On the Nature of Explanations Offered by Network Science: A Perspective From and for Practicing Neuroscientists

Hierarchies of description enable understanding of cognitive phenomena in terms of neuron activity

One objective of neuroscience is to understand a wide range of specific cognitive processes in terms of neuron activity. The huge amount of observational data about the brain makes achieving this objective challenging. Different models on different levels of detail provide some insight, but the relationship between models on different levels is not clear. Complex computing systems with trillions of components like transistors are fully understood in the sense that system features can be precisely related to transistor activity. Such understanding could not involve a designer simultaneously thinking about the ongoing activity of all the components active in the course of carrying out some system feature. Brain modeling approaches like dynamical systems are inadequate to support understanding of computing systems, because their use relies on approximations like treating all components as more or less identical. Understanding computing systems needs a much more sophisticated use of approximation, involving creation of hierarchies of description in which the higher levels are more approximate, with effective translation between different levels in the hierarchy made possible by using the same general types of information processes on every level. These types are instruction and data read/write. There are no direct resemblances between computers and brains, but natural selection pressures have resulted in brain resources being organized into modular hierarchies and in the existence of two general types of information processes called condition definition/detection and behavioral recommendation. As a result, it is possible to create hierarchies of description linking cognitive phenomena to neuron activity, analogous with but qualitatively different from the hierarchies of description used to understand computing systems. An intuitively satisfying understanding of cognitive processes in terms of more detailed brain activity is then possible.

Using network control theory to study the dynamics of the structural connectome

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

Strong intercorrelations among global graph-theoretic indices of structural connectivity in the human brain

Graph-theoretic metrics derived from neuroimaging data have been heralded as powerful tools for uncovering neural mechanisms of psychological traits, psychiatric disorders, and neurodegenerative diseases. In N = 8,185 human structural connectomes from UK Biobank, we examined the extent to which 11 commonly-used global graph-theoretic metrics index distinct versus overlapping information with respect to interindividual differences in brain organization. Using unthresholded, FA-weighted networks we found that all metrics other than Participation Coefficient were highly intercorrelated, both with each other (mean |r| = 0.788) and with a topologically-naïve summary index of brain structure (mean edge weight; mean |r| = 0.873). In a series of sensitivity analyses, we found that overlap between metrics is influenced by the sparseness of the network and the magnitude of variation in edge weights. Simulation analyses representing a range of population network structures indicated that individual differences in global graph metrics may be intrinsically difficult to separate from mean edge weight. In particular, Closeness, Characteristic Path Length, Global Efficiency, Clustering Coefficient, and Small Worldness were nearly perfectly collinear with one another (mean |r| = 0.939) and with mean edge weight (mean |r| = 0.952) across all observed and simulated conditions. Global graph-theoretic measures are valuable for their ability to distill a high-dimensional system of neural connections into summary indices of brain organization, but they may be of more limited utility when the goal is to index separable components of interindividual variation in specific properties of the human structural connectome.

Case report: "Proust phenomenon" after right posterior cerebral artery occlusion

Odors evoking vivid and intensely felt autobiographical memories are known as the "Proust phenomenon," delineating the particularity of olfaction in being more effective with eliciting emotional memories than other sensory modalities. The phenomenon has been described extensively in healthy participants as well as in patients during pre-epilepsy surgery evaluation after focal stimulation of the amygdalae and post-traumatic stress disorder (PTSD). In this study, we provide the inaugural description of aversive odor-evoked autobiographical memories after stroke in the right hippocampal, parahippocampal, and thalamic nuclei. As potential underlying neural signatures of the phenomenon, we discuss the disinhibition of limbic circuits and impaired communication between the major networks, such as saliency, central executive, and default mode network.

