Network Connectivity Patterns Research Articles

Dynamic coherence analysis of resting fMRI data to jointly capture state-based phase, frequency, and time-domain information

Many approaches for estimating functional connectivity among brain regions or networks in fMRI have been considered in the literature. More recently, studies have shown that connectivity which is usually estimated by calculating correlation between time series or by estimating coherence as a function of frequency has a dynamic nature, during both task and resting conditions. Sliding-window methods have been commonly used to study these dynamic properties although other approaches such as instantaneous phase synchronization have also been used for similar purposes.Some studies have also suggested that spectral analysis can be used to separate the distinct contributions of motion, respiration and neurophysiological activity from the observed correlation. Several recent studies have merged analysis of coherence with study of temporal dynamics of functional connectivity though these have mostly been limited to a few selected brain regions and frequency bands.Here we propose a novel data-driven framework to estimate time-varying patterns of whole-brain functional network connectivity of resting state fMRI combined with the different frequencies and phase lags at which these patterns are observed. We show that this analysis identifies both broad-band cluster centroids that summarize connectivity patterns observed in many frequency bands, as well as clusters consisting only of functional network connectivity (FNC) from a narrow range of frequencies along with associated phase profiles. The value of this approach is demonstrated by its ability to reveal significant group differences in males versus females regarding occupancy rates of cluster that would not be separable without considering the frequencies and phase lags. The method we introduce provides a novel and informative framework for analyzing time-varying and frequency specific connectivity which can be broadly applied to the study of the healthy and diseased human brain.

Self-organization of grid fields under supervision of place cells in a neuron model with associative plasticity

Abstract The grid cells (GCs) of the medial entorhinal cortex (MEC) and place cells (PCs) of the hippocampus are assumed to be the key elements of the brain network for the metric representation of space. Existing theoretical models of GC network rely on specific hypotheses about the network connectivity patterns. How these patterns could be formed during the network development is not fully understood. It was previously suggested, within the feedforward network models, that activity of PCs could provide the basis for development of GC-like activity patterns. Supporting this hypothesis is the finding that PC activity remains spatially stable after disruption of the GC firing patterns and that the grid fields almost disappear when hippocampal cells are deactivated. Here a new theoretical model of this type is proposed, allowing for grid fields formation due to synaptic plasticity in synapses connecting PCs in hippocampus with neurons in MEC. Learning of the hexagonally symmetric fields in this model occurs due to complex action of several simple biologically motivated synaptic plasticity rules. These rules include associative synaptic plasticity rules similar to BCM rule, and homeostatic plasticity rules that constrain synaptic weights. In contrast to previously described models, a short-term navigational experience in a novel environment is sufficient for the network to learn GC activity patterns. We suggest that learning on the basis of simple and biologically plausible associative synaptic plasticity rules could contribute to the formation of grid fields in early development and to maintenance of normal GCs activity patterns in the familiar environments.

Brain connectivity and prediction of relapse after cognitive-behavioral therapy in obsessive-compulsive disorder.

Intensive cognitive-behavioral therapy (CBT) can effectively reduce symptoms in obsessive-compulsive disorder (OCD). However, many relapse after treatment. Few studies have investigated biological markers predictive of follow-up clinical status. The objective was to determine if brain network connectivity patterns prior to intensive CBT predict worsening of clinical symptoms during follow-up. We acquired resting-state functional magnetic resonance imaging data from 17 adults with OCD prior to and following 4 weeks of intensive CBT. Functional connectivity data were analyzed to yield graph-theory metrics. We examined the relationship between pre-treatment connectome properties and OCD clinical symptoms before and after treatment and during a 12-month follow-up period. Mean OCD symptom decrease was 40.4 ± 16.4% pre- to post-treatment (64.7% responded; 58.8% remitted), but 35.3% experienced clinically significant worsening during follow-up. From pre- to post-treatment, small-worldness and clustering coefficient significantly increased. Decreases in modularity correlated with decreases in OCD symptoms. Higher pre-treatment small-world connectivity was significantly associated with worsening of OCD symptoms during the follow-up period. Psychometric and neurocognitive measures pre- and post-treatment were not significant predictors. This is the first graph-theory connectivity study of the effects of CBT in OCD, and the first to test associations with follow-up clinical status. Results show functional network efficiency as a biomarker of CBT response and relapse in OCD. CBT increases network efficiency as it alleviates symptoms in most patients, but those entering therapy with already high network efficiency are at greater risk of relapse. Results have potential clinical implications for treatment selection.

