Abstract
Computational neuroscience has traditionally focused on isolated scales, limiting understanding of brain function across multiple levels. While microscopic models capture biophysical details of neurons, macroscopic models describe large-scale network dynamics. Integrating these scales, however, remains a significant challenge. In this study, we present a novel co-simulation framework that bridges these levels by integrating the neural simulator Arbor with The Virtual Brain (TVB) platform. Arbor enables detailed simulations from single-compartment neurons to populations of such cells, while TVB models whole-brain dynamics based on anatomical features and the mean neural activity of a brain region. By linking these simulators for the first time, we provide an example of how to model and investigate the onset of seizures in specific areas and their propagation to the whole brain. This framework employs an MPI intercommunicator for real-time bidirectional interaction, translating between discrete spikes from Arbor and continuous TVB activity. Its fully modular design enables independent model selection for each scale, requiring minimal effort to translate activity across simulators. The novel Arbor-TVB co-simulator allows replacement of TVB nodes with biologically realistic neuron populations, offering insights into seizure propagation and potential intervention strategies. The integration of Arbor and TVB marks a significant advancement in multi-scale modeling, providing a comprehensive computational framework for studying neural disorders and optimizing treatments.
1 Introduction
The human brain consists of billions of neurons and an equally vast population of non-neuronal cells, intricately organized into layers and regions (Herculano-Houzel, 2009, 2012). Each neuron operates as a highly sophisticated biochemical machinery (West et al., 2002; Augustine et al., 2003; Darnell, 2013; Lu et al., 2025), coordinating signal transmission within an extensive network in health (Reyes, 2003; Barral et al., 2019; Dicks, 2022) and disease (see e.g., Tetzlaff et al., 2025). Ever since the Hodgkin-Huxley model was introduced to describe membrane potential dynamics (Hodgkin and Huxley, 1952c), computational neuroscience has played a pivotal role in enhancing our understanding of brain function. Yet, due to the immense complexity of the brain and computational constraints, most modeling studies focus on a single scale simulator or rely on standalone simulation codes.
Modeling the large-scale electrical activity of the brain is a complex task. It not only demands familiarity with advanced mathematical methods, but also a solid grasp of the brain's physiology and anatomy. Neural field theory offers a way to study the nonlinear behavior of large groups of neurons at a population level, while still keeping the mathematics manageable. These models give us a strong theoretical framework for understanding key processes in neural tissue, including how the brain transitions between different activity states, such as those seen in sleep or during epileptic events, see e.g., Cook et al. (2022) for a recent review. Moreover, multi-scale computational modeling provides a framework for connecting neural mechanisms with measurements ranging from unit recordings to electroencephalogram (EEG), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI). Such models clarify how neural systems compute and interact, and they are essential for integrating empirical findings into a robust theoretical understanding of brain function, see e.g., Deco et al. (2008) and Cooray et al. (2023). Along this direction, recently in Cooray et al. (2025), the authors also studied oscillatory activity in cortical tissue arising from not uniform neural connections. and showed that oscillations can be maintained under a wide range of anisotropic and time-varying connectivity patterns.
Simulations incorporating biophysical properties and neural morphology typically concentrate on individual neurons using simulators such as the NEURON simulator (Carnevale and Hines, 2006). At the microscopic level, studies have explored questions such as how the ion channel kinetics influence neural excitability (see e.g., Gurkiewicz et al., 2011; Suma et al., 2024), how proteins, enzymes, and calcium concentration are distributed among neighboring spines to impact plasticity (see e.g., Luboeinski and Tetzlaff, 2021; Chater et al., 2024), and how signal propagation along axonal fibers relates to neuropathic pain (see e.g., Tigerholm et al., 2014, 2015). Some studies examine how neural morphology—such as dendritic tree growth (see e.g., Yasumatsu et al., 2008) and morphology-dependent plastic interactions (see e.g., Hananeia et al., 2024)—affects function. These studies, while often limited to small patches of the neural membrane, a few dendritic segments, or a small local network, provide valuable approximations of broader neural phenomena.
At the mesoscopic level, researchers simplify neuronal representations using point leaky-integrate-and-fire neurons [based on simulators such as NEST (Gewaltig and Diesmann, 2007) or Brian/Brian2 (Stimberg et al., 2019)], allowing studies on larger networks without explicit neuronal morphology or with some degree of self-customized morphology, using, e.g., NESTML (Linssen et al., 2024). This approach has advanced our understanding of neural heterogeneity (Demirtaş et al., 2019; Nayebi et al., 2021; Gast et al., 2024), self-organization (Zheng et al., 2013; Diaz-Pier et al., 2016; Miner and Triesch, 2016), neural capacity (Emina and Kropff, 2022), energy efficiency (Sacramento et al., 2015), and neural plasticity in disease and health (Manos et al., 2021; Lu et al., 2024). Most microscopic and mesoscopic models remain theory-driven, using mathematical approximations to infer neural behavior rather than directly establishing model based on large datasets (see e.g., Popovych et al., 2019 for a recent review).
Data-driven modeling has gained traction at the macroscopic level with the rise of open-source brain imaging databases [such as OpenfMRI (Poldrack and Gorgolewski, 2017)]. High-resolution structural and functional data from magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) enable whole-brain modeling based on real anatomical features. The Virtual Brain (TVB) (Sanz Leon et al., 2013; Sanz-Leon et al., 2015; Ritter et al., 2013) and Virtual Brain Twin (VBT) (Hashemi et al., 2025) platforms, for instance, integrate functional MRI and DTI datasets to build individualized models, using coupled oscillators to represent regional activity. TVB has contributed significantly to the understanding of neurological disorders and serves as a testing ground for therapeutic interventions (see e.g., Stefanovski et al., 2021; Monteverdi et al., 2023; Courson et al., 2024 and references therein), for studying self-organization on macroscale (see e.g., Fousek et al., 2024), consciousness (see e.g., Breyton et al., 2024) and healthy aging (see e.g., Lavanga et al., 2023).
With advances in computing resources and simulation technologies, the integration of models across different scales has become both feasible and essential to strike a balance between retaining detailed information and achieving a broad-scale understanding. Recently, a co-simulation framework was introduced that employs NEST and TVB to bridge mesoscopic and macroscopic modeling. This work has pioneered cross-scale modeling (Kusch et al., 2024) and has demonstrated the benefits of integrating models across spatial levels. A notable application of the virtual deep brain stimulation model (Meier et al., 2022; Shaheen et al., 2022; Wang et al., 2025) demonstrated its utility in multiscale simulations. Similar tools have been made available within the European digital neuroscience platform, EBRAINS (Schirner et al., 2022).
However, integrating microscopic and macroscopic models remains technically challenging. At the core lies the vast amount of information being processed, billions of cells with thousands of connections, and the immense gap in timescales, from microseconds in ion channel dynamics to minutes or hours for plastic changes of the connectome. To solve this challenge, we used the Arbor simulator (Abi Akar et al., 2019) and, more specifically, its most recent next-generation version (Cumming et al., 2024), at the microscopic end. Designed for single-neuron and large-scale network simulations, Arbor leverages GPU resources to enhance computational speed and energy efficiency.
