Search

The need for reproducible, credible, multiscale biological modeling has led to the development of standardized simulation platforms, such as the widely-used NEURON environment for computational neuroscience. Developing and maintaining NEURON over several decades has required attention to the competing needs of backwards compatibility, evolving computer architectures, the addition of new scales and physical processes, accessibility to new users, and efficiency and flexibility for specialists. In order to meet these challenges, we have now substantially modernized NEURON, providing continuous integration, an improved build system and release workflow, and better documentation. With the help of a new source-to-source compiler of the NMODL domain-specific language we have enhanced NEURON's ability to run efficiently, via the CoreNEURON simulation engine, on a variety of hardware platforms, including GPUs. Through the implementation of an optimized in-memory transfer mechanism this performance optimized backend is made easily accessible to users, providing training and model-development paths from laptop to workstation to supercomputer and cloud platform. Similarly, we have been able to accelerate NEURON's reaction-diffusion simulation performance through the use of just-in-time compilation. We show that these efforts have led to a growing developer base, a simpler and more robust software distribution, a wider range of supported computer architectures, a better integration of NEURON with other scientific workflows, and substantially improved performance for the simulation of biophysical and biochemical models., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2022 Awile, Kumbhar, Cornu, Dura-Bernal, King, Lupton, Magkanaris, McDougal, Newton, Pereira, Săvulescu, Carnevale, Lytton, Hines and Schürmann.)

The first compartmental computer models of brain neurons using the Rall method predicted novel and unexpected dendrodendritic interactions between mitral and granule cells in the olfactory bulb. We review the models from a 50-year perspective on the work that has challenged, supported, and extended the original proposal that these interactions mediate both lateral inhibition and oscillatory activity, essential steps in the neural basis of olfactory processing and perception. We highlight strategies behind the neurophysiological experiments and the Rall methods that enhance the ability of detailed compartmental modeling to give counterintuitive predictions that lead to deeper insights into neural organization at the synaptic and circuit level. The application of these methods to mechanisms of neurogenesis and plasticity are exciting challenges for the future.

Magneto- and electro-encephalography (MEG/EEG) non-invasively record human brain activity with millisecond resolution providing reliable markers of healthy and disease states. Relating these macroscopic signals to underlying cellular- and circuit-level generators is a limitation that constrains using MEG/EEG to reveal novel principles of information processing or to translate findings into new therapies for neuropathology. To address this problem, we built Human Neocortical Neurosolver (HNN, https://hnn.brown.edu) software. HNN has a graphical user interface designed to help researchers and clinicians interpret the neural origins of MEG/EEG. HNN's core is a neocortical circuit model that accounts for biophysical origins of electrical currents generating MEG/EEG. Data can be directly compared to simulated signals and parameters easily manipulated to develop/test hypotheses on a signal's origin. Tutorials teach users to simulate commonly measured signals, including event related potentials and brain rhythms. HNN's ability to associate signals across scales makes it a unique tool for translational neuroscience research., Competing Interests: SN, DD, BC, RM, NC, MJ, CM, MH, MH, SJ No competing interests declared, (© 2020, Neymotin et al.)

Computational models are powerful tools for exploring the properties of complex biological systems. In neuroscience, data-driven models of neural circuits that span multiple scales are increasingly being used to understand brain function in health and disease. But their adoption and reuse has been limited by the specialist knowledge required to evaluate and use them. To address this, we have developed Open Source Brain, a platform for sharing, viewing, analyzing, and simulating standardized models from different brain regions and species. Model structure and parameters can be automatically visualized and their dynamical properties explored through browser-based simulations. Infrastructure and tools for collaborative interaction, development, and testing are also provided. We demonstrate how existing components can be reused by constructing new models of inhibition-stabilized cortical networks that match recent experimental results. These features of Open Source Brain improve the accessibility, transparency, and reproducibility of models and facilitate their reuse by the wider community., (Copyright © 2019 The Authors. Published by Elsevier Inc. All rights reserved.)

