges

c Disease models Empirical atrophy

Misfolded protein

Annotation-guided Predicted
propagation along atrophy
structural connection

Fig. 3 | Annotation-enhanced models of the brain. Adding biological network formation. Annotations can be used to guide the progressive placement
heterogeneity to brain regions improves structural and functional models of of structural edges according to biological similarity between brain regions
the brain. a, Biophysical models of the brain fit local microcircuit parameters (indicated by nodes of a similar colour)174. c, Annotated networks can be used to
at every brain region to predict regional functional dynamics. Models are often model the spread of disease. Misfolded proteins are initially located in one or
evaluated according to how well they can generate patterns of co-activation more brain regions. Misfolded proteins then diffuse via structural connections,
(predicted functional connectivity (FC)) that resemble experimentally measured but their propagation is guided by biological annotations (indicated by node
patterns of co-activation (empirical FC). Circuit models that tune local parameters colour). For example, misfolded proteins may spread preferentially to regions
according to heterogeneous biological annotations (indicated by the different with high expression of a specific gene178. The aggregation of misfolded proteins
coloured nodes, right) predict functional connectivity between all pairs of brain causes cell death and atrophy. Model fit is evaluated by comparing predicted
regions better than models that retain default homogeneous local parameters atrophy patterns with empirically measured atrophy patterns. No data were used
(left)141,155. b, Generative models seek to identify the principles that govern to generate this figure.

regional dynamics are allowed to vary across cortex show an identical example, Demirtaş and colleagues enriched a biophysical neural
unimodal–transmodal gradient of dynamical parameters, represent- field model using a map of intracortical myelin141,164,165. They fitted
ing regional differences in excitation–inhibition ratio and subcortical two classes of model: a homogeneous model in which all node
input155. Collectively, these results open the possibility that taking into parameters are identical and a heterogeneous model in which dif-
account regional heterogeneity can help us to build better models of ferent nodes may take on different local microcircuit characteristic
brain function. (excitatory-to-inhibitory strength and recurrent excitatory strength)
Indeed, several recent studies have shown that models that that were scaled according to the intracortical myelin map. The authors
combine structural connectivity and local biological annotations found that the heterogeneous model better predicted resting-state
generate more accurate predictions about functional dynamics156, blood oxygen level-dependent (BOLD) functional connectivity and
including models enriched with intracortical myelin141,157, neurotrans- MEG-estimated power spectral density, demonstrating the utility of
mitter receptors49,158–162 and gene expression163 (Fig. 3a). As a practical annotation-informed dynamical models.

Nature Reviews Neuroscience

Review article

A parallel literature has emerged in which biological annotations overlapping neural circuits184,185. Therefore, areas that are physically
are used to guide generative models of network formation (Fig. 3b). close together have a higher tendency to be connected with each other,
In these models, the goal is to recapitulate the architecture and devel- and the strength of these connections is probably greater186,187. At the
opment of the connectome by placing nodes and/or edges according same time, areas that are close together also have a higher tendency to
to simple rules166, such as cost minimization (placing connections share similar biological features, such as gene transcription, cell types
between spatially proximal regions) or the optimization of some topo- and intrinsic electrophysiological rhythms. In other words, spatial
logical feature, such as clustering or the degree distribution167–169. These proximity influences both annotations and connectivity. As a result,
models can be used to study sequences of anatomical changes over phy- inferential procedures that take spatial autocorrelation into account
logeny or ontogeny, allowing one to test theories about how network are necessary to disentangle relationships between connectivity,
formation and arealization interact over time166,170. Earlier empirical annotations and geometry.
work has shown that the existence of connections can be predicted Fortunately, this has been an active topic of research, and multi-
from correlated gene expression or cellular type26,99,171–173. A natural next ple methods have been developed that randomize either the spatial
step is to build formal generative models in which edge placement is layout of annotations or wiring topology while controlling for spatial
governed by biological rules in addition to spatial (cost-minimizing) autocorrelation139,188. For almost any type of analysis, such as correlat-
and topological rules. For instance, a recent report explored an ensem- ing biological annotations and network features, spatial null models
ble of generative models in which the placement of connections is allow us to build a distribution of statistical parameters (for example,
also governed by similarity of gene expression, intracortical myelin or correlation coefficients between randomized annotations and network
laminar differentiation174. However, there still exist many annotations, features) that embody the null hypothesis that any observed relationship
such as metabolism, chemoarchitecture or cell types, whose roles between annotations and connectivity is because of the passive influence
in network formation have not been directly studied. Furthermore, of spatial autocorrelation. A popular set of methods for randomizing
generative models typically embody homophilic mixing rules, in which annotation maps are non-parametric ‘spin tests’ that project annota-
edge placement is more probable between brain regions with similar tions to a sphere, apply random angular rotations and bring annotation
annotations. The same methodology can also be readily implemented values back onto the cortical surface189. An alternative family of methods
to test the contribution of heterophilic mixing rules. are parameterized models that estimate specific statistical features of
Just as biological features shape the formation of networks and the
emergence of functional dynamics, they also impart local vulnerability
to pathology. As a result, annotation-enriched connectomes have seen Glossary
substantial use in disease modelling. Accumulating evidence has dem-
onstrated that multiple syndromes, particularly neurodegenerative
diseases, are caused by trans-synaptic propagation and aggregation of Annotations Multilayer community
misfolded proteins42,175. Although the specific proteins differ between Any measurements or features that can detection
syndromes (tau in Alzheimer disease, α-synuclein in Parkinson disease be attached to the nodes of a network, Techniques for identifying communities
and TDP-43 in fronto-temporal dementia), the spreading principle is either as a unidimensional scalar or that simultaneously take into account
the same. The propagation of pathology is, therefore, typically mod- multidimensional vector. multiple types of connectivity or other
elled as a guided diffusion process on white matter connectomes interactions between nodes, such as
(Fig. 3c). A common strategy has been to use regional differences in Arealization annotation similarity.
the transcription of genes associated with disease-specific proteins to The developmental process by which
scale the synthesis and clearance of proteins in silico176,177. For example, cortical cell types and circuits come to Stochastic block models
models of Parkinson’s disease or other synucleinopathies in which support unique specialized functions. Statistical models of network
α-synuclein synthesis and clearance are scaled by the expression of organization that formally take into
disease-related genes (SNCA and GBA) better predict atrophy pat- Cliques and cavities account both connection patterns and
terns compared with a homogeneous model in which synthesis and Cliques are groups of all-to-all local node annotations.
clearance are uniform across the brain178–180. Taking this approach connected nodes, and cavities are
further, separate dynamic models of disease propagation can be built groups of mutually unconnected nodes Transmodal networks
on connectomes annotated with expression of risk genes associated that participate in cliques. Networks that respond to multiple
with specific subgroups of a particular syndrome, such as in genetic sensory modalities and specialize for
variants of fronto-temporal dementia103. More generally, biological Community higher-order cognitive function, such as
annotations may help to model local vulnerability and explain the A group of nodes densely connected the salience and default networks.
spatial patterning of many other brain diseases or disorders that do with each other but sparsely connected
not necessarily involve misfolded proteins but may involve a different with the rest of the network. Unimodal networks
network-mediated mechanism, such as spreading excitotoxicity or the Networks that specialize for one primary
propagation of aberrant signals or trophic factors80,181,182. Connection profiles sensory or motor function, such as the
Vectors describing the connectivity visual and somatomotor networks.
Geometry, topology and annotations of a brain region, detailing all of its
How do we test whether a biological annotation is correlated with a pairwise connections with other Vertex or voxel values
network feature? The principal challenge is to disambiguate relation- brain regions. The main units of brain images,
ships between annotations and connectivity from the background representing a spatial location in the
effect of spatial proximity183. Guided by genomic gradients, the gradual Hub brain, defined either for a brain volume
process of axonal growth and synapse formation results in a series of A region that has many connections. (voxel) or on the surface (vertex).