In this Methods paper, we successfully established efficient communication between Arbor and the TVB that respects their different operational time scales and provided a use case example of the cross-scale interaction. To demonstrate a first showcase, we used a mouse brain connectome provided by TVB, where each region represents the mean mass neural activity of a brain area modeled by a macroscopic model. Using the co-simulation interface, we replaced one TVB node with a network of detailed neurons modeled in Arbor. An extended Hodgkin-Huxley-based neuron model (Depannemaecker et al., 2022) was utilized in Arbor to simulate different neural activity patterns (e.g., spiking, bursting, seizure-like, etc.). By tweaking a single parameter, we showcased that the seizure-like events generated in Arbor propagated to other nodes modeled in TVB. This platform provides users with the freedom to use existing models across scales with minimal additions and enabled future development in building brain digital twins that contain both micro- and macroscopic information for therapeutic applications.
2 Materials and methods
2.1 The Arbor simulator
Arbor is an open-source library for building simulations of biophysically detailed neuron models (Abi Akar et al., 2019). It provides an alternative to software like NEURON (Carnevale and Hines, 2006), but with a strong emphasis on modern hardware and scalability to large-scale systems (Hines, 1984). Its overall set of capabilities allows Arbor to model neural networks at a level of resolution beyond point models to explore phenomena like dendritic computation. Thus, support for bulk-synchronous parallelism via MPI, shared memory parallelism by utilizing a thread-pool and job system is central to Arbor, and certain cell types—primarily cable cells—can further leverage SIMD and GPU hardware. Arbor is written in C++, though most users interface with it through an intuitive, high-level Python interface built on top of the lower level implementation. The underlying numerical model of Arbor is the cable equation:
where the membrane potential U is computed over the morphological structure of the neural tree; the spatial coordinate x and the derivative are to be understood within this structure (Von Helmholtz, 1850; Thompson and Kelvin, 1855; Hodgkin et al., 1952; Hodgkin and Huxley, 1952a,b; Loligo, 1952; Hodgkin and Huxley, 1952c; Scott, 1975). The parameters c and σ define the membrane capacitance and longitudinal resistance. The trans-membrane current density i models the entirety of ionic and non-ionic currents. In both NEURON and Arbor, these are calculated from user-specified sets of differential equations, potentially varying along the morphology. The equations for i and U are solved in alternation (Lie-Trotter splitting) using a first-order implicit method.
2.2 Single neuron and network models in Arbor
We begin by selecting a dynamical model that allows for relatively easy yet realistic simulation of a broad spectrum of neural activity at the single-neuron level, governed by a small set of biophysical parameters. The neural model from Depannemaecker et al. (2022) was formulated for Arbor in the Neuron MODeling Language (NMODL). The following equations form the slow part of the system, describing the evolution of ion concentrations due to voltage-gated channels, active pumps, and buffering by an external bath, see Figure 1 for a schematic of the dynamics. It describes the ionic exchanges between the intracellular and extracellular spaces (ICS, ECS) of a neuron immersed within an external bath, acting as a potassium buffer of concentration Kbath. Ions flow between the ICS and ECS through a sodium-potassium pump and the sodium, potassium and chloride voltage-gated channels, driving changes in the internal (Ki, Nai, Cli) and external (Ko, Nao, Clo) ionic concentrations. By gradually increasing the external bath concentration of potassium ions Kbath, the model sequentially presents these patterns: resting state (RS), spike train (ST), tonic spiking (TS), bursting, seizure-like events (SLE), sustained ictal activity (SIA) and depolarization block (DB), see Figure 2. The fast dynamics of the membrane potential V is modeled in Arbor via the cable equations, see above, which require computing the ion current densities iX used in the simulator update as:
with the ion species X = {K, Na, Cl} and a non-ion current density:
These currents enter the cable equation Equation 1 as the trans-membrane current i via:
Following the Hodgkin-Huxley model, conductivities gX are written as:
The—internal i and external o—ion concentrations are modeled as:
The variables {ΔKi, Kg} evolve as:
where γ converts the current density iX to molar flux, summarizing the effect of the ion pump in Figure 1A and the external buffer. Finally, fast dynamics were reduced and adjusted to mammalian neurons:
based on the observations that the reaction of the sodium gating variable to changes in V is nigh instantaneous and h(t) + n(t) = const.
Figure 1
Figure 2
The resulting ion channel was added to a basic, spherical, single-compartment neuron. After implementing this biophysical model, we reproduced the firing patterns using the parameters of the reported model (Figures 2A–F), see also Depannemaecker et al. (2022) for more details and motivation for model parameter choices. Note that despite the values given in the original publication, neither the Arbor nor the published reference implementation produces the depolarization block pattern at Kbath = 20mM but only at around Kbath = 22.5mM. From here, a simple model network was developed, comprising N total cells, with a mixture of tonic f · N and SLE (1 − f) · N cells, where both sub-populations are assigned individual values for Kbath, sketched in Figure 1. Cells are connected using exponential synapses with an internal weight of w = 0.5 chosen to produce an activity similar to Rabuffo et al. (2025) which uses delta synapses.
In addition to the elementary spherical morphology, we also investigated two multi-compartmental neuron models. The first model included a single dendritic segment of 25μm, subdivided into 5μm compartments. The second model extended the dendrite into a randomly generated tree composed of 5μm compartments. To introduce variability across cells, the random number generator was seeded with each cell's unique identifier (see Figure 3 for examples). In future work, these synthetic morphologies will be replaced with reconstructions derived from neural imaging data available in public databases. Supporting such models in Arbor will require only a simple command to load per-cell morphology data from disk.
Figure 3
2.3 TVB network model
Following Deco et al. (2013), we used the reduced Wong-Wang model (Wang, 2002) to simulate resting-state activity and to investigate the dynamics of local brain regions embedded within a large-scale brain network. The mean firing rate H(xI) and mean synaptic gating variable SI of region I are described by:
where xI is the synaptic input to the I-th region, ω = 1 denotes the local excitatory recurrence, cIK is the strength of the structural connection from the local area I to K, and G is a global coupling strength. The parameters are set to the same values as those used in the TVB implementation. JN = 0.2609nA is the synaptic coupling of NMDA receptors and I0 = 0.33nA is the baseline external input. The kinetic parameters are τs = 100ms and γ = 0.641. The parameters of the input-output function H are a = 0.27nC−1, b = 0.108kHz, and d = 154ms. Depending on the tuning of G, the system exhibits a multi-stable regime, with steady states of high and low spiking activity. Here, we set G = 0.096.
2.4 Co-simulation framework of Arbor and TVB
Both Arbor and TVB offer support for attaching a second simulator to perform co-simulation, potentially at different scales. Co-simulation from TVB's viewpoint is the simpler technology of the two frameworks, since TVB is designed to execute as a single process. TVB allows for exchange of any variable relevant to the region models and any number of variables. The co-simulation partner is encapsulated in one or more TVB regions, called proxy nodes, see Figure 4. These proxies present a conforming interface to TVB, exchanging the salient variables as a table, one row per time-step, one column per variable. As TVB advances in lockstep on a global time-step, this is almost identical to normal operation. However, co-simulation introduces the concept of an ‘epoch’ to TVB, i.e., the length of time that conforms to the smallest delay τmin in the set of inter-region connections delays τIJ, with I and J referring to two connected regions. These delays are part of the connectome data used to construct a TVB simulation. In the case that a connectome contains zero-valued delays these must be replaced with a pre-defined finite value. Further, it is required that the time-step evenly divides τmin. Co-simulation thus can integrate all nodes' state, including the proxy, for one epoch τmin without exchanging data. This is correct as an event emanating from any region I at time t influences any other region J at time t + τIJ ≥ t + τmin. Only after an epoch, data need to be exchanged between the proxy and the rest of the TVB regions. A TVB–NEST demonstration has been published to showcase the interaction between a local network of spiking neurons and the whole-brain network dynamics (Kusch et al., 2024).