Precision in neuron names is increasingly needed. We are entering a new era in which classical anatomical criteria are only the beginning toward defining the identity of a neuron as carried in its name. New criteria include patterns of gene expression, membrane properties of channels and receptors, pharmacology of neurotransmitters and neuropeptides, physiological properties of impulse firing, and state-dependent variations in expression of characteristic genes and proteins. These gene and functional properties are increasingly defining neuron types and subtypes. Clarity will therefore be enhanced by conveying as much as possible the genes and properties in the neuron name. Using a tested format of parent-child relations for the region and subregion for naming a neuron, we show how the format can be extended so that these additional properties can become an explicit part of a neuron's identity and name, or archived in a linked properties database. Based on the mouse, examples are provided for neurons in several brain regions as proof of principle, with extension to the complexities of neuron names in the cerebral cortex. The format has dual advantages, of ensuring order in archiving the hundreds of neuron types across all brain regions, as well as facilitating investigation of a given neuron type or given gene or property in the context of all its properties. In particular, we show how the format is extensible to the variety of neuron types and subtypes being revealed by RNA-seq and optogenetics. As current research reveals increasingly complex properties, the proposed approach can facilitate a consensus that goes beyond traditional neuron types.

Development of credible clinically-relevant brain simulations has been slowed due to a focus on electrophysiology in computational neuroscience, neglecting the multiscale whole-tissue modeling approach used for simulation in most other organ systems. We have now begun to extend the NEURON simulation platform in this direction by adding extracellular modeling. The extracellular medium of neural tissue is an active medium of neuromodulators, ions, inflammatory cells, oxygen, NO and other gases, with additional physiological, pharmacological and pathological agents. These extracellular agents influence, and are influenced by, cellular electrophysiology, and cellular chemophysiology-the complex internal cellular milieu of second-messenger signaling and cascades. NEURON's extracellular reaction-diffusion is supported by an intuitive Python-based where/who/what command sequence, derived from that used for intracellular reaction diffusion, to support coarse-grained macroscopic extracellular models. This simulation specification separates the expression of the conceptual model and parameters from the underlying numerical methods. In the volume-averaging approach used, the macroscopic model of tissue is characterized by free volume fraction -the proportion of space in which species are able to diffuse, and tortuosity- the average increase in path length due to obstacles. These tissue characteristics can be defined within particular spatial regions, enabling the modeler to account for regional differences, due either to intrinsic organization, particularly gray vs. white matter, or to pathology such as edema. We illustrate simulation development using spreading depression, a pathological phenomenon thought to play roles in migraine, epilepsy and stroke. Simulation results were verified against analytic results and against the extracellular portion of the simulation run under FiPy. The creation of this NEURON interface provides a pathway for interoperability that can be used to automatically export this class of models into complex intracellular/extracellular simulations and future cross-simulator standardization.

The olfactory bulb (OB) transforms sensory input into spatially and temporally organized patterns of activity in principal mitral (MC) and middle tufted (mTC) cells. Thus far, the mechanisms underlying odor representations in the OB have been mainly investigated in MCs. However, experimental findings suggest that MC and mTC may encode parallel and complementary odor representations. We have analyzed the functional roles of these pathways by using a morphologically and physiologically realistic three-dimensional model to explore the MC and mTC microcircuits in the glomerular layer and deeper plexiform layer. The model makes several predictions. MCs and mTCs are controlled by similar computations in the glomerular layer but are differentially modulated in deeper layers. The intrinsic properties of mTCs promote their synchronization through a common granule cell input. Finally, the MC and mTC pathways can be coordinated through the deep short-axon cells in providing input to the olfactory cortex. The results suggest how these mechanisms can dynamically select the functional network connectivity to create the overall output of the OB and promote the dynamic synchronization of glomerular units for any given odor stimulus.

Stochastic simulation of chemical reactions and diffusion in a neuron helps to provide a realistic view of the molecular dynamics within a neuron. We developed a multi-threaded PDES simulator, Neuron Time Warp-Multi Thread, suitable for the stochastic simulation of reaction and diffusion in a neuron. In this paper we make use of Q-Learning and Simulated Annealing to determine the parameters for a dynamic load balancing algorithm and for dynamic window control. During the simulation, the runtime statistics of each thread are collected and used to determine the execution time of the simulation. Based upon this assessment, workload is migrated from the most overloaded threads to the most under-load ones. As the results for a calcium wave model show, both approaches can improve the execution time for small simulations by up to 31% (Q-Learning) and 19% (SA). The simulated annealing approach is more suitable for larger populations, decreasing execution time by 41%.