Nature Reviews Neuroscience

 Review article

a Participant similarity b Time-series similarity c Multidimensional feature similarity

(for example, anatomical covariance) (for example, functional connectivity) (for example, correlated gene expression)

Expression
Time Genes
Participants
Regions

Regions

Regions

Regions Regions Regions

Increasing similarity

Fig. 4 | Annotation similarity networks. New networks of the brain can be and correlated across multiple time points. Bottom: the resulting annotation
derived by quantifying annotation similarity between pairs of brain regions. similarity network represents inter-regional synchronization between the
Blue and orange nodes indicate an example pair of brain regions; all other two nodes (for example, functional connectivity). c, Top: multidimensional
regions are shown as grey dots. Orange and blue arrowheads in the similarity annotations (for example, gene expression measurements for many genes)
matrix mark the row and column that correspond to the orange and blue nodes, are measured and correlated. Bottom: the resulting annotation similarity
respectively. The matrix entry at their intersection represents the correlation, or network represents multidimensional feature similarity between pairs of nodes
annotation similarity, between the measurements made at the orange and blue (for example, correlated gene expression200). Panel a was generated using open
nodes. a, Top: a single annotation (for example, cortical thickness) is measured MRI-derived cortical thickness fsLR-32k maps for 48 random participants in the
and correlated across multiple individuals. Bottom: the resulting annotation Human Connectome Project S1200 data release211. Images were downloaded,
similarity network represents how pairs of brain regions are correlated across parcellated to 800 cortical regions according to the Schaefer atlas213 and
participants (for example, anatomical covariance network196). b, Top: a single correlated (Pearson’s r) to generate the region × region correlation matrix. Heat
annotation (for example, blood oxygenation from functional MRI214) is measured maps in panels b and c were generated by plotting networks published in ref. 204.

annotation maps, such as the variogram, and generate random annota- We close this section by emphasizing that the effect of spatial
tion maps that approximately match those features59,190,191. Although proximity should not be treated as a confound but as a natural and fun-
spin-based nulls can be applied only to cortical surfaces, parameter- damental feature of this spatially embedded system. Indeed, many ana-
ized nulls are more versatile and can be applied to cortical volumes, as lytic procedures in network neuroscience explicitly take into account
well as extracortical structures, such as the subcortex or cerebellum. geometry. For example, state-of-the-art generative models of brain net-
Conversely, methods that randomize connectivity typically build on con- work formation include a geometric term that promotes the placement
ventional edge randomization algorithms, such that edges are randomly of connections among spatially proximal populations167,168,170. Similarly,
swapped if and only if the resulting swaps approximately preserve the in dynamic models of communication or disease propagation, spread-
edge length distribution192. Collectively, spatial null models test whether ing agents (signals or misfolded proteins) are endowed with velocities
an observed relationship between biological annotations and network defined by the lengths of anatomical connections178,193,194. Alternatively,
features exists above and beyond the effects of spatial proximity. analyses that do not explicitly take into account geometry can be

Nature Reviews Neuroscience

Review article

readily reformulated to do so. For instance, homophilic and hetero- measurement that is measured across individuals (Fig. 4a). A second,
philic attachment can be assessed with respect to permuted annotation and probably the most common, conceptualization of annotation
maps to benchmark the extent to which these principles extend above similarity is ‘functional connectivity’, that is, how annotations of neu-
and beyond the effect of spatial proximity110. ral activity are synchronized over time. This definition is most often
applied to the BOLD signal from functional MRI (fMRI) but can also
Annotation similarity networks be applied to other time courses such as glucose metabolism from
Although the connectome is considered the full set of network elements fluorodeoxyglucose-PET (FDG-PET)197,198 and electromagnetic neural
and connections in the brain2, there is an alternative perspective of activity from EEG and MEG199 (Fig. 4b). Finally, a third form of annota-
brain connectivity in which the definition is not based on structural tion similarity has emerged, whereby multiple measurements of an
connectivity: annotation similarity. In this case, a network edge is not annotation are collected at each brain region, and these regional finger­
physical but rather represents statistical associations between brain prints are correlated (Fig. 4c). These annotation similarity networks
regions with respect to some annotation (for example, cortical thick- have been constructed for phenotypes such as genes (correlated gene
ness, BOLD signal and gene transcription) across some independent expression200), receptors (neurotransmitter receptor similarity49) and
variable (for example, participants, time and genes). In other words, cell types (laminar similarity57,100,201), and tell us how regions share
instead of annotating brain regions using micro-architectural or molec- similar biological building blocks. Interestingly, regardless of the anno-
ular features, local annotations are used to construct an entirely new tation or data type, annotation similarity consistently decreases with
type of network (Fig. 4). Annotation similarity is generally represented distance, is larger for structurally connected regions and is symmetric
as a brain region × brain region matrix in which elements are the chosen across hemispheres, suggesting common organizational principles in
statistical measure of similarity. the brain71,114,200,202.
Every version of annotation similarity provides a new perspec- With annotation similarity networks, we can ask new questions
tive on the relationships between brain regions, while preserving about how brain circuitry emerges from a confluence of local proper-
a set of ubiquitous organizational principles. Perhaps the earliest ties. For example, the rich club of the brain — densely interconnected
representation of an annotation similarity network is metabolic con- brain regions that are thought to support signal integration —
nectivity, derived by correlating regional glucose metabolism across was originally viewed only from the perspective of the structural
participants195. This type of annotation similarity — that is, a network connectome203. Recent studies have found that structural rich club
that tells us how brain regions are coupled across individuals with links exist between brain regions with similar gene expression and
respect to a specific phenotype — has also been applied to cortical receptor density profiles96,200,204. By defining edges on the basis of
thickness (anatomical covariance196) and can be extended to any information about local biological similarities, annotation similarity

Box 2

From population to individual annotated connectomes

Current work in annotated connectomics typically uses datasets the brain is divided into brain regions — can influence connectome
sampled from different populations. Sometimes, this is because reconstruction229,230, may not appropriately represent individual-
of the cost associated with acquiring data. At other times, it is specific biological annotations and can have downstream effects
because of the ethical considerations and/or invasiveness of the on analyses, including modelling cognition231 and disease232.
imaging or tracing technique. For instance, PET imaging involves For example, if parcel boundaries are defined functionally, these
injecting radioactive tracers, precluding comprehensive sampling parcels may be misaligned with histological boundaries, resulting
in the same individuals, whereas gene expression, cell staining and in parcels with non-homogeneous annotation measurements233.
autoradiography are measured post-mortem. Annotation-enriched Therefore, it is necessary to define nodes to reflect both the individual
connectomic analyses, therefore, involve comparing annotations and the annotation.
from groups that are not necessarily matched with respect to sample Going forward, the acquisition of highly sampled and multimodal
size, chronological age, biological sex or disease status. This induces datasets combining a wide range of imaging modalities in individual
statistical challenges and limits possible inferences and the extent to participants is highly desirable. Such deep phenotyping datasets
which findings can be extrapolated to other populations. For example, would capture not only the connectivity profiles of individual
disease phenotypes in clinical populations (for example, cortical brain regions but also their local genetic and cellular make-up,
thinning) are often referenced against connectomes and annotations and are in line with recent work aimed at better capturing
generated from healthy populations (for example, connectomes from individual variability234–236. Furthermore, precision imaging and
the Human Connectome Project and gene transcription from the Allen deep phenotyping can guide the development of personalized
Human Brain Atlas). Furthermore, annotations are frequently shared multimodal parcellations that take into account multiple levels of
and analysed as group averages, blurring individual differences. Thus, description of brain anatomy233–238. With the emergence of annotated
the diversity of a population should be explicitly taken into account connectomes that represent both individuals and multiple brain
when building annotated connectomes. phenotypes, we can ask new questions about how annotations and
An added challenge is how to align individual-specific annotations neural wiring interact throughout development, during ageing
with group-level parcellations. The choice of parcellation — how and in disease.