Figure 4
Arbor has a different design in terms of connectivity. Interaction between physically separate cells is mediated by action potentials, e.g., when the membrane potential triggered by dedicated sources crosses a configurable threshold. Cells are connected by wiring these sources to corresponding sinks like synapses via an abstract connection object comprising a delay and weight, modeling transmission and attenuation via an axon. In contrast to TVB, Arbor is fundamentally a distributed system and internally employs the same approach to decoupling via the minimum network delay as explained above. To initiate co-simulation, Arbor utilizes an additional interface to manage the spike exchange from external connections that are originated from outside but terminate at cells simulated in Arbor. On the technical side, the latter part leverages MPI_Allgatherv through an inter-communicator and effects that the concatenation of all spikes sent from all MPI ranks running TVB arrive on all ranks running Arbor and vice versa (Figure 4B). This allows co-simulation in conjunction with arbitrary numbers of ranks on both sides and even in compounds with more than two simulators.
Finally, bi-directional translation between TVB's variable concept and Arbor's representation of action potentials is required. As the former depends on the region models used, we chose to bundle this with the remaining TVB functionality as part of the Arbor proxy node. For the TVB models used in this study, the main variable is the per-region mean activity rate νI which is conceptually compatible with the concept of spike generation. For each region I connected to the proxy node P, i.e., with connectome weight cIP > 0, a set of synthetic events must be generated such that the mean activity conforms to νI. This is an ambiguous process, even if we prescribe a population (list of cell identifiers) and a per-cell distribution, e.g., a Poisson point process, from which to draw events, which likely must be resolved by ensembles of simulations. In general, this is both model dependent and mathematically intractable, so we leave the general case as a customization point in the framework. For our running example, however, we make the following choice: Event timings for the current step k will be drawn from a uniform distribution and dispatched to all cells in the Arbor network. Note that while these events are created at given time a per-connection delay is applied and thus delivery occurs at a later time.
The inverse direction, converting spike events to mean rates, while being well-defined, is still subject to customization. We explore two options here. First, simple running averages, i.e., all spikes that originate within the Arbor network during the current epoch, are collected and sorted into bins of width Δt. This list is then normalized to the cell count and time step and sent to TVB as the mean activities as a function of time. Although straightforward, this can lead to unrealistically rough activity traces, especially if cell populations are small. Second, as inspired by high-speed calcium imaging experiments (Grewe et al., 2010), a mechanism to track cell activity through calcium level is implemented as:
with per cell p having a decay parameter τ and a weight β. Computing the activity becomes the average:
yielding a smooth trace. This method of converting discrete spiking events into a continuous interval variable is also used in a few plasticity models recruiting a negative feedback control mechanism such as synaptic scaling (Van Rossum et al., 2000) and homeostatic structural plasticity (Butz and Van Ooyen, 2013; Diaz-Pier et al., 2016; Lu et al., 2024) models. The choice of the calcium kernel parameters is based on trial-and-error to match the output of Arbor and the magnitude of activity generated by TVB. Figure 5 compares the impact of this choice on the macro-scale network. In small networks and over short timescales defined by the epoch length as shown in the example, spiking activity occurs in noncontinuous bursts, which is dubious in conjunction with the smooth dynamics of the chosen TVB model. Figure 5A shows the propagation of this noncontinuous activity into the TVB regions, while using the Ca-like model (B) provides smooth dynamics in both the Arbor and TVB models. We thus will use the latter in all simulations from here on out. In general, both methods require scaling by the number of cells in the proxy region to arrive at a scale-free activity measure. A local scaling factor GA is used to convert between the activity of the detailed network and the activity of the region modeled in TVB. In general, GA needs to be adjusted to the choice of connectome and TVB model, similar to the choice of the global coupling strength G in the RWW model. In this study, GA = 100 is used as it produces seizure-like propagation patterns comparable to those found in similar studies, see e.g., Melozzi et al. (2017) and Courson et al. (2024) and references therein.
Figure 5
3 Results
So far, we have described the components of the co-simulation framework. It consists of the following components: a TVB model based on the connectome and node dynamics, a specified set of TVB nodes where the Arbor models will be placed, one or more internally connected network models in Arbor, a defined mechanism for routing events from TVB to individual cells in Arbor, a method for translating Arbor-generated events into TVB variables, and a translation process for converting TVB variables into events originating from synthetic cells. Each of these components serves as a customization point for the user. While reasonable default configurations can be provided for some, others require user-defined specifications to suit specific modeling needs.
The single cell model has been demonstrated to exhibit the necessary range of behaviors. We have also motivated our choice for converting spikes to rates of using a biologically-inspired exponential smoothing filter via a Ca-like activity over simple binning and fix parameters to τ = 100ms and . This normalization is important, as it produces results that are invariant under changes in the number of cells N in the detailed network.
To induce seizure propagation in mice brain, we use the structural connectivity derived from the Allen mouse brain atlas (Oh et al., 2014), also used in Melozzi et al. (2017), to embed the TVB nodes. Following the work presented in Courson et al. (2024), the proxy node modeling the Arbor population of tonic-spiking and SLE point neurons is set within left the Hippocampus and in particular in the left-field CA1 (l CA1), a region that is prone to generate widespread seizures.
3.1 Seizure induction among morphologically detailed cells
As an initial step, we examined a network of morphologically detailed cells (comprising of a single-compartment soma and a random dendritic tree, see bottom-right inset panel of Figure 6) without embedding them into a co-simulation framework. The objective was to assess how interactions among neurons exhibiting distinct firing patterns influence network dynamics within a small population of SLE neurons at the local (Arbor) level. We constructed a network comprising 80 tonic-spiking neurons (Kbath = 9.5mM) and 20 SLE neurons (Kbath = 17.0mM), initially without internal synaptic connectivity. The system was simulated for T = 5s, and the resulting membrane potentials are presented in Figure 6 (this initial phase is indicated by the gray-shaded region). During this initial phase, all neurons independently showed their intrinsic firing behavior, consistent with the activity shown in Figure 6. Following this baseline simulation, integration was paused, the network was reconfigured to full connectivity, and the simulation was resumed. The introduction of connectivity produced an immediate response across the network. The SLE neurons progressively transitioned toward tonic-like spiking, mirroring the dominant (80%) tonic-spiking subpopulation. At the same time, tonic-spiking neurons began to exhibit the bursting activity characteristic of the SLE phenotype (Figure 6, T ≥ 5s)). Note that the use of different Kbath values for neurons within the same group is, at this stage, driven by computational considerations and the need to validate the performance of the Arbor-TVB co-simulator. This choice can be adjusted to study more biologically realistic scenarios which involve the interaction of different neural populations or transition of network dynamics, which is further addressed in the Discussion section.