Abnormal connectivity between the default mode and the visual system underlies the manifestation of visual hallucinations in Parkinson's disease: a task-based fMRI study.

Background:The neural substrates of visual hallucinations remain an enigma, due primarily to the difficulties associated with directly interrogating the brain during hallucinatory episodes.Aims:To delineate the functional patterns of brain network activity and connectivity underlying visual hallucinations in Parkinson’s disease.Methods:In this study, we combined functional magnetic resonance imaging (MRI) with a behavioral task capable of eliciting visual misperceptions, a confirmed surrogate for visual hallucinations, in 35 patients with idiopathic Parkinson’s disease. We then applied an independent component analysis to extract time series information for large-scale neuronal networks that have been previously implicated in the pathophysiology of visual hallucinations. These data were subjected to a task-based functional connectivity analysis, thus providing the first objective description of the neural activity and connectivity during visual hallucinations in patients with Parkinson’s disease.Results:Correct performance of the task was associated with increased activity in primary visual regions; however, during visual misperceptions, this same visual network became actively coupled with the default mode network (DMN). Further, the frequency of misperception errors on the task was positively correlated with the strength of connectivity between these two systems, as well as with decreased activity in the dorsal attention network (DAN), and with impaired connectivity between the DAN and the DMNs, and ventral attention networks. Finally, each of the network abnormalities identified in our analysis were significantly correlated with two independent clinical measures of hallucination severity.Conclusions:Together, these results provide evidence that visual hallucinations are due to increased engagement of the DMN with the primary visual system, and emphasize the role of dysfunctional engagement of attentional networks in the pathophysiology of hallucinations.

Age-related changes in resting-state networks of a large sample size of healthy elderly.

Population aging is burdening the society globally, and the evaluation of functional networks is the key toward understanding cognitive changes in normal aging. However, the effect of age on default mode subnetworks has not been documented well, and age-related changes in many resting-state networks remain debatable. The purpose of this study was to propose more precise results for these issues using a large sample size. We used group-level meta-ICA analysis and dual regression approach for identifying resting-state networks from functional magnetic resonance imaging data of 430 healthy elderly participants. Partial correlation was used to observe age-related correlations within and between resting-state networks. In the default mode network, only the ventral subnetwork negatively correlated with age. Age-related decrease in functional connectivity was also noted in the auditory, right frontoparietal, sensorimotor, and visual medial networks. Further, some age-related increases and decreases were observed for between-network correlations. The results of this study suggest that only the ventral default mode subnetwork had age-related decline in functional connectivity and several reverse patterns of resting-state networks for network development. Understanding age-related network changes may provide solutions for the impact of population aging and diagnosis of neurodegenerative diseases.

Cross-National Patterns of Social Capital Accumulation

Cultural and institutional context has the potential to moderate life course patterns of social interaction and network connectivity, yet few have attempted to empirically assess this claim. Contrasting collectivist versus individualistic cultural traditions, as well as socialist versus capitalist institutions, we develop and test a set of propositions regarding age-based variation in daily contact, occupational connections, and organizational memberships in China, Taiwan, and the United States. Analyses from cross-sectional survey data reveal how the cultural and institutional differences help structure access to social capital across age. Specifically, the data show how social capital accumulation in the individualistic societies is facilitated by employment and civic institutions, whereas family institutions form the basis for social capital accumulation in collectivist societies.

Role Analysis in Networks Using Mixtures of Exponential Random Graph Models

This article introduces a novel and flexible framework for investigating the roles of actors within a network. Particular interest is in roles as defined by local network connectivity patterns, identified using the ego-networks extracted from the network. A mixture of exponential-family random graph models (ERGM) is developed for these ego-networks to cluster the nodes into roles. We refer to this model as the ego-ERGM. An expectation-maximization algorithm is developed to infer the unobserved cluster assignments and to estimate the mixture model parameters using a maximum pseudo-likelihood approximation. We demonstrate the flexibility and utility of the method using examples of simulated and real networks.