Nature Reviews Neuroscience

Review article

networks shed light on the biological mechanisms of macroscale imaging and deep phenotyping efforts will enhance our capacity to build
architectural features. Likewise, using the graph theoretic approaches individualized annotated connectomes (Box 2).
presented above (see ‘Annotations and network architecture’), In summary, biological annotations add a new dimension to brain
we can ask whether annotation similarity networks present novel connectomes. The resulting mathematical object — the annotated
architectural features not found in the more standard structural or connectome — unlocks an entirely new way of thinking about the inter-
fMRI-derived functional connectomes. Indeed, FDG-PET-derived digitation of global connectivity and local features. The datasets,
functional connectivity has revealed hubs of neural synchronization analytics and theory are all in place to start an exciting new chapter in
in frontoparietal regions: regions not traditionally considered hubs network neuroscience.
from the perspective of the classical structural and fMRI-derived
functional connectomes12,197. Additionally, network architecture of Data availability
morphometric similarity, an annotation similarity network in which Data used to generate Fig. 1 are available in the neuromaps toolbox
the annotations are MRI-derived white-matter and grey-matter (https://netneurolab.github.io/neuromaps/ (ref. 35)) and the Allen
measurements, has been shown to vary with individual differences Human Brain Atlas35,44. Synthetic data were used to generate Fig. 2.
in cognitive performance101. Figure 3 is a conceptual illustration and is not based on real data. Data
Ultimately, each annotation similarity network is only a single used to generate Fig. 4 can be found at https://github.com/netneurolab/
perspective of inter-regional relationships. The integrated, multiscale hansen_many_networks and the Human Connectome Project204,211.
nature of the brain is reflected by these annotation similarity net-
works, from shared molecular mechanisms to dynamics to function. Published online: xx xx xxxx
How these annotation similarity networks interact with one another
References
and with the structural backbone of the brain remains an exciting 1. DeWeerdt, S. How to map the brain. Nature 571, S6 (2019).
direction for future research. Indeed, with the emergence of new 2. Sporns, O., Tononi, G. & Kötter, R. The human connectome: a structural description of the
datasets, diverse annotation similarity networks are rapidly being human brain. PLoS Comput. Biol. 1, e42 (2005).
3. Bassett, D. S. & Sporns, O. Network neuroscience. Nat. Neurosci. 20, 353–364 (2017).
introduced57,71,80,205,206. This sets the stage for generating new inte- 4. Bullmore, E. & Sporns, O. Complex brain networks: graph theoretical analysis of
grative networks derived either by combining multiple annotation structural and functional systems. Nat. Rev. Neurosci. 10, 186–198 (2009).
similarity networks into a single fused network204 or by concatenating 5. Sporns, O. Network attributes for segregation and integration in the human brain.
Curr. Opin. Neurobiol. 23, 162–171 (2013).
multiscale annotations to create a single comprehensive annotation 6. Passingham, R. E., Stephan, K. E. & Kötter, R. The anatomical basis of functional
fingerprint at each node104,207. Thus, we envision that the future of localization in the cortex. Nat. Rev. Neurosci. 3, 606–616 (2002).
annotated connectomics involves a holistic, multiscale perspective 7. Mars, R. B. et al. Whole brain comparative anatomy using connectivity blueprints. eLife 7,
e35237 (2018).
of inter-regional relationships. 8. Mars, R. B., Passingham, R. E. & Jbabdi, S. Connectivity fingerprints: from areal
descriptions to abstract spaces. Trends Cogn. Sci. 22, 1026–1037 (2018).
Concluding remarks 9. Hilgetag, C. C. & Kaiser, M. Clustered organization of cortical connectivity.
Neuroinformatics 2, 353–360 (2004).
The present Review summarizes a single, consistent and altogether 10. Sporns, O., Honey, C. J. & Kötter, R. Identification and classification of hubs in brain
unsurprising principle: that including information about the underly- networks. PLoS ONE 2, e1049 (2007).
11. Van Den Heuvel, M. P., Kahn, R. S., Goñi, J. & Sporns, O. High-cost, high-capacity backbone
ing biology can help us to build more realistic connectome represen-
for global brain communication. Proc. Natl Acad. Sci. USA 109, 11372–11377 (2012).
tations. Annotations add depth and meaning to the roles of hallmark 12. Heuvel, M. P. & van den, Sporns, O. Network hubs in the human brain. Trends Cogn. Sci.
architectural features, such as hubs and modules. Annotations boost 17, 683–696 (2013).
13. Zamora-López, G., Zhou, C. & Kurths, J. Cortical hubs form a module for multisensory
virtually all models of brain network formation, neural dynamics and
integration on top of the hierarchy of cortical networks. Front. Neuroinform. 4, 1 (2010).
disease propagation. Finally, annotations can be used to recover new 14. Van den Heuvel, M. P., Bullmore, E. T. & Sporns, O. Comparative connectomics.
inter-regional relationships that complement existing structural and Trends Cogn. Sci. 20, 345–361 (2016).
15. Barsotti, E., Correia, A. & Cardona, A. Neural architectures in the light of comparative
functional networks. connectomics. Curr. Opin. Neurobiol. 71, 139–149 (2021).
It is with surpassing wonder that we consider how modern advances 16. White, J. G. et al. The structure of the nervous system of the nematode Caenorhabditis
in imaging and tracing technology will shape next-generation connec- elegans. Philos. Trans. R. Soc. Lond. B Biol. Sci. 314, 1–340 (1986).
17. Chiang, A. S. et al. Three-dimensional reconstruction of brain-wide wiring networks in
tomics. Going forward, we envision the incorporation of other bio- Drosophila at single-cell resolution. Curr. Biol. 21, 1–11 (2011).
logically meaningful annotations that can be reliably measured but have 18. Towlson, E. K., Vértes, P. E., Ahnert, S. E., Schafer, W. R. & Bullmore, E. T. The rich club
seen little uptake in connectomics. These include data on the vascular of the C. elegans neuronal connectome. J. Neurosci. 33, 6380–6387 (2013).
19. Worrell, J. C., Rumschlag, J., Betzel, R. F., Sporns, O. & Mišić, B. Optimized connectome
system (particularly measures of neurovascular coupling), spatially com- architecture for sensory-motor integration. Netw. Neurosci. 1, 415–430 (2017).
prehensive gene transcription and cell type deconvolution estimated 20. Shanahan, M., Bingman, V. P., Shimizu, T., Wild, M. & Güntürkün, O. Large-scale network
using single-cell RNA sequencing (as opposed to microarray), receptor organization in the avian forebrain: a connectivity matrix and theoretical analysis.
Front. Comput. Neurosci. 7, 89 (2013).
density estimated using autoradiography, synapse types, enzymes 21. Oh, S. W. et al. A mesoscale connectome of the mouse brain. Nature 508, 207–214
and more specific markers of pathology (for example, tau imaging) (2014).
or inflammation (for example, microglial activation). To complement 22. Bota, M., Sporns, O. & Swanson, L. W. Architecture of the cerebral cortical association
connectome underlying cognition. Proc. Natl Acad. Sci. USA 112, E2093–E2101 (2015).
annotation-enriched connectomic analyses in humans, experiments in 23. Rubinov, M., Ypma, R. J., Watson, C. & Bullmore, E. T. Wiring cost and topological
animal models — for which we have access to additional techniques and participation of the mouse brain connectome. Proc. Natl Acad. Sci. USA 112,
a wider variety of measurements — will generate a more complete set of 10032–10037 (2015).
24. Scannell, J. W., Blakemore, C. & Young, M. P. Analysis of connectivity in the cat cerebral
annotations, such as synaptomes208, metabolomes209 and direct assays cortex. J. Neurosci. 15, 1463–1483 (1995).
of pathology210. More broadly, whereas numerous studies have looked at 25. de Reus M. A. & van den Heuvel M. P. Rich club organization and intermodule
single annotations, the time is right for more comprehensive studies that communication in the cat connectome. J. Neurosci. 33, 12929–12939 (2013).
26. Beul, S. F., Grant, S. & Hilgetag, C. C. A predictive model of the cat cortical
simultaneously look at entire classes or even multiple classes of anno- connectome based on cytoarchitecture and distance. Brain Struct. Funct. 220,
tations in a single analytic framework110,204. Finally, modern precision 3167–3184 (2015).