Figure 6
3.2 Seizure induction and propagation employing point neurons in Arbor-TVB co-simulator
Next, we embedded a network of detailed neurons as a proxy node in a simulation of neural mass models in TVB. The general setup is similar as before, however, the proxy node now consists of only SLE-type point neurons in a fully connected network. Figure 7A illustrates the evolution of the membrane potential of individual neurons within the Arbor network, with all neurons exhibiting identical dynamical behavior. The pattern of SLE activity is modulated by neuronal coupling.
Figure 7
We run simulations over 20s, and investigate the effect of SLE emergence in the network once all brain areas have reached their steady state. The production of these patterns in the Arbor node generates changes in the firing rates of local TVB brain areas. Even though changes in activity occur in most brain areas, these fluctuations occur within different ranges.
In Figure 7B, we show the time-series of TVB nodes' firing rate throughout the simulation, here with N = 100 in the Arbor detailed network. A transient period is necessary before all nodes reach their steady-state. The emergence of recurrent SLE patterns in the Arbor node triggers a mixture of periodic and seizure-like patterns in the dynamics of the TVB nodes. We highlight traces corresponding to the l CA1 region modeled using detailed cells (orange) and the four regions with the highest activity modeled by the neural mass model. The inset shows the highlighted traces, excluding l CA1, during a single period.
In Figure 8, we present the propagation of SLE originating in l CA1 (star marker), represented as a fully connected network of N = 10000 SLE point neurons. Colors on the brain template show the time distance between seizure emergence in the diseased area and seizure arrival in the different brain regions. The systematic detection of SLE onset in other regions of the whole-brain network (if and when present within the simulation period) is performed as follows. First, the baseline firing activity of each region is defined (first time window of the simulation). The non-SLE nodes initially exhibit a relatively stable firing rate, which gradually becomes influenced by the SLE originating from the Arbor node. When connected to the Arbor node, spiking-tonic brain areas are repeatedly recruited into SLE activity patterns, followed by periods of relaxation. For each region, we define the baseline firing rate as the minimum firing rate observed during the relaxation phase immediately preceding any event in the onset region. When a robust increase in firing rate is detected in a region—indicative of a seizure-like event—the corresponding onset timestamp is recorded. The baseline firing rate serves as a reference, and seizure activity is identified based on a sustained increase in firing rate over a defined duration. We define the onset of a seizure as the beginning of the high-amplitude bursting regime, specifically the point at which the firing rate begins to rise consistently after the first small activity peak. To ensure that only significant deviations are classified as seizures, we require the firing rate to exceed the baseline by at least 10% and remain above this threshold for a minimum of 100 ms. Our seizure onset detection follows a similar (though not identical) approach to that used in related studies—for example, Melozzi et al. (2017) and Courson et al. (2024)—where seizure-like or bursting activity in mouse brain models is identified by monitoring a model variable and applying a threshold to detect the onset. In Figure 8, we also depict the activity time-series of four initially non-SLE nodes of the brain network being recruited in the seizure, namely the ventral part of left Lateral Septal Nucleus (l LSv), l CA3, l ENTCl and r ENTCl. In the main panel (mouse brain template), non-colored regions correspond to areas where seizures either did not occur or had relatively weak effects.
Figure 8
The Arbor network was initialized with fixed connection weights (w = 0.5). To assess robustness, we also tested weights drawn from a normal distribution (μ = 0.5, σ = 0.5), truncated to positive values. Neural activity was averaged over 20 independent realizations (see Supplementary Figures S2, S3). Experiments with inhibitory/excitatory network variants are also currently in progress. Both baseline and seizure-evoked activity differ across brain regions. Due to the symmetrical inter-hemispheric connections in the Allen Mouse Brain SC, l ENTCl and r ENTCl share the same baseline firing rate. As the SLE pattern emerges in the left hippocampus, l ENTCl exhibits higher spiking rates.
3.3 Computational performance of the Arbor-TVB co-simulation framework
We next evaluated the performance of the running example on a single Apple M1 (2021) laptop. Arbor was built with MPI and SIMD (Arm Neon/SVE) support, with cells organized into groups of ten to fully exploit SIMD capabilities. The overall runtime consists of four primary components: (1) Arbor model update, (2) Conversion from spikes to rates, (3) TVB model update, and (4) Conversion from rates to spikes.
The Arbor update runs in parallel with the conversions between rates and spikes, as well as the TVB update. During spike exchange, both simulations synchronize, meaning the slower part must wait in the MPI collective, which accounts for the primary time spent in the collective call. Figure 9 illustrates the total runtime of a 10s simulation for the entire model described above, along with the relative contributions, for system sizes ranging from one to 10,000 cells. Notably, at 10,000 cells, nearly all computational time is spent within the Arbor network model. In future experiments, we plan to leverage additional hardware, including GPUs, to accelerate the Arbor side of the simulation. At this scale, TVB and the spike/rate conversions are potential bottlenecks that will require optimization, potentially through TVB's JIT compilation and GPU acceleration. Additionally, further parallelization and porting of the conversion steps to a more performant programming environment remain promising avenues for improvement.
Figure 9
4 Discussion
In this work, we presented a co-simulation framework that offers a novel approach to bridging the gap between microscopic (spiking neuron) and macroscopic (mean-field) models. This framework integrates simulators Arbor and TVB within a parallelized MPI environment, enabling a detailed yet computationally feasible representation of neural dynamics across scales. To model large-scale brain network dynamics, we used the mean-field reduced Wong-Wang model to reproduce resting-state dynamics. Simultaneously, a detailed spiking neural network was simulated with Arbor, employing a physiological model of seizures at the neuron level (Figure 2). The spiking activity of the population was then converted into a smooth trace for the proxy node, which was communicated with TVB (Figure 4B). This co-simulation approach successfully captured the interplay between spiking activity and large-scale brain dynamics, where local neuron dynamics generate global activity wave-fronts. At the microscopic scale, we demonstrated that the structure of the detailed neural network influences its activity patterns, thereby affecting the shape of the activity wavefront (Figure 5).
As a proof of concept, and as a technical showcase, we simulated the emergence of seizure-like activity patterns (SLE) in the mouse hippocampus, using the Allen Mouse Brain Structural Connectivity data. By tuning the single neural parameter, we modeled the target brain area with a small, fully-connected network of (point and detailed-compartmental) neurons SLE, which produces scale-free activity patterns. This network design can be further adapted to investigate more bio-inspired scenarios, such as the study of: (i) ensembles of neurons within a given brain area that belong to different sub-regions (i.e., are relatively distant from one another), whose dynamical activity may differ significantly from that of other sub-groups and (ii) dynamical transition phenomena where the Kbath value is set near a critical threshold—e.g., inducing a transition from regular spiking to bursting activity, see Figures 2A, B. The latter consideration is more general, in the sense that for a different dynamical model of neural activity, one can choose a relevant control parameter that drives transitions in neural activity patterns—for example, from slow spiking (representing healthy activity) to fast bursting (associated with pathological activity) in a given region, when the system operates near a critical transition point (e.g., from spike trains to bursting). Our approach offers a well-understood and easily controlled platform by showcasing its technical usability. By adapting the Arbor nodes with different cell composition and connectivity, we demonstrated its significant potential for more complex cell models untapped.