Comparing representations and computations in single neurons versus neural networks

Single-neuron-level explanations have been the gold standard in neuroscience for decades. Recently, however, neural-network-level explanations have become increasingly popular. This increase in popularity is driven by the fact that the analysis of neural networks can solve problems that cannot be addressed by analyzing neurons independently. In this opinion article, I argue that while both frameworks employ the same general logic to link physical and mental phenomena, in many cases the neural network framework provides better explanatory objects to understand representations and computations related to mental phenomena. I discuss what constitutes a mechanistic explanation in neural systems, provide examples, and conclude by highlighting a number of the challenges and considerations associated with the use of analyses of neural networks to study brain function.

Assessing the Repeatability of Multi-Frequency Multi-Layer Brain Network Topologies Across Alternative Researcher's Choice Paths

There is a growing interest in the neuroscience community on the advantages of multilayer functional brain networks. Researchers usually treated different frequencies separately at distinct functional brain networks. However, there is strong evidence that these networks share complementary information while their interdependencies could reveal novel findings. For this purpose, neuroscientists adopt multilayer networks, which can be described mathematically as an extension of trivial single-layer networks. Multilayer networks have become popular in neuroscience due to their advantage to integrate different sources of information. Here, Ι will focus on the multi-frequency multilayer functional connectivity analysis on resting-state fMRI (rs-fMRI) recordings. However, constructing a multilayer network depends on selecting multiple pre-processing steps that can affect the final network topology. Here, I analyzed the rs-fMRI dataset from a single human performing scanning over a period of 18 months (84 scans in total), and the rs-fMRI dataset containing 25 subjects with 3 repeat scans. I focused on assessing the reproducibility of multi-frequency multilayer topologies exploring the effect of two filtering methods for extracting frequencies from BOLD activity, three connectivity estimators, with or without a topological filtering scheme, and two spatial scales. Finally, I untangled specific combinations of researchers' choices that yield consistently brain networks with repeatable topologies, giving me the chance to recommend best practices over consistent topologies.

Effects of Acute Resistance Exercise on Executive Function: A Systematic Review of the Moderating Role of Intensity and Executive Function Domain

BACKGROUND

Research has demonstrated that there is a beneficial effect of acute exercise on cognitive function; however, the moderators of the acute resistance exercise (RE) effect on executive function (EF) are underestimated. This systematic review aims to clarify the effects of acute RE on EF by examining the moderating effect of exercise intensity (light, moderate, and vigorous) and EF domains (inhibitory control, working memory, and cognitive flexibility), as well as their interactions.

METHODS

The search strategy was conducted in four databases (PubMed, Scopus, PsycARTICLES, and Cochrane Library) prior to January 29, 2022. Included studies had to: (1) investigate acute RE in adults with normal cognition and without diagnosed disease; (2) include a control group or control session for comparison; (3) include outcomes related to the core EF domains; and (4) be published in English. The methodological quality of the included studies was judged according to the PEDro scale guidelines.

RESULTS

Nineteen studies were included which included a total of 692 participants. More than half of the outcomes (24/42, 57.14%) indicate that acute RE had a statistically significant positive effect on overall EF. In terms of RE intensity and EF domain, moderate intensity acute RE benefited EF more consistently than light and vigorous intensity acute RE. Acute RE-induced EF benefits were more often found for inhibitory control than for working memory and cognitive flexibility. When considering moderators simultaneously, measuring inhibitory control after light or moderate intensity RE and measuring working memory or cognitive flexibility after moderate intensity RE most often resulted in statistically significant positive outcomes.

CONCLUSION

Acute RE has a beneficial effect on EF, observed most consistently for inhibitory control following moderate intensity RE. Future studies should include all exercise intensities and EF domains as well as investigate other potential moderators to enable a better understanding of the benefits of acute RE on EF.