Nature Reviews Neuroscience

Review article

27. Markov, N. T. et al. A weighted and directed interareal connectivity matrix for macaque 63. Bethlehem, R. A. et al. Brain charts for the human lifespan. Nature 604, 525–533
cerebral cortex. Cereb. Cortex 24, 17–36 (2012). (2022).
28. Majka, P. et al. Towards a comprehensive atlas of cortical connections in a primate brain: 64. Hill, J. et al. Similar patterns of cortical expansion during human development and
mapping tracer injection studies of the common marmoset into a reference digital evolution. Proc. Natl Acad. Sci. USA 107, 13135–13140 (2010).
template. J. Comp. Neurol. 524, 2161–2181 (2016). 65. Wei, Y. et al. Genetic mapping and evolutionary analysis of human-expanded cognitive
29. Liu, Z. Q., Zheng, Y. Q. & Misic, B. Network topology of the marmoset connectome. networks. Nat. Commun. 10, 4839 (2019).
Netw. Neurosci. 4, 1181–1196 (2020). This study shows that brain regions that are anatomically connected tend to have
30. Suárez, L. E., Markello, R. D., Betzel, R. F. & Misic, B. Linking structure and function in similar laminar differentiation.
macroscale brain networks. Trends Cogn. Sci. 24, 302–315 (2020). 66. Xu, T. et al. Cross-species functional alignment reveals evolutionary hierarchy within
31. Heuvel, M. P. & van den, Yeo, B. T. A spotlight on bridging microscale and macroscale the connectome. Neuroimage 223, 117346 (2020).
human brain. architecture. Neuron 93, 1248–1251 (2017). 67. Reardon, P. et al. Normative brain size variation and brain shape diversity in humans.
32. García-Cabezas, M. Á., Zikopoulos, B. & Barbas, H. The structural model: a theory linking Science 360, 1222–1227 (2018).
connections, plasticity, pathology, development and evolution of the cerebral cortex. 68. Vaishnavi, S. N. et al. Regional aerobic glycolysis in the human brain. Proc. Natl Acad.
Brain Struct. Funct. 224, 985–1008 (2019). Sci. USA 107, 17757–17762 (2010).
33. Van Essen, D. C. et al. The brain analysis library of spatial maps and atlases (BALSA) 69. Castrillon G. et al. An energy costly architecture of neuromodulators for human brain.
database. Neuroimage 144, 270–274 (2017). evolution and cognition. Preprint at bioRxiv https://doi.org/10.1101/2023.04.25.538209
34. Gorgolewski, K. J. et al. NeuroVault.org: a repository for sharing unthresholded statistical (2023).
maps, parcellations, and atlases of the human brain. Neuroimage 124, 1242–1244 (2016). 70. Murray, J. D. et al. A hierarchy of intrinsic timescales across primate cortex. Nat. Neurosci.
35. Markello R. D. et al. neuromaps: structural and functional interpretation of brain maps. 17, 1661–1663 (2014).
Nat. Methods. 19, 1472–1479 (2022). 71. Shafiei, G. et al. Topographic gradients of intrinsic dynamics across neocortex. eLife 9,
This paper accompanies a toolbox for accessing and analysing many different brain e62116 (2020).
annotations. 72. Gao, R., Brink, R. L., van den, Pfeffer, T. & Voytek, B. Neuronal timescales are functionally
36. Rubinov, M. & Sporns, O. Complex network measures of brain connectivity: uses and dynamic and shaped by cortical microarchitecture. eLife 9, e61277 (2020).
interpretations. Neuroimage 52, 1059–1069 (2010). 73. Frauscher, B. et al. Atlas of the normal intracranial electroencephalogram:
37. Finn, E. S. et al. Functional connectome fingerprinting: identifying individuals using neurophysiological awake activity in different cortical areas. Brain 141, 1130–1144 (2018).
patterns of brain connect. Nat. Neurosci. 18, 1664–1671 (2015). 74. Wang, X. J. Macroscopic gradients of synaptic excitation and inhibition in the neocortex.
38. Silva Castanheira, J., da, Orozco Perez, H. D., Misic, B. & Baillet, S. Brief segments of Nat. Rev. Neurosci. 21, 169–178 (2020).
neurophysiological activity enable individual differentiation. Nat. Commun. 12, 5713 (2021). This paper is a comprehensive review of how the brain can be topographically
39. Amico, E. & Goñi, J. The quest for identifiability in human functional connectomes. annotated using features of intrinsic dynamics.
Sci. Rep. 8, 8254 (2018). 75. Wang, B., Chen, Y., Chen, K., Lu, H. & Zhang, Z. From local properties to brain-wide
40. Marek, S. et al. Reproducible brain-wide association studies require thousands of organization: a review of intraregional temporal features in functional magnetic
individuals. Nature 603, 654–660 (2022). resonance imaging data. Hum. Brain Mapp. 44, 3926–3938 (2023).
41. Shen, X. et al. Using connectome-based predictive modeling to predict individual 76. Yarkoni, T., Poldrack, R. A., Nichols, T. E., Van Essen, D. C. & Wager, T. D. Large-scale
behavior from brain connectivity. Nat. Protoc. 12, 506–518 (2017). automated synthesis of human functional neuroimaging data. Nat. Methods 8, 665–670
42. Fornito, A., Zalesky, A. & Breakspear, M. The connectomics of brain disorders. Nat. Rev. (2011).
Neurosci. 16, 159–172 (2015). 77. Voytek, J. B. & Voytek, B. Automated cognome construction and semi-automated
43. Suarez, L. E. et al. A connectomics-based taxonomy of mammals. eLife 11, e78635 (2022). hypothesis generation. J. Neurosci. Methods 208, 92–100 (2012).
44. Hawrylycz, M. J. et al. An anatomically comprehensive atlas of the adult human brain 78. Dockès, J. et al. NeuroQuery, comprehensive meta-analysis of human brain mapping.
transcriptome. Nature 489, 391–399 (2012). eLife 9, e53385 (2020).
45. Akbarian, S. et al. The PsychENCODE project. Nat. Neurosci. 18, 1707–1712 (2015). 79. Thompson, P. M. et al. ENIGMA and global neuroscience: a decade of large-scale studies
46. Zilles, K. & Palomero-Gallagher, N. Multiple transmitter receptors in regions and layers of the brain in health and disease across more than 40 countries. Transl. Psychiatry 10,
of the human cerebral cortex. Front. Neuroanat. 11, 78 (2017). 100 (2020).
47. Beliveau, V. et al. A high-resolution in vivo atlas of the human brain’s serotonin system. 80. Hansen, J. Y. et al. Local molecular and global connectomic contributions to
J. Neurosci. 37, 120–128 (2017). cross-disorder cortical abnormalities. Nat. Commun. 13, 4682 (2022).
48. Nørgaard, M. et al. A high-resolution in vivo atlas of the human brain’s benzodiazepine 81. Larivière, S. et al. The ENIGMA toolbox: multiscale neural contextualization of multisite
binding site of GABAA receptors. Neuroimage 232, 117878 (2021). neuroimaging datasets. Nat. Methods 18, 698–700 (2021).
49. Hansen, J. Y. et al. Mapping neurotransmitter systems to the structural and functional 82. Yeo, B. T. et al. The organization of the human cerebral cortex estimated by intrinsic
organization of the human neocortex. Nat. Neurosci. 25, 1569–1581 (2022). functional connectivity. J. Neurophysiol. 106, 1125–1165 (2011).
50. Hansen, J. Y. et al. Correspondence between gene expression and neurotransmitter 83. Vértes, P. E. et al. Gene transcription profiles associated with inter-modular hubs and
receptor and transporter density in the human brain. Neuroimage 264, 119671 (2022). connection distance in human functional magnetic resonance imaging networks.
51. Froudist-Walsh, S. et al. Gradients of neurotransmitter receptor expression in the Philos. Trans. R. Soc. Lond. B Biol. Sci. 371, 20150362 (2016).
macaque cortex. Nat. Neurosci. 26, 1281–1294 (2023). 84. Scholtens, L. H., de Reus, M. A., deLange, S. C., Schmidt, R. & van den Heuvel, M. P.
52. Amunts, K. et al. BigBrain: an ultrahigh-resolution 3D human brain model. Science 340, An MRI von Economo–Koskinas atlas. Neuroimage 170, 249–256 (2018).
1472–1475 (2013). 85. Foit, N. A. et al. A whole-brain 3D myeloarchitectonic atlas: mapping the Vogt-Vogt
53. Vidal-Pineiro, D. et al. Cellular correlates of cortical thinning throughout the lifespan. legacy to the cortical surface. Neuroimage 263, 119617 (2022).
Sci. Rep. 10, 21803 (2020). 86. Ciric R. et al. TemplateFlow: FAIR-sharing of multi-scale, multi-species brain models.
54. Seidlitz, J. et al. Transcriptomic and cellular decoding of regional brain vulnerability Nat. Methods 19, 1568–1571 (2022).
to neurogenetic disorders. Nat. Commun. 11, 3358 (2020). 87. Schirner, M. et al. Brain simulation as a cloud service: the virtual brain on EBRAINS.
55. Scholtens, L. H., Schmidt, R., de Reus, M. A. & van den Heuvel, M. P. Linking macroscale Neuroimage 251, 118973 (2022).
graph analytical organization to microscale neuroarchitectonics in the macaque 88. Markello, R. D. et al. Standardizing workflows in imaging transcriptomics with the abagen
connectome. J. Neurosci. 34, 12192–12205 (2014). toolbox. eLife 10, e72129 (2021).
This paper demonstrates links between macroscale connectome features 89. Esteban, O. et al. fMRIPrep: a robust preprocessing pipeline for functional MRI.
(for example, number of connections) and micro-architecture (for example, Nat. Methods 16, 111–116 (2019).
dendritic spine count). 90. Cieslak, M. et al. QSIPrep: an integrative platform for preprocessing and reconstructing
56. Wagstyl, K. et al. BigBrain 3D atlas of cortical layers: cortical and laminar thickness diffusion MRI data. Nat. Methods 18, 775–778 (2021).
gradients diverge in sensory and motor cortices. PLoS Biol. 18, e3000678 (2020). 91. Adebimpe, A. et al. ASLPrep: a platform for processing of arterial spin labeled MRI and
57. Paquola, C. et al. Microstructural and functional gradients are increasingly dissociated quantification of regional brain perfusion. Nat. Methods 19, 683–686 (2022).
in transmodal cortices. PLoS Biol. 17, e3000284 (2019). 92. Arnatkeviciute, A., Fulcher, B. D. & Fornito, A. A practical guide to linking brain-wide gene
58. Glasser, M. F. & Van Essen, D. C. Mapping human cortical areas in vivo based on myelin expression and neuroimaging data. Neuroimage 189, 353–367 (2019).
content as revealed by T1- and T2-weighted MRI. J. Neurosci. 31, 11597–11616 (2011). 93. Khambhati, A. N., Sizemore, A. E., Betzel, R. F. & Bassett, D. S. Modeling and interpreting
59. Burt, J. B. et al. Hierarchy of transcriptomic specialization across human cortex captured mesoscale network dynamics. Neuroimage 180, 337–349 (2018).
by structural neuroimaging topography. Nat. Neurosci. 21, 1251–1259 (2018). 94. Lariviere, S. et al. Microstructure-informed connectomics: enriching large-scale
60. Whitaker, K. J. et al. Adolescence is associated with genomically patterned consolidation descriptions of healthy and diseased brains. Brain Connect. 9, 113–127 (2019).
of the hubs of the human brain. connectome. Proc. Natl Acad. Sci. USA 113, 9105–9110 95. French, L. & Pavlidis, P. Relationships between gene expression and brain wiring in the
(2016). adult rodent brain. PLoS Comput. Biol. 7, e1001049 (2011).
61. Huntenburg, J. M. et al. A systematic relationship between functional connectivity 96. Fulcher, B. D. & Fornito, A. A transcriptional signature of hub connectivity in the mouse
and intracortical myelin in the human cerebral cortex. Cereb. Cortex 27, 981–997 (2017). connectome. Proc. Natl Acad. Sci. USA 113, 1435–1440 (2016).
62. Dale, A. M., Fischl, B. & Sereno, M. I. Cortical surface-based analysis: I. Segmentation and This paper demonstrates that connectivity between hub regions is supported by gene
surface reconstruction. Neuroimage 9, 179–194 (1999). co-expression.