Although using seizure propagation as a use case, the seizure activity patterns established in the present study may not capture precisely the complex nature of all epileptic seizures. For example, there are other types of seizures that occur in different brain disorders, i.e., acute symptomatic seizures, which are not like those caused in epilepsy. For example, patients with Alzheimer's disease may also experience seizures, which are classified as progressive symptomatic seizures and typically arise from underlying neurodegenerative processes., see e.g., (Mauritz et al., 2022) for a recent relevant review.
The implementation in the present study extended the functionality and application scenarios of Arbor. Arbor has embarked on many types of computational studies as a new-generation simulator. It enables seamless conversion and simulation of single-neuron models from the NEURON simulator and supports simulations of both individual neurons and large-scale networks. Arbor accommodates various plasticity models, including spike-timing-dependent plasticity, calcium-based synaptic tagging and capture, and structural plasticity. It has been used to study synaptic tagging and capture via the built-in diffusion functionality (Luboeinski et al., 2024, under review). Recent developments focus on co-simulation with membrane dynamics and external kernels, enabling dynamic connectivity modifications in a distance-dependent manner. With its high flexibility and scalability, Arbor stands out as a promising platform developed within the EBRAINS initiative to advance cross-scale simulations in computational neuroscience. Arbor is available as part of the EBRAINS software distribution (ESD) on connected HPC centers and the EBRAINS collab via Jupyter Lab. Providing a bridge between morphologically detailed neurons and neural mass models encompassing the full brain spans a gap of scale from sub-micrometer to decimeter. It allows for placing the resolution—using Arbor and detailed models—where needed and using realistic, data driven environments everywhere else via TVB.
Despite its successes, the Arbor-TVB framework has some limitations. The co-simulation requires careful exploration and calibration of coupling parameters to ensure meaningful interactions between Arbor and TVB, which remains a challenge when generalizing to diverse neural models. The TVB network we used here is homogeneous, in the sense that all model parameters are set to be identical. While this is not a highly realistic assumption—particularly when assigning a dynamical mean field model to simulate the activity of a brain region—it is a common choice among researchers when modeling whole-brain resting-state dynamics, see e.g., Popovych et al. (2021) and Manos et al. (2023) and references therein. Note that even with identical initial settings, firing rates vary across regions due to the influence of long-distance connectivity weight values and the resulting complex dynamics. Our use-case simulation of seizure propagation is not yet directly compared to experimental data from mice or humans. Nevertheless, the current Arbor-TVB implementation is capable of indirectly capturing biologically realistic brain dynamics and activity propagation, as reported for example in Melozzi et al. (2017) and Courson et al. (2024). The same applies to the choice of calcium kernel parameters. This kernel is inspired by experimental observations via calcium imaging experiments (Grewe et al., 2010) but no exact values are available. Due care should be taken when matching the magnitudes of both Arbor output and TVB output. Moreover, within the Arbor-TVB framework, the chosen dynamical model can be further tuned to generate neural activity—such as BOLD signals or firing rates—that more closely aligns with neuroimaging data, thereby enabling the simulation of more realistic brain dynamics. Specifically, computational costs for large neural networks may necessitate further optimization in model parallelization and data handling. A near-term goal would be to incorporate new features, such as synaptic plasticity, which could offer valuable insights into how brain networks adapt and reorganize in response to disrupted activity.
As stated in the Introduction section, it is natural to compare the Arbor-TVB co-simulation framework presented in the current Method paper with the established NEST-TVB co-simulation (Kusch et al., 2024). Both simulators have their own merits and application scope. NEST has a longer history (Gewaltig and Diesmann, 2007) and a rich profile of neural models and plasticity models. Its performance is excellent in mesoscopic modeling of spiking neural networks. Arbor is younger but has made it possible to efficiently simulate networks of the highly expensive biophysical HH model with multi-compartments (Abi Akar et al., 2019). It also accommodates a variety of plasticity rules, including heterosynaptic dendritic plasticity rules inside the dendritic shaft (Luboeinski et al., 2024). The Arbor-TVB co-simulation framework has thereby inherited those merits of Arbor, while the NEST-TVB co-simulation framework opens the venue for users to freely use NEST functions for co-simulation. Those two tools are complementary for distinct research purposes. Moreover, since Arbor is designed to make the best use of both GPU and CPU, Arbor-TVB can also be easily adapted to make use of the cutting-edge exascale GPU computing resources.
From an epilepsy-seizure perspective, while the framework provides insights into seizure propagation, additional validation against empirical data would enhance its applicability to clinical settings. This co-simulation framework could enable a detailed investigation of the physiological sources of seizures. Understanding the impact of the structure of the diseased area on seizure patterns and propagation would be of great interest (see e.g., Netoff et al., 2004; Garcia-Ramos et al., 2016). Specifically, we expect the inhibition and excitation ratios in the detailed neural network to play a critical role in seizure patterns (see e.g., Engel, 1996; Liu et al., 2020). Moreover, the Arbor-TVB user can implement various neural models and configuration topologies to simulate different brain regions and to computationally investigate diverse dynamic activities or the effects of medical interventions–for example, modeling subthalamic neurons along with synaptic and structural plasticity under stimulation in Parkinson's disease (Manos et al., 2021; Meier et al., 2022; Shaheen et al., 2022), Alzheimer's disease (Stefanovski et al., 2019; Manos et al., 2023) or tinnitus (Manos et al., 2018a,b) etc. Evidently, comparison, parameter tuning, and validation are also feasible using empirical neuroimaging data, however this was not the primary goal of this work. The simulated neural activity generated in each brain region of the connectome can be transformed into a BOLD signal-similar to the built-in functionality of TVB (Sanz Leon et al., 2013; Melozzi et al., 2017), which computes the hemodynamic response function (HRF) kernel (i.e., fMRI activity) using the Balloon-Windkessel model (Friston et al., 2000)—and can ultimately be aligned with neuroimaging time series data. Hence, a framework like Arbor-TVB can be extended to investigate various brain conditions.
Statements
Data availability statement
The codes necessary to reproduce selected results of this work are available on GitHub: https://github.com/arbor-contrib/arbor-tvb-cosim.
Author contributions
TH: Writing – review & editing, Software, Validation, Resources, Writing – original draft, Formal analysis, Data curation, Methodology, Investigation, Visualization. JC: Investigation, Writing – review & editing, Software, Writing – original draft, Validation, Visualization, Data curation, Methodology, Formal analysis. HL: Writing – review & editing, Investigation, Validation, Writing – original draft, Formal analysis, Methodology, Data curation, Software, Visualization. SD-P: Project administration, Funding acquisition, Validation, Conceptualization, Supervision, Writing – review & editing, Methodology. TM: Validation, Formal analysis, Methodology, Conceptualization, Supervision, Project administration, Resources, Writing – original draft, Writing – review & editing, Funding acquisition.
Funding
The author(s) declared that financial support was received for this work and/or its publication. JC was supported by the LABEX MME-DII (ANR-16-IDEX-0008) PhD grant. This work was supported by the NIH 1R21AG087888-01 grant. HL was supported by EBRAINS2.0. EBRAINS 2.0 has received funding from the European Union's Research and Innovation Program Horizon Europe under Grant Agreement No. 101147319. Open access publication funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—491111487.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors TM, HL declared that they were an editorial board member of Frontiers, at the time of submission. This had no impact on the peer review process and the final decision.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncom.2025.1731161/full#supplementary-material
References
1
Abi AkarN.CummingB.KarakasisV.KüstersA.KlijnW.PeyserA.et al. (2019). “Arbor – a morphologically-detailed neural network simulation library for contemporary high-performance computing architectures,” in 2019 27th Euromicro International Conference on Parallel, Distributed and Network-Based Processing (PDP) (New York City, NY: IEEE), 274–282.