Cells exhibit stochastic behavior when the number of molecules is small. Hence a stochastic reaction-diffusion simulator capable of working at scale can provide a more accurate view of molecular dynamics within the cell. This paper describes a parallel discrete event simulator, Neuron Time Warp-Multi Thread (NTW-MT), developed for the simulation of reaction diffusion models of neurons. To the best of our knowledge, this is the first parallel discrete event simulator oriented towards stochastic simulation of chemical reactions in a neuron. The simulator was developed as part of the NEURON project. NTW-MT is optimistic and thread-based, which attempts to capitalize on multi-core architectures used in high performance machines. It makes use of a multi-level queue for the pending event set and a single roll-back message in place of individual anti-messages to disperse contention and decrease the overhead of processing rollbacks. Global Virtual Time is computed asynchronously both within and among processes to get rid of the overhead for synchronizing threads. Memory usage is managed in order to avoid locking and unlocking when allocating and de-allocating memory and to maximize cache locality. We verified our simulator on a calcium buffer model. We examined its performance on a calcium wave model, comparing it to the performance of a process based optimistic simulator and a threaded simulator which uses a single priority queue for each thread. Our multi-threaded simulator is shown to achieve superior performance to these simulators. Finally, we demonstrated the scalability of our simulator on a larger CICR model and a more detailed CICR model.

Neuron modeling may be said to have originated with the Hodgkin and Huxley action potential model in 1952 and Rall's models of integrative activity of dendrites in 1964. Over the ensuing decades, these approaches have led to a massive development of increasingly accurate and complex data-based models of neurons and neuronal circuits. ModelDB was founded in 1996 to support this new field and enhance the scientific credibility and utility of computational neuroscience models by providing a convenient venue for sharing them. It has grown to include over 1100 published models covering more than 130 research topics. It is actively curated and developed to help researchers discover and understand models of interest. ModelDB also provides mechanisms to assist running models both locally and remotely, and has a graphical tool that enables users to explore the anatomical and biophysical properties that are represented in a model. Each of its capabilities is undergoing continued refinement and improvement in response to user experience. Large research groups (Allen Brain Institute, EU Human Brain Project, etc.) are emerging that collect data across multiple scales and integrate that data into many complex models, presenting new challenges of scale. We end by predicting a future for neuroscience increasingly fueled by new technology and high performance computation, and increasingly in need of comprehensive user-friendly databases such as ModelDB to provide the means to integrate the data for deeper insights into brain function in health and disease., Competing Interests: The authors declare no competing financial interests.

Large multiscale neuronal network simulations are of increasing value as more big data are gathered about brain wiring and organization under the auspices of a current major research initiative, such as Brain Research through Advancing Innovative Neurotechnologies. The development of these models requires new simulation technologies. We describe here the current use of the NEURON simulator with message passing interface (MPI) for simulation in the domain of moderately large networks on commonly available high-performance computers (HPCs). We discuss the basic layout of such simulations, including the methods of simulation setup, the run-time spike-passing paradigm, and postsimulation data storage and data management approaches. Using the Neuroscience Gateway, a portal for computational neuroscience that provides access to large HPCs, we benchmark simulations of neuronal networks of different sizes (500-100,000 cells), and using different numbers of nodes (1-256). We compare three types of networks, composed of either Izhikevich integrate-and-fire neurons (I&F), single-compartment Hodgkin-Huxley (HH) cells, or a hybrid network with half of each. Results show simulation run time increased approximately linearly with network size and decreased almost linearly with the number of nodes. Networks with I&F neurons were faster than HH networks, although differences were small since all tested cells were point neurons with a single compartment.

Accurate estimation of action potential (AP)-related metabolic cost is essential for understanding energetic constraints on brain connections and signaling processes. Most previous energy estimates of the AP were obtained using the Na(+)-counting method, which seriously limits accurate assessment of metabolic cost of ionic currents that underlie AP conduction along the axon. Here, we first derive a full cable energy function for cortical axons based on classic Hodgkin-Huxley (HH) neuronal equations and then apply the cable energy function to precisely estimate the energy consumption of AP conduction along axons with different geometric shapes. Our analytical approach predicts an inhomogeneous distribution of metabolic cost along an axon with either uniformly or nonuniformly distributed ion channels. The results show that the Na(+)-counting method severely underestimates energy cost in the cable model by 20-70%. AP propagation along axons that differ in length may require over 15% more energy per unit of axon area than that required by a point model. However, actual energy cost can vary greatly depending on axonal branching complexity, ion channel density distributions, and AP conduction states. We also infer that the metabolic rate (i.e. energy consumption rate) of cortical axonal branches as a function of spatial volume exhibits a 3/4 power law relationship.