High frequency functional brain networks in neonates revealed by rapid acquisition resting state fMRI

Understanding how spatially remote brain regions interact to form functional brain networks, and how these develop during the neonatal period, provides fundamental insights into normal brain development, and how mechanisms of brain disorder and recovery may function in the immature brain. A key imaging tool in characterising functional brain networks is examination of T2*-weighted fMRI signal during rest (resting state fMRI, rs-fMRI). The majority of rs-fMRI studies have concentrated on slow signal fluctuations occurring at <0.1 Hz, even though neuronal rhythms, and haemodynamic responses to these fluctuate more rapidly, and there is emerging evidence for crucial information about functional brain connectivity occurring more rapidly than these limits. The characterisation of higher frequency components has been limited by the sampling frequency achievable with standard T2* echoplanar imaging (EPI) sequences. We describe patterns of neonatal functional brain network connectivity derived using accelerated T2*-weighted EPI MRI. We acquired whole brain rs-fMRI data, at subsecond sampling frequency, from preterm infants at term equivalent age and compared this to rs-fMRI data acquired with standard EPI acquisition protocol. We provide the first evidence that rapid rs-fMRI acquisition in neonates, and adoption of an extended frequency range for analysis, allows identification of a substantial proportion of signal power residing above 0.2 Hz. We thereby describe changes in brain connectivity associated with increasing maturity which are not evident using standard rs-fMRI protocols. Development of optimised neonatal fMRI protocols, including use of high speed acquisition sequences, is crucial for understanding the physiology and pathophysiology of the developing brain.

A splitting brain: Imbalanced neural networks in schizophrenia

Dysconnectivity between key brain systems has been hypothesized to underlie the pathophysiology of schizophrenia. The present study examined the pattern of functional dysconnectivity across whole-brain neural networks in 121 first-episode, treatment-naïve patients with schizophrenia by using resting-state functional magnetic resonance imaging (rsfMRI). Group independent component analysis (ICA) was first applied to rsfMRI data to extract 90 functional components of the brain. The functional connectivity between these ICA components was then evaluated and compared between the patient and control groups. To examine the functional roles of significantly altered between-component connections in patients, each ICA component was ascribed to one of 10 previously well-defined brain networks/areas. Relative to findings in healthy controls (n=103), 29 altered functional connections including 19 connections with increased connectivity and 10 connections with decreased connectivity in schizophrenia patients were found. Increased connectivity was mainly within the default mode network (DMN) and between the DMN and cognitive networks, whereas decreased connectivity was predominantly associated with sensory networks. Given the key roles of the DMN in internal mental processes and sensory networks in inputs from the external environment, these patterns of altered brain network connectivity could suggest imbalanced neural processing of internal and external information in schizophrenia.

Differential relations between juvenile psychopathic traits and resting state network connectivity.

Traditionally, neurobiological research on psychopathy has focused on categorical differences in adults. However, there is evidence that psychopathy is best described by a set of relatively independent personality dimensions, that is, callous-unemotional, grandiose-manipulative, and impulsive-irresponsible traits, which can be reliably detected in juveniles, allowing investigation of the neural mechanisms leading to psychopathy. Furthermore, complex psychiatric disorders like psychopathy are increasingly being conceptualized as disorders of brain networks. The intrinsic organization of the brain in such networks is reflected by coherent fluctuations in resting state networks (RSNs), but these have not been studied in sufficient detail in relation to juvenile psychopathic traits yet. The current study investigated the distinct associations of juvenile psychopathic traits dimensions with RSN connectivity. Resting-state functional MRI and independent component analysis were used to assess RSN connectivity in a large sample of adolescents (n = 130, mean age 17.8 years) from a childhood arrestee cohort. Associations between scores on each of the three psychopathic traits dimensions and connectivity within and between relevant RSNs were investigated. Callous-unemotional traits were related to aberrant connectivity patterns of the default mode network, which has been implicated in self-referential and moral processes. Impulsive-irresponsible traits were associated with altered connectivity patterns in the frontoparietal cognitive control networks. Grandiose-manipulative traits were not associated with altered connectivity patterns. These findings confirm the association between psychopathic traits and brain network connectivity, and considerably add to emerging evidence supporting neurobiological heterogeneity in the processes leading to psychopathy.