2
AugustineG. J.SantamariaF.TanakaK. (2003). Local calcium signaling in neurons. Neuron40, 331–346. doi: 10.1016/S0896-6273(03)00639-1
3
BarralJ.WangX.-J.ReyesA. D. (2019). Propagation of temporal and rate signals in cultured multilayer networks. Nat. Commun. 10:3969. doi: 10.1038/s41467-019-11851-0
4
BreytonM.FousekJ.RabuffoG.SorrentinoP.KuschL.MassiminiM.et al. (2024). Spatiotemporal brain complexity quantifies consciousness outside of perturbation paradigms. eLife, 13:RP98920. doi: 10.7554/eLife.98920
5
ButzM.Van OoyenA. (2013). A simple rule for dendritic spine and axonal bouton formation can account for cortical reorganization after focal retinal lesions. PLoS Comput. Biol. 9:e1003259. doi: 10.1371/journal.pcbi.1003259
6
CarnevaleN. T.HinesM. L. (2006). The NEURON Book. Cambridge: Cambridge University Press.
7
ChaterT. E.EgglM. F.GodaY.TchumatchenkoT. (2024). Competitive processes shape multi-synapse plasticity along dendritic segments. Nat. Commun. 15:7572. doi: 10.1038/s41467-024-51919-0
8
CookB. J.PetersonA. D. H.WoldmanW.TerryJ. R. (2022). Neural field models: A mathematical overview and unifying framework. Mathem. Neurosci. Appl. 2, 1–67. doi: 10.46298/mna.7284
9
CoorayG. K.CoorayV.FristonK. J. (2025). Cortical dynamics of neural-connectivity fields. J. Comput. Neurosci. 53, 373–391. doi: 10.1007/s10827-025-00903-8
10
CoorayG. K.RoschR. E.FristonK. J. (2023). Global dynamics of neural mass models. PLoS Comput. Biol. 19:e1010915. doi: 10.1371/journal.pcbi.1010915
11
CoursonJ.QuoyM.TimofeevaY.ManosT. (2024). An exploratory computational analysis in mice brain networks of widespread epileptic seizure onset locations along with potential strategies for effective intervention and propagation control. Front. Comput. Neurosci. 18:1360009. doi: 10.3389/fncom.2024.1360009
12
CummingB.YatesS.HaterT.LuH.HuismanB.WouterK.et al. (2024). Arbor v0.10.0. Zenodo.
13
DarnellR. B. (2013). Rna protein interaction in neurons. Annu. Rev. Neurosci. 36, 243–270. doi: 10.1146/annurev-neuro-062912-114322
14
DecoG.JirsaV. K.RobinsonP. A.BreakspearM.FristonK. (2008). The dynamic brain: from spiking neurons to neural masses and cortical fields. PLoS Comput. Biol. 4:e1000092. doi: 10.1371/journal.pcbi.1000092
15
DecoG.Ponce-AlvarezA.MantiniD.RomaniG. L.HagmannP.CorbettaM. (2013). Resting-state functional connectivity emerges from structurally and dynamically shaped slow linear fluctuations. J. Neurosci. 33, 11239–11252. doi: 10.1523/JNEUROSCI.1091-13.2013
16
DemirtaşM.BurtJ. B.HelmerM.JiJ. L.AdkinsonB. D.GlasserM. F.et al. (2019). Hierarchical heterogeneity across human cortex shapes large-scale neural dynamics. Neuron101:1181–1194. doi: 10.1016/j.neuron.2019.01.017
17
DepannemaeckerD.IvanovA.LilloD.SpekL.BernardC.JirsaV. (2022). A unified physiological framework of transitions between seizures, sustained ictal activity and depolarization block at the single neuron level. J. Comput. Neurosci. 2022, 1–17. doi: 10.1007/s10827-022-00811-1
18
Diaz-PierS.NaveauM.Butz-OstendorfM.MorrisonA. (2016). Automatic generation of connectivity for large-scale neuronal network models through structural plasticity. Front. Neuroanat. 10:57. doi: 10.3389/fnana.2016.00057
19
DicksL. M. (2022). Gut bacteria and neurotransmitters. Microorganisms10:1838. doi: 10.3390/microorganisms10091838
20
EminaF.KropffE. (2022). Selective connectivity enhances storage capacity in attractor models of memory function. Front. Systems Neurosci. 104:983147. doi: 10.3389/fnsys.2022.983147
21
EngelJ. (1996). Excitation and inhibition in epilepsy. Canadian J.Neurol. Sci. 23, 167–174. doi: 10.1017/S0317167100038464
22
FousekJ.RabuffoG.GudibandaK.SheheitliH.PetkoskiS.JirsaV. (2024). Symmetry breaking organizes the brain's resting state manifold. Sci. Rep. 14:31970. doi: 10.1038/s41598-024-83542-w
23
FristonK.MechelliA.TurnerR.PriceC. (2000). Nonlinear responses in fmri: The balloon model, volterra kernels, and other hemodynamics. Neuroimage12:466–477. doi: 10.1006/nimg.2000.0630
24
Garcia-RamosC.LinJ. J.KellermannT. S.BonilhaL.PrabhakaranV.HermannB. P. (2016). Graph theory and cognition: a complementary avenue for examining neuropsychological status in epilepsy. Epilepsy Behav. 64, 329–335. doi: 10.1016/j.yebeh.2016.02.032
25
GastR.SollaS. A.KennedyA. (2024). Neural heterogeneity controls computations in spiking neural networks. Proc. Nat. Acad. Sci. 121:e2311885121. doi: 10.1073/pnas.2311885121
26
GewaltigM.-O.DiesmannM. (2007). Nest (neural simulation tool). Scholarpedia2:1430. doi: 10.4249/scholarpedia.1430
27
GreweB. F.LangerD.KasperH.KampaB. M.HelmchenF. (2010). High-speed in vivo calcium imaging reveals neuronal network activity with near-millisecond precision. Nat. Methods7, 399–405. doi: 10.1038/nmeth.1453
28
GurkiewiczM.KorngreenA.WaxmanS. G.LampertA. (2011). Kinetic modeling of nav1. 7 provides insight into erythromelalgia-associated f1449v mutation. J Neurophysiol. 105, 1546–1557. doi: 10.1152/jn.00703.2010
29
HananeiaN.EbnerC.GalanisC.CuntzH.OpitzA.VlachosA.et al. (2024). Multi-scale modelling of location-and frequency-dependent synaptic plasticity induced by transcranial magnetic stimulation in the dendrites of pyramidal neurons. bioRxiv. doi: 10.1101/2024.07.03.601851
30
HashemiM.DepannemaeckerD.SaggioM.TriebkornP.RabuffoG.FousekJ.et al. (2025). Principles and operation of virtual brain twins. IEEE Rev Biomed Eng. 1–29. doi: 10.1109/RBME.2025.3562951
31
Herculano-HouzelS. (2009). The human brain in numbers: a linearly scaled-up primate brain. Front. Hum. Neurosci. 3:857. doi: 10.3389/neuro.09.031.2009
32
Herculano-HouzelS. (2012). The remarkable, yet not extraordinary, human brain as a scaled-up primate brain and its associated cost. Proc. Nat. Acad. Sci. 109, 10661–10668. doi: 10.1073/pnas.1201895109
33
HinesG. S. (1984). Advances in simulation software technology. Simulation42, 36–36. doi: 10.1177/003754978404200105
34
HodgkinA. L.HuxleyA. F. (1952a). The components of membrane conductance in the giant axon of loligo. J. Physiol. 116:473. doi: 10.1113/jphysiol.1952.sp004718
35
HodgkinA. L.HuxleyA. F. (1952b). Currents carried by sodium and potassium ions through the membrane of the giant axon of loligo. J. Physiol. 116:449. doi: 10.1113/jphysiol.1952.sp004717
36
HodgkinA. L.HuxleyA. F. (1952c). A quantitative description of membrane current and its application to conduction and excitation in nerve. J. Physiol. 117:500. doi: 10.1113/jphysiol.1952.sp004764
37
HodgkinA. L.HuxleyA. F.KatzB. (1952). Measurement of current-voltage relations in the membrane of the giant axon of loligo. J. Physiol. 116:424. doi: 10.1113/jphysiol.1952.sp004716
38
KuschL.Diaz-PierS.KlijnW.SontheimerK.BernardC.MorrisonA.et al. (2024). Multiscale co-simulation design pattern for neuroscience applications. Front. Neuroinform. 18:1156683. doi: 10.3389/fninf.2024.1156683
39
LavangaM.StummeJ.YalcinkayaB. H.FousekJ.JockwitzC.SheheitliH.et al. (2023). The virtual aging brain: Causal inference supports interhemispheric dedifferentiation in healthy aging. Neuroimage283:120403. doi: 10.1016/j.neuroimage.2023.120403
40
LinssenC.BabuP. N.BouhadjarY.EwertL.WyboW.LoberM.et al. (2024). Nestml 8.0.0. Zenodo.