The olfactory bulb processes inputs from olfactory receptor neurons (ORNs) through two levels: the glomerular layer at the site of input, and the granule cell level at the site of output to the olfactory cortex. The sequence of action of these two levels has not yet been examined. We analyze this issue using a novel computational framework that is scaled up, in three-dimensions (3D), with realistic representations of the interactions between layers, activated by simulated natural odors, and constrained by experimental and theoretical analyses. We suggest that the postulated functions of glomerular circuits have as their primary role transforming a complex and disorganized input into a contrast-enhanced and normalized representation, but cannot provide for synchronization of the distributed glomerular outputs. By contrast, at the granule cell layer, the dendrodendritic interactions mediate temporal decorrelation, which we show is dependent on the preceding contrast enhancement by the glomerular layer. The results provide the first insights into the successive operations in the olfactory bulb, and demonstrate the significance of the modular organization around glomeruli. This layered organization is especially important for natural odor inputs, because they activate many overlapping glomeruli.

Neuronal persistent activity has been primarily assessed in terms of electrical mechanisms, without attention to the complex array of molecular events that also control cell excitability. We developed a multiscale neocortical model proceeding from the molecular to the network level to assess the contributions of calcium (Ca(2+)) regulation of hyperpolarization-activated cyclic nucleotide-gated (HCN) channels in providing additional and complementary support of continuing activation in the network. The network contained 776 compartmental neurons arranged in the cortical layers, connected using synapses containing AMPA/NMDA/GABAA/GABAB receptors. Metabotropic glutamate receptors (mGluR) produced inositol triphosphate (IP3) which caused the release of Ca(2+) from endoplasmic reticulum (ER) stores, with reuptake by sarco/ER Ca(2+)-ATP-ase pumps (SERCA), and influence on HCN channels. Stimulus-induced depolarization led to Ca(2+) influx via NMDA and voltage-gated Ca(2+) channels (VGCCs). After a delay, mGluR activation led to ER Ca(2+) release via IP3 receptors. These factors increased HCN channel conductance and produced firing lasting for ∼1min. The model displayed inter-scale synergies among synaptic weights, excitation/inhibition balance, firing rates, membrane depolarization, Ca(2+) levels, regulation of HCN channels, and induction of persistent activity. The interaction between inhibition and Ca(2+) at the HCN channel nexus determined a limited range of inhibition strengths for which intracellular Ca(2+) could prepare population-specific persistent activity. Interactions between metabotropic and ionotropic inputs to the neuron demonstrated how multiple pathways could contribute in a complementary manner to persistent activity. Such redundancy and complementarity via multiple pathways is a critical feature of biological systems. Mediation of activation at different time scales, and through different pathways, would be expected to protect against disruption, in this case providing stability for persistent activity., (Copyright © 2015 The Authors. Published by Elsevier Ltd.. All rights reserved.)

ModelDB ( modeldb.yale.edu ), a searchable repository of source code of more than 950 published computational neuroscience models, seeks to promote model reuse and reproducibility. Code sharing is a first step; however, model source code is often large and not easily understood. To aid users, we have developed ModelView, a web application for ModelDB that presents a graphical view of model structure augmented with contextual information for NEURON and NEURON-runnable (e.g. NeuroML, PyNN) models. Web presentation provides a rich, simulator-independent environment for interacting with graphs. The necessary data is generated by combining manual curation, text-mining the source code, querying ModelDB, and simulator introspection. Key features of the user interface along with the data analysis, storage, and visualization algorithms are explained. With this tool, researchers can examine and assess the structure of hundreds of models in ModelDB in a standardized presentation without installing any software, downloading the model, or reading model source code.

How the olfactory bulb organizes and processes odor inputs through fundamental operations of its microcircuits is largely unknown. To gain new insight we focus on odor-activated synaptic clusters related to individual glomeruli, which we call glomerular units. Using a 3D model of mitral and granule cell interactions supported by experimental findings, combined with a matrix-based representation of glomerular operations, we identify the mechanisms for forming one or more glomerular units in response to a given odor, how and to what extent the glomerular units interfere or interact with each other during learning, their computational role within the olfactory bulb microcircuit, and how their actions can be formalized into a theoretical framework in which the olfactory bulb can be considered to contain "odor operators" unique to each individual. The results provide new and specific theoretical and experimentally testable predictions.