From calcium imaging to graph topology

Systems neuroscience is facing an ever-growing mountain of data. Recent advances in protein engineering and microscopy have together led to a paradigm shift in neuroscience; using fluorescence, we can now image the activity of every neuron through the whole brain of behaving animals. Even in larger organisms, the number of neurons that we can record simultaneously is increasing exponentially with time. This increase in the dimensionality of the data is being met with an explosion of computational and mathematical methods, each using disparate terminology, distinct approaches, and diverse mathematical concepts. Here we collect, organize, and explain multiple data analysis techniques that have been, or could be, applied to whole-brain imaging, using larval zebrafish as an example model. We begin with methods such as linear regression that are designed to detect relations between two variables. Next, we progress through network science and applied topological methods, which focus on the patterns of relations among many variables. Finally, we highlight the potential of generative models that could provide testable hypotheses on wiring rules and network progression through time, or disease progression. While we use examples of imaging from larval zebrafish, these approaches are suitable for any population-scale neural network modeling, and indeed, to applications beyond systems neuroscience. Computational approaches from network science and applied topology are not limited to larval zebrafish, or even to systems neuroscience, and we therefore conclude with a discussion of how such methods can be applied to diverse problems across the biological sciences.

Network analysis of neuroimaging in mice

Graph theory allows assessing changes of neuronal connectivity and interactions of brain regions in response to local lesions, e.g., after stroke, and global perturbations, e.g., due to psychiatric dysfunctions or neurodegenerative disorders. Consequently, network analysis based on constructing graphs from structural and functional MRI connectivity matrices is increasingly used in clinical studies. In contrast, in mouse neuroimaging, the focus is mainly on basic connectivity parameters, i.e., the correlation coefficient or fiber counts, whereas more advanced network analyses remain rarely used. This review summarizes graph theoretical measures and their interpretation to describe networks derived from recent in vivo mouse brain studies. To facilitate the entry into the topic, we explain the related mathematical definitions, provide a dedicated software toolkit, and discuss practical considerations for the application to rs-fMRI and DTI. This way, we aim to foster cross-species comparisons and the application of standardized measures to classify and interpret network changes in translational brain disease studies.

Brain Network Segregation and Integration during Painful Thermal Stimulation

The present study aimed to determine changes in brain network integration/segregation during thermal pain using methods optimized for network connectivity events with high temporal resolution. Participants (n = 33) actively judged whether thermal stimuli applied to the volar forearm were painful or not and then rated the warmth/pain intensity after each trial. We show that the temporal evolution of integration/segregation within trials correlates with the subjective ratings of pain. Specifically, the brain shifts from a segregated state to an integrated state when processing painful stimuli. The association with subjective pain ratings occurred at different time points for all networks. However, the degree of association between ratings and integration/segregation vanished for several brain networks when time-varying functional connectivity was measured at lower temporal resolution. Moreover, the increased integration associated with pain is explained to some degree by relative increases in between-network connectivity. Our results highlight the importance of investigating the relationship between pain and brain network connectivity at a single time point scale, since commonly used temporal aggregations of connectivity data may result in that fine-scale changes in network connectivity may go unnoticed. The interplay between integration/segregation reflects shifting demands of information processing between brain networks and this adaptation occurs both for cognitive tasks and nociceptive processing.

The time-varying networks of the wrist extension in post-stroke hemiplegic patients

Hemiplegia is a common dysfunction caused by the brain stroke and leads to movement disability. Although the lateralization of movement-related potential, the event-related desynchronization, and more complicated inter-regional information coupling have been investigated, seldom studies have focused on investigating the dynamic information exchanging among multiple brain regions during motor execution for post-stroke hemiplegic patients. With high temporal-resolution electroencephalogram (EEG), the time-varying network is able to reflect the dynamical complex network modalities corresponding to the movements at a millisecond level. In our present study, the wrist extension experiment was designed, along with related EEG datasets being collected. Thereafter, the corresponding time-varying networks underlying the wrist extension were accordingly constructed by adopting the adaptive directed transfer function and then statistically explored, to further uncover the dynamic network deficits (i.e., motor dysfunction) in post-stroke hemiplegic patients. Results of this study found the effective connectivity between the stroked motor area and other areas decreased in patients when compared to healthy controls; on the contrary, the enhanced connectivity between non-stroked motor areas and other areas, especially the frontal and parietal-occipital lobes, were further identified for patients during their accomplishing the designed wrist extension, which might dynamically compensate for the deficited patients' motor behaviors. These findings not only helped deepen our knowledge of the mechanism underlying the patients' motor behaviors, but also facilitated the real-time strategies for clinical therapy of brain stroke, as well as providing a reliable biomarker to predict the future rehabilitation.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11571-021-09738-2.