Nature Reviews Neuroscience

Review article

97. Arnatkeviciute, A., Fulcher, B. D., Pocock, R. & Fornito, A. Hub connectivity, neuronal 129. Crossley, N. A. et al. The hubs of the human connectome are generally implicated in the
diversity, and gene expression in the Caenorhabditis elegans connectome. anatomy of brain disorders. Brain 137, 2382–2395 (2014).
PLoS Comput. Biol. 14, e1005989 (2018). This study demonstrates that pathological grey matter lesions across multiple
98. Beul, S. F., Barbas, H. & Hilgetag, C. C. A predictive structural model of the primate disorders have a higher tendency to target hub regions of brain networks.
connectome. Sci. Rep. 7, 43176 (2017). 130. Seeley, W. W., Crawford, R. K., Zhou, J., Miller, B. L. & Greicius, M. D. Neurodegenerative
99. Beul, S. F. & Hilgetag, C. C. Neuron density fundamentally relates to architecture and diseases target large-scale human brain networks. Neuron 62, 42–52 (2009).
connectivity of the primate cerebral cortex. Neuroimage 189, 777–792 (2019). 131. Buckner, R. L. et al. Cortical hubs revealed by intrinsic functional connectivity: mapping,
100. Wei, Y., Scholtens, L. H., Turk, E. & Van Den Heuvel, M. P. Multiscale examination of assessment of stability, and relation to Alzheimer’s disease. J. Neurosci. 29, 1860–1873
cytoarchitectonic similarity and human brain connect. Netw. Neurosci. 3, 124–137 (2009).
(2018). 132. Romme, I. A., de Reus, M. A., Ophoff, R. A., Kahn, R. S. & van den Heuvel, M. P.
101. Seidlitz, J. et al. Morphometric similarity networks detect microscale cortical Connectome disconnectivity and cortical gene expression in patients with
organization and predict inter-individual cognitive variation. Neuron 97, 231–247 schizophrenia. Biol. Psychiatry 81, 495–502 (2017).
(2018). 133. Sporns, O. & Betzel, R. F. Modular brain networks. Annu. Rev. Psychol. 67, 613–640 (2016).
102. Shafiei, G. et al. Spatial patterning of tissue volume loss in schizophrenia reflects brain 134. Power, J. D. et al. Functional network organization of the human brain. Neuron 72,
network architecture. Biol. Psychiatry 87, 727–735 (2020). 665–678 (2011).
103. Shafiei, G. et al. Network structure and transcriptomic vulnerability shape atrophy in 135. Damoiseaux, J. S. et al. Consistent resting-state networks across healthy subjects.
frontotemporal dementia. Brain 146, 321–336 (2023). Proc. Natl Acad. Sci. USA 103, 13848–13853 (2006).
104. Dipasquale, O. et al. Receptor-enriched analysis of functional connectivity by targets 136. Bellec, P., Rosa-Neto, P., Lyttelton, O. C., Benali, H. & Evans, A. C. Multi-level bootstrap
(REACT): a novel, multimodal analytical approach informed by PET to study the analysis of stable clusters in resting-state fMRI. Neuroimage 51, 1126–1139 (2010).
pharmacodynamic response of the brain under MDMA. Neuroimage 195, 252–260 137. Richiardi, J. et al. Correlated gene expression supports synchronous activity in brain
(2019). networks. Science 348, 1241–1244 (2015).
105. Gollo, L. L., Zalesky, A., Hutchison, R. M., Van Den Heuvel, M. & Breakspear, M. Dwelling 138. Sydnor, V. J. et al. Neurodevelopment of the association cortices: patterns, mechanisms,
quietly in the rich club: brain network determinants of slow cortical fluctuations. and implications for psychopathology. Neuron 109, 2820–2846 (2021).
Philos. Trans. R. Soc. Lond. B Biol. Sci. 370, 20140165 (2015). This paper reviews brain annotations and their relationship with the unimodal–
106. Ramon y Cajal S. Studies on Vertebrate Neurogenesis (Blackwell Scientific, 1960). transmodal hierarchy.
107. Sperry, R. W. Chemoaffinity in the orderly growth of nerve fiber patterns and 139. Markello, R. D. & Misic, B. Comparing spatial null models for brain maps. Neuroimage
connections. Proc. Natl Acad. Sci. USA 50, 703–710 (1963). 236, 118052 (2021).
108. Newman, M. E. Mixing patterns in networks. Phys. Rev. E 67, 026126 (2003). 140. Paquola C. et al. The unique cytoarchitecture and wiring of the human default mode
109. Bassett, D. S. et al. Hierarchical organization of human cortical networks in health and network. Preprint at bioRxiv https://doi.org/10.1101/2021.11.22.469533 (2021).
schizophrenia. J. Neurosci. 28, 9239–9248 (2008). 141. Demirtaş, M. et al. Hierarchical heterogeneity across human cortex shapes large-scale
110. Bazinet, V. et al. Assortative mixing in micro-architecturally annotated brain neural dynamics. Neuron 101, 1181–1194 (2019).
connectomes. Nat. Commun. 14, 2850 (2023). This study shows that dynamical models based on connectomes annotated with
This paper formally studies the assortative mixing of multiple biological attributes intracortical myelin enhance prediction of neural dynamics.
across different species. 142. Newman, M. E. & Clauset, A. Structure and inference in annotated networks.
111. Rockland, K. S. & Pandya, D. N. Laminar origins and terminations of cortical connections Nat. Commun. 7, 11863 (2016).
of the occipital lobe in the rhesus monkey. Brain Res. 179, 3–20 (1979). 143. Mucha, P. J., Richardson, T., Macon, K., Porter, M. A. & Onnela, J. P. Community structure
112. Barbas, H. General cortical and special prefrontal connections: principles from structure in time-dependent, multiscale, and multiplex networks. Science 328, 876–878 (2010).
to function. Annu. Rev. Neurosci. 38, 269–289 (2015). 144. Vézquez-Rodríguez, B., Liu, Z. Q., Hagmann, P. & Misic, B. Signal propagation via cortical
113. Goulas, A., Zilles, K. & Hilgetag, C. C. Cortical gradients and laminar projections in hierarchies. Netw. Neurosci. 4, 1072–1090 (2020).
mammals. Trends Neurosci. 41, 775–788 (2018). 145. Milo, R. et al. Network motifs: simple building blocks of complex networks. Science 298,
114. Hilgetag, C. C., Beul, S. F., Albada, S. J. & van, Goulas, A. An architectonic type principle 824–827 (2002).
integrates macroscopic cortico-cortical connections with intrinsic cortical circuits of the 146. Sporns, O. & Kötter, R. Motifs in brain networks. PLoS Biol. 2, e369 (2004).
primate brain. Netw. Neurosci. 3, 905–923 (2019). 147. Shen, K. et al. Information processing architecture of functionally defined clusters in the
This paper reviews how cytoarchitecture guides neural wiring across species. macaque cortex. J. Neurosci. 32, 17465–17476 (2012).
115. Shine, J. M. Neuromodulatory influences on integration and segregation in the brain. 148. Sizemore, A. E. et al. Cliques and cavities in the human connectome. J. Comput. Neurosci.
Trends Cogn. Sci. 23, 572–583 (2019). 44, 115–145 (2018).
116. Barbas, H. Pattern in the laminar origin of corticocortical connections. J. Comp. Neurol. 149. Giusti, C., Ghrist, R. & Bassett, D. S. Two’s company, three (or more) is a simplex:
252, 415–422 (1986). algebraic-topological tools for understanding higher-order structure in neural data.
117. Barbas, H. & Pandya, D. Architecture and intrinsic connections of the prefrontal cortex J. Comput. Neurosci. 41, 1–14 (2016).
in the rhesus monkey. J. Comp. Neurol. 286, 353–375 (1989). 150. Vázquez-Rodrı́guez, B. et al. Gradients of structure–function tethering across neocortex.
118. Sethi, S. S., Zerbi, V., Wenderoth, N., Fornito, A. & Fulcher, B. D. Structural connectome Proc. Natl Acad. Sci. USA 116, 21219–27. (2019).
topology relates to regional BOLD signal dynamics in the mouse brain. Chaos 27, 047405 151. Preti, M. G. & Van De Ville, D. Decoupling of brain function from structure reveals regional
(2017). behavioral specialization in humans. Nat. Commun. 10, 4747 (2019).
119. Fallon, J. et al. Timescales of spontaneous fMRI fluctuations relate to structural 152. Baum, G. L. et al. Development of structure–function coupling in human brain. networks
connectivity in the brain. Netw. Neurosci. 4, 788–806 (2020). during youth. Proc. Natl Acad. Sci. USA 117, 771–778 (2020).
120. Theodoni, P. et al. Structural attributes and principles of the neocortical connectome 153. Liu Z. Q., Shafiei G., Baillet S., & Misic B. Spatially heterogeneous structure-function coupling
in the marmoset monkey. Cereb. Cortex 32, 15–28 (2022). in haemodynamic and electromagnetic brain networks. NeuroImage 278, 120276 (2023).
121. Xu, Z. et al. Meta-connectomic analysis maps consistent, reproducible, and 154. Zamani Esfahlani, F., Faskowitz, J., Slack, J., Mišić, B. & Betzel, R. F. Local structure-function
transcriptionally relevant functional connectome hubs in the human brain. relationships in human brain. networks across the lifespan. Nat. Commun. 13, 2053 (2022).
Commun. Biol. 5, 1056 (2022). 155. Wang, P. et al. Inversion of a large-scale circuit model reveals a cortical hierarchy in the
122. van den Heuvel, M. P., Scholtens, L. H., Barrett, L. F., Hilgetag, C. C. & de Reus, M. A. dynamic resting human brain. Sci. Adv. 5, eaat7854 (2019).
Bridging cytoarchitectonics and connectomics in human cerebral cortex. J. Neurosci. 156. Sip, V. et al. Characterization of regional differences in resting-state fMRI with
35, 13943–13948 (2015). a data-driven network model of brain dynamics. Sci. Adv. 9, eabq7547 (2023).
123. Hagmann, P. et al. Mapping the structural core of human cerebral cortex. PLoS Biol. 6, 157. Kong, X. et al. Sensory-motor cortices shape functional connectivity dynamics in the
e159 (2008). human brain. Nat. Commun. 12, 6373 (2021).
124. Várkuti, B. et al. Quantifying the link between anatomical connectivity, gray matter 158. Froudist-Walsh, S. et al. A dopamine gradient controls access to distributed working
volume and regional cerebral blood flow: an integrative MRI study. PLoS ONE 6, e14801 memory in the large-scale monkey cortex. Neuron 109, 3500–3520 (2021).
(2011). 159. Burt, J. B. et al. Transcriptomics-informed large-scale cortical model captures
125. Liang, X., Zou, Q., He, Y. & Yang, Y. Coupling of functional connectivity and regional topography of pharmacological neuroimaging effects of LSD. eLife 10, e69320 (2021).
cerebral blood flow reveals a physiological basis for network hubs of the human brain. 160. Singleton, S. P. et al. Receptor-informed network control theory links LSD and psilocybin
Proc. Natl Acad. Sci. USA 110, 1929–1934 (2013). to a flattening of the brain’s control energy landscape. Nat. Commun. 13, 5812 (2022).
126. Bullmore, E. & Sporns, O. The economy of brain network organization. Nat. Rev. Neurosci. 161. Deco, G. et al. Whole-brain multimodal neuroimaging model using serotonin receptor
13, 336–349 (2012). maps explains non-linear functional effects of LSD. Curr. Biol. 28, 3065–3074 (2018).
127. Collin, G., Sporns, O., Mandl, R. C. & Van Den Heuvel, M. P. Structural and functional 162. Shine, J. M. et al. Computational models link cellular mechanisms of neuromodulation
aspects relating to cost and benefit of rich club organization in the human cerebral to large-scale neural dynamics. Nat. Neurosci. 24, 765–776 (2021).
cortex. Cereb. Cortex 24, 2258–2267 (2014). 163. Deco, G. et al. Dynamical consequences of regional heterogeneity in the brain’s
128. Chen, Y., Lin, Q., Liao, X., Zhou, C. & He, Y. Association of aerobic glycolysis with the transcriptional landscape. Sci. Adv. 7, eabf4752 (2021).
structural connectome reveals a benefit–risk balancing mechanism in the human brain. 164. Deco, G. et al. How local excitation–inhibition ratio impacts the whole brain dynamics.
Proc. Natl Acad. Sci. USA 118, e2013232118 (2021). J. Neurosci. 34, 7886–7898 (2014).