Calcium (Ca²⁺) waves provide a complement to neuronal electrical signaling, forming a key part of a neuron's second messenger system. We developed a reaction-diffusion model of an apical dendrite with diffusible inositol triphosphate (IP₃), diffusible Ca²⁺, IP₃ receptors (IP₃Rs), endoplasmic reticulum (ER) Ca²⁺ leak, and ER pump (SERCA) on ER. Ca²⁺ is released from ER stores via IP₃Rs upon binding of IP₃ and Ca²⁺. This results in Ca²⁺-induced-Ca²⁺-release (CICR) and increases Ca²⁺ spread. At least two modes of Ca²⁺ wave spread have been suggested: a continuous mode based on presumed relative homogeneity of ER within the cell and a pseudo-saltatory model where Ca²⁺ regeneration occurs at discrete points with diffusion between them. We compared the effects of three patterns of hypothesized IP₃R distribution: (1) continuous homogeneous ER, (2) hotspots with increased IP₃R density (IP₃R hotspots), and (3) areas of increased ER density (ER stacks). All three modes produced Ca²⁺ waves with velocities similar to those measured in vitro (approximately 50-90 μm /sec). Continuous ER showed high sensitivity to IP₃R density increases, with time to onset reduced and speed increased. Increases in SERCA density resulted in opposite effects. The measures were sensitive to changes in density and spacing of IP₃R hotspots and stacks. Increasing the apparent diffusion coefficient of Ca²⁺ substantially increased wave speed. An extended electrochemical model, including voltage-gated calcium channels and AMPA synapses, demonstrated that membrane priming via AMPA stimulation enhances subsequent Ca²⁺ wave amplitude and duration. Our modeling suggests that pharmacological targeting of IP₃Rs and SERCA could allow modulation of Ca²⁺ wave propagation in diseases where Ca²⁺ dysregulation has been implicated.

The precise mechanism by which synaptic excitation and inhibition interact with each other in odor coding through the unique dendrodendritic synaptic microcircuits present in olfactory bulb is unknown. Here a scaled-up model of the mitral-granule cell network in the rodent olfactory bulb is used to analyze dendrodendritic processing of experimentally determined odor patterns. We found that the interaction between excitation and inhibition is responsible for two fundamental computational mechanisms: (1) a balanced excitation/inhibition in strongly activated mitral cells, leading to a sparse representation of odorant input, and (2) an unbalanced excitation/inhibition (inhibition dominated) in surrounding weakly activated mitral cells, leading to lateral inhibition. These results suggest how both mechanisms can carry information about the input patterns, with optimal level of synaptic excitation and inhibition producing the highest level of sparseness and decorrelation in the network response. The results suggest how the learning process, through the emergent development of these mechanisms, can enhance odor representation of olfactory bulb., (Copyright © 2014 the authors 0270-6474/14/3413701-13$15.00/0.)

Cardiac morbidity and mortality increases with the population age. To investigate the underlying pathological mechanisms, and suggest new ways to reduce clinical risks, computational approaches complementing experimental and clinical investigations are becoming more and more important. Here we explore the possible processes leading to the occasional onset and termination of the (usually) non-fatal arrhythmias widely observed in the heart. Using a computational model of a two-dimensional network of cardiac cells, we tested the hypothesis that an ischemia alters the properties of the gap junctions inside the ischemic area. In particular, in agreement with experimental findings, we assumed that an ischemic episode can alter the gap junctions of the affected cells by reducing their average conductance. We extended these changes to include random fluctuations with time, and modifications in the gap junction rectifying conductive properties of cells along the edges of the ischemic area. The results demonstrate how these alterations can qualitatively give an account of all the main types of non-fatal arrhythmia observed experimentally, and suggest how premature beats can be eliminated in three different ways: a) with a relatively small surgical procedure, b) with a pharmacological reduction of the rectifying conductive properties of the gap-junctions, and c) by pharmacologically decreasing the gap junction conductance. In conclusion, our model strongly supports the hypothesis that non-fatal arrhythmias can develop from post-ischemic alteration of the electrical connectivity in a relatively small area of the cardiac cell network, and suggests experimentally testable predictions on their possible treatments.

The functional consequences of the laminar organization observed in cortical systems cannot be easily studied using standard experimental techniques, abstract theoretical representations, or dimensionally reduced models built from scratch. To solve this problem we have developed a full implementation of an olfactory bulb microcircuit using realistic three-dimensional (3D) inputs, cell morphologies, and network connectivity. The results provide new insights into the relations between the functional properties of individual cells and the networks in which they are embedded. To our knowledge, this is the first model of the mitral-granule cell network to include a realistic representation of the experimentally-recorded complex spatial patterns elicited in the glomerular layer (GL) by natural odor stimulation. Although the olfactory bulb, due to its organization, has unique advantages with respect to other brain systems, the method is completely general, and can be integrated with more general approaches to other systems. The model makes experimentally testable predictions on distributed processing and on the differential backpropagation of somatic action potentials in each lateral dendrite following odor learning, providing a powerful 3D framework for investigating the functions of brain microcircuits.