Connecting brain and behavior in clinical neuroscience: A network approach

In recent years, there has been an increase in applications of network science in many different fields. In clinical neuroscience and psychopathology, the developments and applications of network science have occurred mostly simultaneously, but without much collaboration between the two fields. The promise of integrating these network applications lies in a united framework to tackle one of the fundamental questions of our time: how to understand the link between brain and behavior. In the current overview, we bridge this gap by introducing conventions in both fields, highlighting similarities, and creating a common language that enables the exploitation of synergies. We provide research examples in autism research, as it accurately represents research lines in both network neuroscience and psychological networks. We integrate brain and behavior not only semantically, but also practically, by showcasing three methodological avenues that allow to combine networks of brain and behavioral data. As such, the current paper offers a stepping stone to further develop multi-modal networks and to integrate brain and behavior.

Two views on the cognitive brain

Cognition can be defined as computation over meaningful representations in the brain to produce adaptive behaviour. There are two views on the relationship between cognition and the brain that are largely implicit in the literature. The Sherringtonian view seeks to explain cognition as the result of operations on signals performed at nodes in a network and passed between them that are implemented by specific neurons and their connections in circuits in the brain. The contrasting Hopfieldian view explains cognition as the result of transformations between or movement within representational spaces that are implemented by neural populations. Thus, the Hopfieldian view relegates details regarding the identity of and connections between specific neurons to the status of secondary explainers. Only the Hopfieldian approach has the representational and computational resources needed to develop novel neurofunctional objects that can serve as primary explainers of cognition.

The cerebellum and its network: Disrupted static and dynamic functional connectivity patterns and cognitive impairment in multiple sclerosis

Background:

The impact of cerebellar damage and (dys)function on cognition remains understudied in multiple sclerosis.

Objective:

To assess the cognitive relevance of cerebellar structural damage and functional connectivity (FC) in relapsing-remitting multiple sclerosis (RRMS) and secondary progressive multiple sclerosis (SPMS).

Methods:

This study included 149 patients with early RRMS, 81 late RRMS, 48 SPMS and 82 controls. Cerebellar cortical imaging included fractional anisotropy, grey matter volume and resting-state functional magnetic resonance imaging (MRI). Cerebellar FC was assessed with literature-based resting-state networks, using static connectivity (that is, conventional correlations), and dynamic connectivity (that is, fluctuations in FC strength). Measures were compared between groups and related to disability and cognition.

Results:

Cognitive impairment (CI) and cerebellar damage were worst in SPMS. Only SPMS showed cerebellar connectivity changes, compared to early RRMS and controls. Lower static FC was seen in fronto-parietal and default-mode networks. Higher dynamic FC was seen in dorsal and ventral attention, default-mode and deep grey matter networks. Cerebellar atrophy and higher dynamic FC together explained 32% of disability and 24% of cognitive variance. Higher dynamic FC was related to working and verbal memory and to information processing speed.

Conclusion:

Cerebellar damage and cerebellar connectivity changes were most prominent in SPMS and related to worse CI.