Nature Reviews Neuroscience

Review article

165. Wong, K. F. & Wang, X. J. A recurrent network mechanism of time integration in 200. Fornito, A., Arnatkeviciute, A. & Fulcher, B. D. Bridging the gap between connectome
perceptual decisions. J. Neurosci. 26, 1314–1328 (2006). and transcriptome. Trends Cogn. Sci. 23, 34–50 (2019).
166. Betzel, R. F. & Bassett, D. S. Generative models for network neuroscience: prospects and 201. Goulas, A. et al. Cytoarchitectonic similarity is a wiring principle of the human
promise. J. R. Soc. Interface 14, 20170623 (2017). connectome. Preprint at BioRxiv https://doi.org/10.1101/068254 (2016).
167. Vértes, P. E. et al. Simple models of human brain functional networks. Proc. Natl Acad. 202. Barbas, H., Hilgetag, C. C., Saha, S., Dermon, C. R. & Suski, J. L. Parallel organization
Sci. USA 109, 5868–5873 (2012). of contralateral and ipsilateral prefrontal cortical projections in the rhesus monkey.
168. Betzel, R. F. et al. Generative models of the human connectome. Neuroimage 124, BMC Neurosci. 6, 32 (2005).
1054–1064 (2016). 203. Van Den Heuvel, M. P. & Sporns, O. Rich-club organization of the human connectome.
169. Goulas, A., Betzel, R. F. & Hilgetag, C. C. Spatiotemporal ontogeny of brain wiring. J. Neurosci. 31, 15775–15786 (2011).
Sci. Adv. 5, eaav9694 (2019). 204. Hansen, J. Y. et al. Integrating multimodal and multiscale connectivity blueprints of the
170. Akarca, D., Vértes, P. E., Bullmore, E. T. & Astle, D. E. A generative network model of human cerebral cortex in health and disease. PLoS Biol. 21, e3002314 (2023).
neurodevelopmental diversity in structural brain organization. Nat. Commun. 12, 4216 This paper systematically benchmarks and integrates seven types of annotation
(2021). similarity networks.
171. Ji, S., Fakhry, A. & Deng, H. Integrative analysis of the connectivity and gene expression 205. Hettwer, M. et al. Coordinated cortical thickness alterations across six
atlases in the mouse brain. Neuroimage 84, 245–253 (2014). neurodevelopmental and psychiatric disorders. Nat. Commun. 13, 6851 (2022).
172. Kaufman, A., Dror, G., Meilijson, I. & Ruppin, E. Gene expression of Caenorhabditis elegans 206. Sebenius, I. et al. Robust estimation of cortical similarity networks from brain MRI.
neurons carries information on their synaptic connectivity. PLoS Comput. Biol. 2, e167 Nat. Neurosci. 26, 1461–1471 (2023).
(2006). 207. Mansour S., Seguin C., Winkler A., Noble S., & Zalesky A. Topological cluster statistic
173. Wolf, L., Goldberg, C., Manor, N., Sharan, R. & Ruppin, E. Gene expression in the rodent (TCS): towards structural-connectivity-guided fMRI cluster enhancement. Preprint at
brain is associated with its regional connectivity. PLoS Comput. Biol. 7, e1002040 (2011). Research Square https://doi.org/10.21203/rs.3.rs-2059418/v1 (2022).
174. Oldham, S. et al. Modeling spatial, developmental, physiological, and topological 208. Cizeron, M. et al. A brainwide atlas of synapses across the mouse life span. Science 369,
constraints on human brain connect. Sci. Adv. 8, eabm6127 (2022). 270–275 (2020).
This study introduces generative models in which edge placement is governed by 209. Ding, J. et al. A metabolome atlas of the aging mouse brain. Nat. Commun. 12, 6021 (2021).
biological rules in addition to spatial and topological rules. 210. Luk, K. C. et al. Pathological α-synuclein transmission initiates Parkinson-like
175. Jucker, M. & Walker, L. C. Self-propagation of pathogenic protein aggregates in neurodegeneration in nontransgenic mice. Science 338, 949–953 (2012).
neurodegenerative diseases. Nature 501, 45–51 (2013). 211. Van Essen, D. C. et al. The WU-Minn human connectome project: an overview.
176. Henderson, M. X. et al. Spread of α-synuclein pathology through the brain connectome Neuroimage 80, 62–79 (2013).
is modulated by selective vulnerability and predicted by network analysis. Nat. Neurosci. 212. Finnema, S. J. et al. Kinetic evaluation and test–retest reproducibility of [11C] UCB-j,
22, 1248–1257 (2019). a novel radioligand for positron emission tomography imaging of synaptic vesicle
177. Rahayel, S. et al. Differentially targeted seeding reveals unique pathological glycoprotein 2A in humans. J. Cereb. Blood Flow Metab. 38, 2041–2052 (2018).
alpha-synuclein propagation patterns. Brain 145, 1743–1756 (2022). 213. Schaefer, A. et al. Local-global parcellation of the human cerebral cortex from intrinsic
178. Zheng, Y. Q. et al. Local vulnerability and global connectivity jointly shape functional connectivity MRI. Cereb. Cortex 28, 3095–3114 (2018).
neurodegenerative disease propagation. PLoS Biol. 17, e3000495 (2019). 214. Friston, K. J. Functional and effective connectivity: a review. Brain Connect. 1, 13–36
179. Abdelgawad, A. et al. Predicting longitudinal brain atrophy in Parkinson’s disease using (2011).
a susceptible-infected-removed agent-based model. Netw. Neurosci. 2022;7:906–925. 215. Caeyenberghs, K., Metzler-Baddeley, C., Foley, S. & Jones, D. K. Dynamics of the human
180. Rahayel, S. et al. Brain atrophy in prodromal synucleinopathy is shaped by structural structural connectome underlying working memory training. J. Neurosci. 36, 4056–4066
connectivity and gene expression. Brain 145, 3162–3178 (2022). (2016).
181. Xie, Y. et al. Alterations in connectome dynamics in autism spectrum disorder: 216. Nelson, M. C. et al. The human brain connectome weighted by the myelin content
a harmonized mega-and meta-analysis study using the autism brain imaging data and total intra-axonal cross-sectional area of white matter tracts. Netw. Neurosci.
exchange dataset. Biol. Psychiatry 91, 945–955 (2022). https://doi.org/10.1162/netn_a_00330 (2023).