In order to support research on the role of cell biological principles (genomics, proteomics, signaling cascades and reaction dynamics) on the dynamics of neuronal response in health and disease, NEURON's Reaction-Diffusion (rxd) module in Python provides specification and simulation for these dynamics, coupled with the electrophysiological dynamics of the cell membrane. Arithmetic operations on species and parameters are overloaded, allowing arbitrary reaction formulas to be specified using Python syntax. These expressions are then transparently compiled into bytecode that uses NumPy for fast vectorized calculations. At each time step, rxd combines NEURON's integrators with SciPy's sparse linear algebra library.

We present an algorithm to form watertight 3D surfaces consistent with the point-and-diameter based neuronal morphology descriptions widely used with spatial electrophysiology simulators. Such morphology descriptions are readily available online and may come from light-microscopy tracings or from an artificial cell grown algorithmically. These files contain only limited information about a neuron's full three-dimensional shape, as they consist mostly of a list of points and diameters with connectivity data. This representation is well-suited for electrophysiology simulations, where the space constants are larger than geometric ambiguities. However, the simple interpretations used for pure electrophysiological simulation produce geometries unsuitable for multi-scale models that also involve three-dimensional reaction-diffusion, as such models have smaller space constants. Although one cannot exactly reproduce an original neuron's full shape from point-and-diameter data, our new constructive tessellated neuronal geometry (CTNG) algorithm uses constructive solid geometry to define a plausible reconstruction without gaps or cul-de-sacs. CTNG then uses "constructive cubes" to produce a watertight triangular mesh of the neuron surface, suitable for use in reaction-diffusion simulations. CTNG provides the correspondence between internal voxels and surface triangles, needed to make connections between cytoplasmic and membrane mechanisms. Optimization of the underlying marching cubes algorithm and distance calculations enhanced the performance of constructive cubes for a neuronal geometry, where a large number of small objects sparsely occupy a large volume., (Copyright © 2013 Elsevier B.V. All rights reserved.)

In the olfactory bulb, lateral inhibition mediated by granule cells has been suggested to modulate the timing of mitral cell firing, thereby shaping the representation of input odorants. Current experimental techniques, however, do not enable a clear study of how the mitral-granule cell network sculpts odor inputs to represent odor information spatially and temporally. To address this critical step in the neural basis of odor recognition, we built a biophysical network model of mitral and granule cells, corresponding to 1/100th of the real system in the rat, and used direct experimental imaging data of glomeruli activated by various odors. The model allows the systematic investigation and generation of testable hypotheses of the functional mechanisms underlying odor representation in the olfactory bulb circuit. Specifically, we demonstrate that lateral inhibition emerges within the olfactory bulb network through recurrent dendrodendritic synapses when constrained by a range of balanced excitatory and inhibitory conductances. We find that the spatio-temporal dynamics of lateral inhibition plays a critical role in building the glomerular-related cell clusters observed in experiments, through the modulation of synaptic weights during odor training. Lateral inhibition also mediates the development of sparse and synchronized spiking patterns of mitral cells related to odor inputs within the network, with the frequency of these synchronized spiking patterns also modulated by the sniff cycle.

On their long lateral dendrites, mitral cells of the olfactory bulb form dendrodendritic synapses with large populations of granule cell interneurons. The mitral-granule cell microcircuit operating through these reciprocal synapses has been implicated in inducing synchrony between mitral cells. However, the specific mechanisms of mitral cell synchrony operating through this microcircuit are largely unknown and are complicated by the finding that distal inhibition on the lateral dendrites does not modulate mitral cell spikes. In order to gain insight into how this circuit synchronizes mitral cells within its spatial constraints, we built on a reduced circuit model of biophysically realistic multi-compartment mitral and granule cells to explore systematically the roles of dendrodendritic synapse location and mitral cell separation on synchrony. The simulations showed that mitral cells can synchronize when separated at arbitrary distances through a shared set of granule cells, but synchrony is optimally attained when shared granule cells form two balanced subsets, each subset clustered near to a soma of the mitral cell pairs. Another constraint for synchrony is that the input magnitude must be balanced. When adjusting the input magnitude driving a particular mitral cell relative to another, the mitral-granule cell circuit served to normalize spike rates of the mitral cells while inducing a phase shift or delay in the more weakly driven cell. This shift in phase is absent when the granule cells are removed from the circuit. Our results indicate that the specific distribution of dendrodendritic synaptic clusters is critical for optimal synchronization of mitral cell spikes in response to their odor input.