Assessing the Relationship between Verbal and Nonverbal Cognitive Abilities Using Resting-State EEG Functional Connectivity

The present study investigates the relationship between individual differences in verbal and non-verbal cognitive abilities and resting-state EEG network characteristics. We used a network neuroscience approach to analyze both large-scale topological characteristics of the whole brain as well as local brain network characteristics. The characteristic path length, modularity, and cluster coefficient for different EEG frequency bands (alpha, high and low; beta1 and beta2, and theta) were calculated to estimate large-scale topological integration and segregation properties of the brain networks. Betweenness centrality, nodal clustering coefficient, and local connectivity strength were calculated as local network characteristics. We showed that global network integration measures in the alpha band were positively correlated with non-verbal intelligence, especially with the more difficult part of the test (Raven's total scores and E series), and the ability to operate with verbal information (the "Conclusions" verbal subtest). At the same time, individual differences in non-verbal intelligence (Raven's total score and C series), and vocabulary subtest of the verbal intelligence tests, were negatively correlated with the network segregation measures. Our results show that resting-state EEG functional connectivity can reveal the functional architecture associated with an individual difference in cognitive performance.

Modeling brain, symptom, and behavior in the winds of change

Neuropsychopharmacology addresses pressing questions in the study of three intertwined complex systems: the brain, human behavior, and symptoms of illness. The field seeks to understand the perturbations that impinge upon those systems, either driving greater health or illness. In the pursuit of this aim, investigators often perform analyses that make certain assumptions about the nature of the systems that are being perturbed. Those assumptions can be encoded in powerful computational models that serve to bridge the wide gulf between a descriptive analysis and a formal theory of a system's response. Here we review a set of three such models along a continuum of complexity, moving from a local treatment to a network treatment: one commonly applied form of the general linear model, impulse response models, and network control models. For each, we describe the model's basic form, review its use in the field, and provide a frank assessment of its relative strengths and weaknesses. The discussion naturally motivates future efforts to interlink data analysis, computational modeling, and formal theory. Our goal is to inspire practitioners to consider the assumptions implicit in their analytical approach, align those assumptions to the complexity of the systems under study, and take advantage of exciting recent advances in modeling the relations between perturbations and system function.

Editors' Review and Introduction: Levels of Explanation in Cognitive Science: From Molecules to Culture

Cognitive science began as a multidisciplinary endeavor to understand how the mind works. Since the beginning, cognitive scientists have been asking questions about the right methodologies and levels of explanation to pursue this goal, and make cognitive science a coherent science of the mind. Key questions include: Is there a privileged level of explanation in cognitive science? How do different levels of explanation fit together, or relate to one another? How should explanations at one level inform or constrain explanations at some other level? Can the different approaches to the mind, brain, and culture be unified? The aim of this issue of topiCS is to provide a platform for discussing different answers to such questions and to facilitate a better understanding between the different strands of thinking about the right levels of explanation in cognitive science.

Introduction to “Levels of Explanation in Cognitive Science: From Molecules to Culture”

This paper introduces the topic “Levels of Explanation in Cognitive Science: From Molecules to Culture”, puts into focus some key questions, and provides an overview of the contributions in this topic.

Data-Driven Approaches to Neuroimaging Analysis to Enhance Psychiatric Diagnosis and Therapy

Combining advanced neuroimaging with novel computational methods in network science and machine learning has led to increasingly meaningful descriptions of structure and function in both the normal and the abnormal brain, thereby contributing significantly to our understanding of psychiatric disorders as circuit dysfunctions. Despite its marked potential for psychiatric care, this approach has not yet extended beyond the research setting to any clinically useful applications. Here we review current developments in the study of neuroimaging data using network models and machine learning methods, with a focus on their promise in offering a framework for clinical translation. We discuss 3 potential contributions of these methods to psychiatric care: 1) a better understanding of psychopathology beyond current diagnostic boundaries; 2) individualized prediction of treatment response and prognosis; and 3) formal theories to guide the development of novel interventions. Finally, we highlight current obstacles and sketch a forward-looking perspective of how the application of machine learning and network modeling methods should proceed to accelerate their potential transformation of clinically useful tools.