41
LiuY.GrigorovskyV.BardakjianB. (2020). Excitation and inhibition balance underlying epileptiform activity. IEEE Trans. Biomed. Eng. 67, 2473–2481. doi: 10.1109/TBME.2019.2963430
42
LoligoA. (1952). The dual effect of membrane potential on sodium conductance in the giant. J. Physiol. 6, 497–506. doi: 10.1113/jphysiol.1952.sp004719
43
LuH.DiazS.LenzM.VlachosA. (2024). The interplay between homeostatic synaptic scaling and homeostatic structural plasticity maintains the robust firing rate of neural networks. eLife. doi: 10.7554/eLife.88376.2
44
LuH.GargS.LenzM.VlachosA. (2025). Repetitive magnetic stimulation with itbs600 induces persistent structural and functional plasticity in mouse organotypic slice cultures. bioRxiv. doi: 10.2139/ssrn.5203072
45
LuboeinskiJ.SchmittS.ShafieeS.HaterT.BöschF.TetzlaffC. (2024). Plastic arbor: a modern simulation framework for synaptic plasticity-from single synapses to networks of morphological neurons. arXiv [preprint] arXiv:2411.16445. doi: 10.48550/arXiv.2411.16445
46
LuboeinskiJ.TetzlaffC. (2021). Memory consolidation and improvement by synaptic tagging and capture in recurrent neural networks. Commun. Biol. 4:275. doi: 10.1038/s42003-021-01778-y
47
ManosT.Diaz-PierS.FortelI.DriscollI.ZhanL.LeowA. (2023). Enhanced simulations of whole-brain dynamics using hybrid resting-state structural connectomes. Front. Comput. Neurosci. 17:1295395. doi: 10.3389/fncom.2023.1295395
48
ManosT.Diaz-PierS.TassP. A. (2021). Long-term desynchronization by coordinated reset stimulation in a neural network model with synaptic and structural plasticity. Front. Physiol. 12:716556. doi: 10.3389/fphys.2021.716556
49
ManosT.ZeitlerM.TassP. A. (2018a). How stimulation frequency and intensity impact on the long-lasting effects of coordinated reset stimulation. PLoS Comput. Biol. 14(5). doi: 10.1371/journal.pcbi.1006113
50
ManosT.ZeitlerM.TassP. A. (2018b). Short-term dosage regimen for stimulation-induced long-lasting desynchronization. Front. Physiol. 9:376. doi: 10.3389/fphys.2018.00376
51
MauritzM.HirschL. J.CamfieldP.ChinR.NardoneR.LattanziS.et al. (2022). Acute symptomatic seizures: an educational, evidence-based review. Epileptic Disord. 24, 26–49. doi: 10.1684/epd.2021.1376
52
MeierJ. M.PerdikisD.BlickensdörferA.StefanovskiL.LiuQ.MaithO.et al. (2022). Virtual deep brain stimulation: Multiscale co-simulation of a spiking basal ganglia model and a whole-brain mean-field model with the virtual brain. Exp. Neurol. 354:114111. doi: 10.1016/j.expneurol.2022.114111
53
MelozziF.WoodmanM.JirsaV.BernardC. (2017). The virtual mouse brain: A computational neuroinformatics platform to study whole mouse brain dynamics. eNeuro. 4. doi: 10.1523/ENEURO.0111-17.2017
54
MinerD.TrieschJ. (2016). Plasticity-driven self-organization under topological constraints accounts for non-random features of cortical synaptic wiring. PLoS Comput. Biol. 12:e1004759. doi: 10.1371/journal.pcbi.1004759
55
MonteverdiA.PalesiF.SchirnerM.ArgentinoF.MeranteM.RedolfiA.et al. (2023). Virtual brain simulations reveal network-specific parameters in neurodegenerative dementias. Front. Aging Neurosci. 15:1204134. doi: 10.3389/fnagi.2023.1204134
56
NayebiA.AttingerA.CampbellM.HardcastleK.LowI.MalloryC. S.et al. (2021). “Explaining heterogeneity in medial entorhinal cortex with task-driven neural networks,” in Advances in Neural Information Processing Systems, eds. M. Ranzato, A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan (Red Hook, NY: Curran Associates Inc), 12167–12179.