Odors are encoded in spatio-temporal patterns within the olfactory bulb, but the mechanisms of odor recognition and discrimination are poorly understood. It is reasonable to postulate that the olfactory code is sculpted by lateral and feedforward inhibition mediated by granule cells onto the mitral cells. Recent viral tracing and physiological studies revealed patterns of distributed granule cell synaptic clusters that provided additional clues to the possible mechanisms at the network level. The emerging properties and functional roles of these patterns, however, are unknown. Here, using a realistic model of 5 mitral and 100 granule cells we show how their synaptic network can dynamically self-organize and interact through an activity-dependent dendrodendritic mechanism. The results suggest that the patterns of distributed mitral-granule cell connectivity may represent the most recent history of odor inputs, and may contribute to the basic processes underlying mixture perception and odor qualities. The model predicts how and why the dynamical interactions between the active mitral cells through the granule cell synaptic clusters can account for a variety of puzzling behavioral results on odor mixtures and on the emergence of synthetic or analytic perception.

Biologically detailed single neuron and network models are important for understanding how ion channels, synapses and anatomical connectivity underlie the complex electrical behavior of the brain. While neuronal simulators such as NEURON, GENESIS, MOOSE, NEST, and PSICS facilitate the development of these data-driven neuronal models, the specialized languages they employ are generally not interoperable, limiting model accessibility and preventing reuse of model components and cross-simulator validation. To overcome these problems we have used an Open Source software approach to develop NeuroML, a neuronal model description language based on XML (Extensible Markup Language). This enables these detailed models and their components to be defined in a standalone form, allowing them to be used across multiple simulators and archived in a standardized format. Here we describe the structure of NeuroML and demonstrate its scope by converting into NeuroML models of a number of different voltage- and ligand-gated conductances, models of electrical coupling, synaptic transmission and short-term plasticity, together with morphologically detailed models of individual neurons. We have also used these NeuroML-based components to develop an highly detailed cortical network model. NeuroML-based model descriptions were validated by demonstrating similar model behavior across five independently developed simulators. Although our results confirm that simulations run on different simulators converge, they reveal limits to model interoperability, by showing that for some models convergence only occurs at high levels of spatial and temporal discretisation, when the computational overhead is high. Our development of NeuroML as a common description language for biophysically detailed neuronal and network models enables interoperability across multiple simulation environments, thereby improving model transparency, accessibility and reuse in computational neuroscience.

Neuroscience simulators allow scientists to express models in terms of biological concepts, without having to concern themselves with low-level computational details of their implementation. The expressiveness, power and ease-of-use of the simulator interface is critical in efficiently and accurately translating ideas into a working simulation. We review long-term trends in the development of programmable simulator interfaces, and examine the benefits of moving from proprietary, domain-specific languages to modern dynamic general-purpose languages, in particular Python, which provide neuroscientists with an interactive and expressive simulation development environment and easy access to state-of-the-art general-purpose tools for scientific computing.

The NEURON simulation program now allows Python to be used, alone or in combination with NEURON's traditional Hoc interpreter. Adding Python to NEURON has the immediate benefit of making available a very extensive suite of analysis tools written for engineering and science. It also catalyzes NEURON software development by offering users a modern programming tool that is recognized for its flexibility and power to create and maintain complex programs. At the same time, nothing is lost because all existing models written in Hoc, including graphical user interface tools, continue to work without change and are also available within the Python context. An example of the benefits of Python availability is the use of the xml module in implementing NEURON's Import3D and CellBuild tools to read MorphML and NeuroML model specifications.

When a multi-compartment neuron is divided into subtrees such that no subtree has more than two connection points to other subtrees, the subtrees can be on different processors and the entire system remains amenable to direct Gaussian elimination with only a modest increase in complexity. Accuracy is the same as with standard Gaussian elimination on a single processor. It is often feasible to divide a 3-D reconstructed neuron model onto a dozen or so processors and experience almost linear speedup. We have also used the method for purposes of load balance in network simulations when some cells are so large that their individual computation time is much longer than the average processor computation time or when there are many more processors than cells. The method is available in the standard distribution of the NEURON simulation program.

Neuron tree topology equations can be split into two subtrees and solved on different processors with no change in accuracy, stability, or computational effort; communication costs involve only sending and receiving two double precision values by each subtree at each time step. Splitting cells is useful in attaining load balance in neural network simulations, especially when there is a wide range of cell sizes and the number of cells is about the same as the number of processors. For compute-bound simulations load balance results in almost ideal runtime scaling. Application of the cell splitting method to two published network models exhibits good runtime scaling on twice as many processors as could be effectively used with whole-cell balancing.