The relation between structural and functional connectivity patterns in complex brain networks

ObjectiveAn important problem in systems neuroscience is the relation between complex structural and functional brain networks. Here we use simulations of a simple dynamic process based upon the susceptible–infected–susceptible (SIS) model of infection dynamics on an empirical structural brain network to investigate the extent to which the functional interactions between any two brain areas depend upon (i) the presence of a direct structural connection; and (ii) the degree product of the two areas in the structural network. MethodsFor the structural brain network, we used a 78×78 matrix representing known anatomical connections between brain regions at the level of the AAL atlas (Gong et al., 2009). On this structural network we simulated brain dynamics using a model derived from the study of epidemic processes on networks. Analogous to the SIS model, each vertex/brain region could be in one of two states (inactive/active) with two parameters β and δ determining the transition probabilities. First, the phase transition between the fully inactive and partially active state was investigated as a function of β and δ. Second, the statistical interdependencies between time series of node states were determined (close to and far away from the critical state) with two measures: (i) functional connectivity based upon the correlation coefficient of integrated activation time series; and (ii) effective connectivity based upon conditional co-activation at different time intervals. ResultsWe find a phase transition between an inactive and a partially active state for a critical ratio τ=β/δ of the transition rates in agreement with the theory of SIS models. Slightly above the critical threshold, node activity increases with degree, also in line with epidemic theory. The functional, but not the effective connectivity matrix closely resembled the underlying structural matrix. Both functional connectivity and, to a lesser extent, effective connectivity were higher for connected as compared to disconnected (i.e.: not directly connected) nodes. Effective connectivity scaled with the degree product. For functional connectivity, a weaker scaling relation was only observed for disconnected node pairs. For random networks with the same degree distribution as the original structural network, similar patterns were seen, but the scaling exponent was significantly decreased especially for effective connectivity. ConclusionsEven with a very simple dynamical model it can be shown that functional relations between nodes of a realistic anatomical network display clear patterns if the system is studied near the critical transition. The detailed nature of these patterns depends on the properties of the functional or effective connectivity measure that is used. While the strength of functional interactions between any two nodes clearly depends upon the presence or absence of a direct connection, this study has shown that the degree product of the nodes also plays a large role in explaining interaction strength, especially for disconnected nodes and in combination with an effective connectivity measure. The influence of degree product on node interaction strength probably reflects the presence of large numbers of indirect connections.

Distinct intrinsic network connectivity patterns of post-traumatic stress disorder symptom clusters.

Post-traumatic stress disorder (PTSD) is considered a multidimensional disorder, with distinct symptom clusters including re-experiencing, avoidance/numbing, hyperarousal, and most recently depersonalization/derealization. However, the extent of differing intrinsic network connectivity underlying these symptoms has not been fully investigated. We therefore investigated the degree of association between resting connectivity of the salience (SN), default mode (DMN), and central executive (CEN) networks and PTSD symptom severity. Using resting-state functional MRI data from PTSD participants (n = 21), we conducted multivariate analyses to test whether connectivity of extracted independent components varied as a function of re-experiencing, avoidance/numbing, hyperarousal, and depersonalization/derealization. Hyperarousal symptoms were associated with reduced connectivity of posterior insula/superior temporal gyrus within SN [peak Montréal Neurological Institute (MNI): -44, -8, 0, t = -4.2512, k = 40]. Depersonalization/derealization severity was associated with decreased connectivity of perigenual anterior cingulate/ventromedial prefrontal cortex within ventral anterior DMN (peak MNI: 8, 40, -4; t = -3.8501; k = 15) and altered synchrony between two DMN components and between DMN and CEN. Our results are consistent with prior research showing intrinsic network disruptions in PTSD and imply heterogeneous connectivity patterns underlying PTSD symptom dimensions. These findings suggest possible biomarkers for PTSD and its dissociative subtype.

Predicting learning plateau of working memory from whole-brain intrinsic network connectivity patterns.

Individual learning performance of cognitive function is related to functional connections within ‘task-activated' regions where activities increase during the corresponding cognitive tasks. On the other hand, since any brain region is connected with other regions and brain-wide networks, learning is characterized by modulations in connectivity between networks with different functions. Therefore, we hypothesized that learning performance is determined by functional connections among intrinsic networks that include both task-activated and less-activated networks. Subjects underwent resting-state functional MRI and a short period of training (80–90 min) in a working memory task on separate days. We calculated functional connectivity patterns of whole-brain intrinsic networks and examined whether a sparse linear regression model predicts a performance plateau from the individual patterns. The model resulted in highly accurate predictions (R2 = 0.73, p = 0.003). Positive connections within task-activated networks, including the left fronto-parietal network, accounted for nearly half (48%) of the contribution ratio to the prediction. Moreover, consistent with our hypothesis, connections of the task-activated networks with less-activated networks showed a comparable contribution (44%). Our findings suggest that learning performance is potentially constrained by system-level interactions within task-activated networks as well as those between task-activated and less-activated networks.

Low-frequency connectivity is associated with mild traumatic brain injury.