182. Xia, M. et al. Connectome gradient dysfunction in major depression and its association with 217. Zhang, H., Schneider, T., Wheeler-Kingshott, C. A. & Alexander, D. C. NODDI: practical
gene expression profiles and treatment outcomes. Mol. Psychiatry 27, 1384–1393 (2022). in vivo neurite orientation dispersion and density imaging of the human brain.
183. Roberts, J. A. et al. The contribution of geometry to the human connectome. Neuroimage Neuroimage 61, 1000–1016 (2012).
124, 379–393 (2016). 218. Assaf, Y., Blumenfeld-Katzir, T., Yovel, Y. & Basser, P. J. AxCaliber: a method for measuring
184. Hébert, J. M. & Fishell, G. The genetics of early telencephalon patterning: some assembly axon diameter distribution from diffusion MRI. Magn. Reson. Med. 59, 1347–1354 (2008).
required. Nat. Rev. Neurosci. 9, 678–685 (2008). 219. Alexander, D. C. et al. Orientationally invariant indices of axon diameter and density from
185. Cadwell, C. R., Bhaduri, A., Mostajo-Radji, M. A., Keefe, M. G. & Nowakowski, T. J. diffusion MRI. Neuroimage 52, 1374–1389 (2010).
Development and arealization of the cerebral cortex. Neuron 103, 980–1004 (2019). 220. Boshkovski, T. et al. The R1-weighted connectome: complementing brain networks with
186. Ercsey-Ravasz, M. et al. A predictive network model of cerebral cortical connectivity a myelin-sensitive measure. Netw. Neurosci. 5, 358–372 (2021).
based on a distance rule. Neuron 80, 184–197 (2013). 221. Stikov, N. et al. In vivo histology of the myelin g-ratio with magnetic resonance imaging.
187. Horvát, S. et al. Spatial embedding and wiring cost constrain the functional layout of the Neuroimage 118, 397–405 (2015).
cortical network of rodents and primates. PLoS Biol. 14, e1002512 (2016). 222. Mancini, M. et al. Introducing axonal myelination in connectomics: a preliminary analysis
188. Váša, F. & Mišić, B. Null models in network neuroscience. Nat. Rev. Neurosci. 23, 493–504 of g-ratio distribution in healthy subjects. Neuroimage 182, 351–359 (2018).
(2022). 223. Drakesmith, M. et al. Estimating axon conduction velocity in vivo from microstructural
This paper reviews inferential methods for disentangling the relationships between MRI. Neuroimage 203, 116186 (2019).
connectivity, geometry and annotations. 224. Luo, L., Callaway, E. M. & Svoboda, K. Genetic dissection of neural circuits: a decade
189. Alexander-Bloch, A. F. et al. On testing for spatial correspondence between maps of of progress. Neuron 98, 256–281 (2018).
human brain structure and function. Neuroimage 178, 540–551 (2018). 225. Frässle, S. et al. Whole-brain estimates of directed connectivity for human connectomics.
190. Breakspear, M., Brammer, M. J., Bullmore, E. T., Das, P. & Williams, L. M. Spatiotemporal Neuroimage 225, 117491 (2021).
wavelet resampling for functional neuroimaging data. Hum. Brain Mapp. 23, 1–25 (2004). 226. Seguin, C. et al. Communication dynamics in the human connectome shape the
191. Burt, J. B., Helmer, M., Shinn, M., Anticevic, A. & Murray, J. D. Generative modeling of cortex-wide propagation of direct electrical stimulation. Neuron 111, 1391–1401 (2023).
brain maps with spatial autocorrelation. Neuroimage 220, 117038 (2020). 227. Veit, M. J. et al. Temporal order of signal propagation within and across intrinsic brain
192. Betzel, R. F. & Bassett, D. S. Specificity and robustness of long-distance connections in networks. Proc. Natl Acad. Sci. USA 118, e2105031118 (2021).
weighted, interareal connectomes. Proc. Natl Acad. Sci. USA 115, E4880–E4889 (2018). 228. Seguin, C., Razi, A. & Zalesky, A. Inferring neural signalling directionality from undirected
193. Mišić, B. et al. Cooperative and competitive spreading dynamics on the human structural connectomes. Nat. Commun. 10, 4289 (2019).
connectome. Neuron 86, 1518–1529 (2015). 229. Reus, M. A., de, Van & den Heuvel, M. P. The parcellation-based connectome: limitations
194. Sorrentino, P. et al. The structural connectome constrains fast brain dynamics. eLife 10, and extensions. Neuroimage 80, 397–404 (2013).
e67400 (2021). 230. Gajwani, M. et al. Can hubs of the human connectome be identified consistently with
195. Horwitz, B., Duara, R. & Rapoport, S. I. Intercorrelations of glucose metabolic rates diffusion MRI? Netw. Neurosci. https://doi.org/10.1162/netn_a_00324 (2023).
between brain regions: application to healthy males in a state of reduced sensory input. 231. Dhamala, E., Jamison, K. W., Jaywant, A., Dennis, S. & Kuceyeski, A. Distinct functional
J. Cereb. Blood Flow Metab. 4, 484–499 (1984). and structural connections predict crystallised and fluid cognition in healthy adults.
196. Evans, A. C. Networks of anatomical covariance. Neuroimage 80, 489–504 (2013). Hum. Brain Mapp. 42, 3102–3118 (2021).
197. Jamadar, S. D. et al. Metabolic and hemodynamic resting-state connectivity of the human 232. Dhamala, E., Yeo, B. T. & Holmes, A. J. One size does not fit all: methodological
brain: a high-temporal resolution simultaneous BOLD-fMRI and FDG-fPET multimodality considerations for brain-based predictive modeling in psychiatry. Biol. Psychiatry 93,
study. Cereb. Cortex 31, 2855–2867 (2021). 717–728 (2023).
198. Sala, A. et al. Brain connectomics: time for a molecular imaging perspective? 233. Eickhoff, S. B., Yeo, B. T. & Genon, S. Imaging-based parcellations of the human brain.
Trends Cogn. Sci. 27, 353–366 (2023). Nat. Rev. Neurosci. 19, 672–686 (2018).
199. Brookes, M. J. et al. Measuring functional connectivity using MEG: methodology and 234. Poldrack, R. A. et al. Long-term neural and physiological phenotyping of a single human.
comparison with fcMRI. Neuroimage 56, 1082–1104 (2011). Nat. Commun. 6, 8885 (2015).