57
NetoffT. I.ClewleyR.ArnoS.KeckT.WhiteJ. A. (2004). Epilepsy in small-world networks. J. Neurosci. 24, 8075–8083. doi: 10.1523/JNEUROSCI.1509-04.2004
58
OhS. W.HarrisJ. A.NgL.WinslowB.CainN.MihalasS.et al. (2014). A mesoscale connectome of the mouse brain. Nature508, 207–214. doi: 10.1038/nature13186
59
PoldrackR. A.GorgolewskiK. J. (2017). Openfmri: Open sharing of task fMRI data. Neuroimage144, 259–261. doi: 10.1016/j.neuroimage.2015.05.073
60
PopovychO. V.JungK.ManosT.Diaz-PierS.HoffstaedterF.SchreiberJ.et al. (2021). Inter-subject and inter-parcellation variability of resting-state whole-brain dynamical modeling. Neuroimage236:118201. doi: 10.1016/j.neuroimage.2021.118201
61
PopovychO. V.ManosT.HoffstaedterF.EickhoffS. B. (2019). What can computational models contribute to neuroimaging data analytics?Front. Syst. Neurosci. 12:68. doi: 10.3389/fnsys.2018.00068
62
RabuffoG.BandyopadhyayA.CalabreseC.GudibandaK.DepannemaeckerD.TakarabeL. M.et al. (2025). Biophysically inspired mean-field model of neuronal populations driven by ion exchange mechanisms. eLife. 14:RP104249. doi: 10.7554/eLife.104249.1
63
ReyesA. D. (2003). Synchrony-dependent propagation of firing rate in iteratively constructed networks in vitro. Nat. Neurosci. 6, 593–599. doi: 10.1038/nn1056
64
RitterP.SchirnerM.McIntoshA. R.JirsaV. K. (2013). The virtual brain integrates computational modeling and multimodal neuroimaging. Brain Connect. 3, 121–145. doi: 10.1089/brain.2012.0120
65
SacramentoJ.WichertA.van RossumM. C. (2015). Energy efficient sparse connectivity from imbalanced synaptic plasticity rules. PLoS Comput. Biol. 11:e1004265. doi: 10.1371/journal.pcbi.1004265
66
Sanz LeonP.KnockS. A.WoodmanM. M.DomideL.MersmannJ.McIntoshA. R.et al. (2013). The virtual brain: a simulator of primate brain network dynamics. Front. Neuroinform. 7:10. doi: 10.3389/fninf.2013.00010
67
Sanz-LeonP.KnockS. A.SpieglerA.JirsaV. K. (2015). Mathematical framework for large-scale brain network modeling in the virtual brain. Neuroimage111:385–430. doi: 10.1016/j.neuroimage.2015.01.002
68
SchirnerM.DomideL.PerdikisD.TriebkornP.StefanovskiL.PaiR.et al. (2022). Brain simulation as a cloud service: the virtual brain on ebrains. Neuroimage251:118973. doi: 10.1016/j.neuroimage.2022.118973
69
ScottA. C. (1975). The electrophysics of a nerve fiber. Rev. Mod. Phys. 47:487. doi: 10.1103/RevModPhys.47.487
70
ShaheenH.PalS.MelnikR. (2022). Multiscale co-simulation of deep brain stimulation with brain networks in neurodegenerative disorders. Brain Multiphysics3:100058. doi: 10.1016/j.brain.2022.100058
71
StefanovskiL.MeierJ. M.PaiR. K.TriebkornP.LettT.MartinL.et al. (2021). Bridging scales in alzheimer's disease: biological framework for brain simulation with the virtual brain. Front. Neuroinform. 15:630172. doi: 10.3389/fninf.2021.630172
72
StefanovskiL.TriebkornP.SpieglerA.Diaz-CortesM.-A.SolodkinA.JirsaV.et al. (2019). Linking molecular pathways and large-scale computational modeling to assess candidate disease mechanisms and pharmacodynamics in alzheimer's disease. Front. Comp. Neurosci. 13:2019. doi: 10.3389/fncom.2019.00054
73
StimbergM.BretteR.GoodmanD. F. (2019). Brian 2, an intuitive and efficient neural simulator. eLife8:e47314. doi: 10.7554/eLife.47314
74
SumaA.SiggD.GallagherS.GonnellaG.CarnevaleV. (2024). Ion channels in critical membranes: clustering, cooperativity, and memory effects. PRX Life2:013007. doi: 10.1103/PRXLife.2.013007
75
TetzlaffS. K.ReyhanE.LayerN.BengtsonC. P.HeuerA.SchroersJ.et al. (2025). Characterizing and targeting glioblastoma neuron-tumor networks with retrograde tracing. Cell188, 390–411. doi: 10.1016/j.cell.2024.11.002
76
ThompsonW.KelvinL. (1855). On the theory of the electric telegraph. Proc. Royal Soc. 7, 382–399. doi: 10.1098/rspl.1854.0093
77
TigerholmJ.PeterssonM. E.ObrejaO.EberhardtE.NamerB.WeidnerC.et al. (2015). C-fiber recovery cycle supernormality depends on ion concentration and ion channel permeability. Biophys. J. 108, 1057–1071. doi: 10.1016/j.bpj.2014.12.034
78
TigerholmJ.PeterssonM. E.ObrejaO.LampertA.CarrR.SchmelzM.et al. (2014). Modeling activity-dependent changes of axonal spike conduction in primary afferent c-nociceptors. J. Neurophysiol. 111, 1721–1735. doi: 10.1152/jn.00777.2012
79
Van RossumM. C.BiG. Q.TurrigianoG. G. (2000). Stable hebbian learning from spike timing-dependent plasticity. J. Neurosci. 20, 8812–8821. doi: 10.1523/JNEUROSCI.20-23-08812.2000
80
Von HelmholtzH. (1850). Messungen über den zeitlichen Verlauf der Zuckung animalischer Muskeln und die Fortpflanzungsgeschwindigkeit der Reizung in den Nerven [Measurements of the time course of contraction in animal muscles and the propagation speed of excitation in the nerves]. Archiv für Anatomie, Physiologie und wissenschaftliche Medicin17, 176–364. German.
81
WangH. E.DollomajaB.TriebkornP.DumaG. M.WilliamsonA.MakhalovaJ.et al. (2025). Virtual brain twins for stimulation in epilepsy. Nat. Comput. Sci. 5, 754–768. doi: 10.1038/s43588-025-00841-6
82
WangX.-J. (2002). Probabilistic decision making by slow reverberation in cortical circuits. Neuron36, 955–968. doi: 10.1016/S0896-6273(02)01092-9
83
WestA. E.GriffithE. C.GreenbergM. E. (2002). Regulation of transcription factors by neuronal activity. Nat. Rev. Neurosci. 3, 921–931. doi: 10.1038/nrn987
84
YasumatsuN.MatsuzakiM.MiyazakiT.NoguchiJ.KasaiH. (2008). Principles of long-term dynamics of dendritic spines. J. Neurosci. 28, 13592–13608. doi: 10.1523/JNEUROSCI.0603-08.2008
85
ZhengP.DimitrakakisC.TrieschJ. (2013). Network self-organization explains the statistics and dynamics of synaptic connection strengths in cortex. PLoS Comput. Biol. 9:e1002848. doi: 10.1371/journal.pcbi.1002848
Summary
Keywords
Arbor, mouse brain connectome, multi-scale neural models, seizures, The Virtual Brain
Citation
Hater T, Courson J, Lu H, Diaz-Pier S and Manos T (2026) Arbor-TVB: a novel multi-scale co-simulation framework with a case study on neural-level seizure generation and whole-brain propagation. Front. Comput. Neurosci. 19:1731161. doi: 10.3389/fncom.2025.1731161
Received
23 October 2025
Revised
15 December 2025
Accepted
17 December 2025
Published
02 February 2026
Volume
19 - 2025
Edited by
Tiago Ribeiro, National Institutes of Health (NIH), United States
Reviewed by
Gerald Kaushallye Cooray, Karolinska Institutet (KI), Sweden
Gian Marco Duma, Eugenio Medea (IRCCS), Italy
Updates
Copyright
© 2026 Hater, Courson, Lu, Diaz-Pier and Manos.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Thorsten Hater, t.hater@fz-juelich.de; Thanos Manos, thanos.manos@cyu.fr
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.