Mild traumatic brain injury (mTBI) occurs from a closed-head impact. Often referred to as concussion, about 20% of cases complain of secondary psychological sequelae, such as disorders of attention and memory. Known as post-concussive symptoms (PCS), these problems can severely disrupt the patient's quality of life. Changes in local spectral power, particularly low-frequency amplitude increases and/or peak alpha slowing have been reported in mTBI, but large-scale connectivity metrics based on inter-regional amplitude correlations relevant for integration and segregation in functional brain networks, and their association with disorders in cognition and behaviour, remain relatively unexplored. Here, we used non-invasive neuroimaging with magnetoencephalography to examine functional connectivity in a resting-state protocol in a group with mTBI (n = 20), and a control group (n = 21). We observed a trend for atypical slow-wave power changes in subcortical, temporal and parietal regions in mTBI, as well as significant long-range increases in amplitude envelope correlations among deep-source, temporal, and frontal regions in the delta, theta, and alpha bands. Subsequently, we conducted an exploratory analysis of patterns of connectivity most associated with variability in secondary symptoms of mTBI, including inattention, anxiety, and depression. Differential patterns of altered resting state neurophysiological network connectivity were found across frequency bands. This indicated that multiple network and frequency specific alterations in large scale brain connectivity may contribute to overlapping cognitive sequelae in mTBI. In conclusion, we show that local spectral power content can be supplemented with measures of correlations in amplitude to define general networks that are atypical in mTBI, and suggest that certain cognitive difficulties are mediated by disturbances in a variety of alterations in network interactions which are differentially expressed across canonical neurophysiological frequency ranges.

Parameter-dependent synchronization of coupled neurons in cold receptor model

In this paper, synchronization in two coupled neurons with spiking, bursting and chaos firings is investigated as the coupling strength gets increased. Synchronization state can be identified by means of the bifurcation diagram, the correlation coefficient and ISI-distance. It is illustrated that the coupled neurons can exhibit different types of synchronization state when the coupling strength increases. The different synchronization processes appear similar, but their detailed processes are different depending on the parameter values. The synchronization of neuronal network with two different network connectivity patterns is also studied. It is shown that chaotic and high period pattern are more difficult to get complete synchronization than the situation in single spike and low period pattern. It is also demonstrated that the synchronization status of multiple neurons is dependent on the network connectivity patterns. These results may be instructive to understand synchronization in neuronal systems.

Identifying and Mapping Connectivity Patterns of Brain Network Hubs in Alzheimer's Disease.

Alzheimer's disease (AD) is associated not only with regional gray matter damages, but also with abnormalities in functional integration between brain regions. Here, we employed resting-state functional magnetic resonance imaging data and voxel-based graph-theory analysis to systematically investigate intrinsic functional connectivity patterns of whole-brain networks in 32 AD patients and 38 healthy controls (HCs). We found that AD selectively targeted highly connected hub regions (in terms of nodal functional connectivity strength) of brain networks, involving the medial and lateral prefrontal and parietal cortices, insula, and thalamus. This impairment was connectivity distance-dependent (Euclidean), with the most prominent disruptions appearing in the long-range connections (e.g., 100-130 mm). Moreover, AD also disrupted functional connections within the default-mode, salience and executive-control modules, and connections between the salience and executive-control modules. These disruptions of hub connectivity and modular integrity significantly correlated with the patients' cognitive performance. Finally, the nodal connectivity strength in the posteromedial cortex exhibited a highly discriminative power in distinguishing individuals with AD from HCs. Taken together, our results emphasize AD-related degeneration of specific brain hubs, thus providing novel insights into the pathophysiological mechanisms of connectivity dysfunction in AD and suggesting the potential of using network hub connectivity as a diagnostic biomarker.

Default mode network connectivity patterns associated with visual processing at different stages of Parkinson's disease.

The default mode network (DMN) decreases its activity when switching from a resting state to a cognitive task condition, while activity of the network engaged in the given task increases. Visual processing is typically disturbed in Parkinson's disease dementia (PDD). Using functional MRI, we studied the DMN effective connectivity patterns in PDD as compared with cognitively normal patients with Parkinson's disease (PD) and healthy controls (HC) when switching from baseline to a visual cognitive task condition. In all, 14 PDD, 18 PD, and 18 age-matched healthy controls participated in this functional MRI study. We used a psychophysiological interaction analysis with the precuneus (PCu) as a seed. The threshold was set at p(FWE) <0.05. The healthy controls showed greater PCu connectivity with the bilateral middle temporal/middle occipital gyri at baseline than during the task condition. The correlation direction changed from positive to negative. Both PD and PDD showed disturbed DMN connectivity with the brain regions that are involved in bottom-up visual processing. In PD, we also found impaired integration of the areas engaged in the ventral attentional network, which might reflect specific attentional deficits observed during the early course of PD. In mild PDD, we detected increased engagement of areas involved in the dorsal attentional network, which corresponds to increased top-down control in this patient group as compared to the healthy controls. Our results show impaired dynamic interplay between large scale brain networks in PD that spread far beyond the motor system.