Nature Reviews Neuroscience

Review article

235. Gordon, E. M. et al. Precision functional mapping of individual human brain. Neuron 95, Competing interests
791–807 (2017). The authors declare no competing interests.
236. Gratton, C. et al. Functional brain networks are dominated by stable group and individual
factors, not cognitive or daily variation. Neuron 98, 439–452 (2018). Additional information
237. Kong, R. et al. Spatial topography of individual-specific cortical networks predicts human Supplementary information The online version contains supplementary material available at
cognition, personality, and emotion. Cereb. Cortex 29, 2533–2551 (2019). https://doi.org/10.1038/s41583-023-00752-3.
238. Glasser, M. F. et al. A multi-modal parcellation of human cerebral cortex. Nature 536,
171–178 (2016). Peer review information Nature Reviews Neuroscience thanks Y. He, A. Kucyi and J. Lerch for
their contribution to the peer review of this work.
Acknowledgements
We thank E. Suarez, A. Luppi, F. Milisav, Z.Q. Liu, E. Ceballos, A. Farahani and M. Pourmajidian Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
for their comments and suggestions. This work was funded by the Natural Sciences and published maps and institutional affiliations.
Engineering Research Council of Canada (NSERC), the Canadian Institutes of Health Research
(CIHR), the Canada Research Chair (CRC) Program, Fonds de Recherche du Quebec–Nature Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this
et Technologie (FRQNT), Helmholtz International BigBrain Analytics and Learning Laboratory article under a publishing agreement with the author(s) or other rightsholder(s); author
(HIBALL) and the Brain Canada Foundation. self-archiving of the accepted manuscript version of this article is solely governed by the
terms of such publishing agreement and applicable law.
Author contributions
V.B. and J.Y.H. researched data for the article. All authors made substantial contributions to the
discussion of content and wrote, reviewed and edited the manuscript before submission. © Springer Nature Limited 2023

Nature Reviews Neuroscience