Disturbed default mode network connectivity patterns in Alzheimer's disease associated with visual processing.

Changes in connectivity of the posterior node of the default mode network (DMN) were studied when switching from baseline to a cognitive task using functional magnetic resonance imaging. In all, 15 patients with mild to moderate Alzheimer's disease (AD) and 18 age-, gender-, and education-matched healthy controls (HC) participated in the study. Psychophysiological interactions analysis was used to assess the specific alterations in the DMN connectivity (deactivation-based) due to psychological effects from the complex visual scene encoding task. In HC, we observed task-induced connectivity decreases between the posterior cingulate and middle temporal and occipital visual cortices. These findings imply successful involvement of the ventral visual pathway during the visual processing in our HC cohort. In AD, involvement of the areas engaged in the ventral visual pathway was observed only in a small volume of the right middle temporal gyrus. Additional connectivity changes (decreases) in AD were present between the posterior cingulate and superior temporal gyrus when switching from baseline to task condition. These changes are probably related to both disturbed visual processing and the DMN connectivity in AD and reflect deficits and compensatory mechanisms within the large scale brain networks in this patient population. Studying the DMN connectivity using psychophysiological interactions analysis may provide a sensitive tool for exploring early changes in AD and their dynamics during the disease progression.

SERIOUS GAMING ENHANCES COGNITIVE FUNCTION IN MCI DUE TO ALZHEIMER'S DISEASE

In this study we are assessing a novel ICT-based cognitive intervention, called Long Lasting Memories that combines physical and cognitive training, using a battery of neuropsychological assessments (like the California Verbal Learning Test (CVLT), the Trail Making Test A and B (TMT-A, TMT-B), the Digit Span test and other) and also resting-state electroencephalography in order to compare cognitive performance at the neuropsychological assessments with Default Mode Network (DMN) connectivity patterns, which are hypothesized to be impaired at patients with MCI due to Alzheimer's disease [2].

Model exchange with the NeuroML model database

The Neural Open Markup Language project, NeuroML, is an international, collaborative initiative to develop a language for describing and sharing complex, multiscale neuron and neuronal network models [1]. The project focuses on the key objects that need to be exchanged among software applications used by computational neuroscientists. Examples of these objects include descriptions of neuronal morphology, the dynamics of ion channels and synaptic mechanisms, and the connectivity patterns of networks of model neurons. This modular approach brings additional benefits: not only can entire models be published and exchanged in this format, but each individual object or component, such as a specific calcium channel or excitatory synapse, can be shared and re-implemented in a different model. The use of a standardized description language based on XML also facilitates the development of tools that promote simulator interoperability. The NeuroML Model Database (NeuroML-db.org) is a relational database that provides a means for exchanging multiscale NeuroML model descriptions and their components. There are several existing, curated model databases that include neuroscience models in multiple formats such as ModelDB [2] and the BioModels Database [3], which provide excellent resources for sharing diverse, published neuroscience models in multiple formats. In contrast, NeuroML-db is more focused, including only NeuroML models. This emphasis allows the database design and search to take advantage of this specific format, and in particular, allows for efficient searches over sub-components of models. In addition, the NeuroML Model Database can be used to search over public models developed as part of the Open Source Brain (OSB) initiative, which provides a software infrastructure for the collaborative development and evaluation of models [4]. NeuroML is the preferred model description format for OSB, and OSB models that are described using NeuroML version 2.0 can be flagged for automatic inclusion in NeuroML-db. An overarching design goal for the database is to provide a simple keyword search interface. Because many search terms involve terminology from neuroscience, we complement a metadata-based keyword search with a mechanism that can provide information about the biological meaning of the keywords, where this semantic information is available through an existing ontology for neuroscience called NeuroLex [5]. Through implicit query reformulation, this semantic information is used to expand the search results. For example, if the query is a single brain region, all cell and network models for cells and networks from that region and its sub-regions are provided to the user.