In an effort to design a simulation environment that is more similar to that of neurophysiology, we introduce a virtual slice setup in the NEURON simulator. The virtual slice setup runs continuously and permits parameter changes, including changes to synaptic weights and time course and to intrinsic cell properties. The virtual slice setup permits shocks to be applied at chosen locations and activity to be sampled intra- or extracellularly from chosen locations. By default, a summed population display is shown during a run to indicate the level of activity and no states are saved. Simulations can run for hours of model time, therefore it is not practical to save all of the state variables. These, in any case, are primarily of interest at discrete times when experiments are being run: the simulation can be stopped momentarily at such times to save activity patterns. The virtual slice setup maintains an automated notebook showing shocks and parameter changes as well as user comments. We demonstrate how interaction with a continuously running simulation encourages experimental prototyping and can suggest additional dynamical features such as ligand wash-in and wash-out-alternatives to typical instantaneous parameter change. The virtual slice setup currently uses event-driven cells and runs at approximately 2 min/h on a laptop.

The increasing complexity of network models poses a growing computational burden. At the same time, computational neuroscientists are finding it easier to access parallel hardware, such as multiprocessor personal computers, workstation clusters, and massively parallel supercomputers. The practical question is how to move a working network model from a single processor to parallel hardware. Here we show how to make this transition for models implemented with NEURON, in such a way that the final result will run and produce numerically identical results on either serial or parallel hardware. This allows users to develop and debug models on readily available local resources, then run their code without modification on a parallel supercomputer.

Neuroscience increasingly uses computational models to assist in the exploration and interpretation of complex phenomena. As a result, considerable effort is invested in the development of software tools and technologies for numerical simulations and for the creation and publication of models. The diversity of related tools leads to the duplication of effort and hinders model reuse. Development practices and technologies that support interoperability between software systems therefore play an important role in making the modeling process more efficient and in ensuring that published models can be reliably and easily reused. Various forms of interoperability are possible including the development of portable model description standards, the adoption of common simulation languages or the use of standardized middleware. Each of these approaches finds applications within the broad range of current modeling activity. However more effort is required in many areas to enable new scientific questions to be addressed. Here we present the conclusions of the "Neuro-IT Interoperability of Simulators" workshop, held at the 11th computational neuroscience meeting in Edinburgh ( July 19-20 2006; http://www.cnsorg.org ). We assess the current state of interoperability of neural simulation software and explore the future directions that will enable the field to advance.

One of the more important recent additions to the NEURON simulation environment is a tool called ModelView, which simplifies the task of understanding exactly what biological attributes are represented in a computational model. Here, we illustrate how ModelView contributes to the understanding of models and discuss its utility as a neuroinformatics tool for analyzing models in online databases and as a means for facilitating interoperability among simulators in computational neuroscience.

The NEURON simulation environment has been extended to support parallel network simulations. Each processor integrates the equations for its subnet over an interval equal to the minimum (interprocessor) presynaptic spike generation to postsynaptic spike delivery connection delay. The performance of three published network models with very different spike patterns exhibits superlinear speedup on Beowulf clusters and demonstrates that spike communication overhead is often less than the benefit of an increased fraction of the entire problem fitting into high speed cache. On the EPFL IBM Blue Gene, almost linear speedup was obtained up to 100 processors. Increasing one model from 500 to 40,000 realistic cells exhibited almost linear speedup on 2,000 processors, with an integration time of 9.8 seconds and communication time of 1.3 seconds. The potential for speed-ups of several orders of magnitude makes practical the running of large network simulations that could otherwise not be explored.

Realistic neural networks involve the coexistence of stiff, coupled, continuous differential equations arising from the integrations of individual neurons, with the discrete events with delays used for modeling synaptic connections. We present here an integration method, the local variable time-step method (lvardt), that uses separate variable-step integrators for individual neurons in the network. Cells that are undergoing excitation tend to have small time steps, and cells that are at rest with little synaptic input tend to have large time steps. A synaptic input to a cell causes reinitialization of only that cell's integrator without affecting the integration of other cells. We illustrated the use of lvardt on three models: a worst-case synchronizing mutual-inhibition model, a best-case synfire chain model, and a more realistic thalamocortical network model.

One of the first and most important stages of odor processing occurs in the glomerular units of the olfactory bulb and most likely involves mitral cell synchronization. Using a detailed model constrained by a number of experimental findings, we show how the intercellular coupling mediated by intraglomerular gap junctions (GJs) in the tuft dendrites could play a major role in sychronization of mitral cell action potential output in spite of their distal dendritic location. The model suggests that the high input resistance and active properties of the fine tuft dendrites are instrumental in generating local spike synchronization and an efficient forward and backpropagation of action potentials between the tuft and the soma. The model also gives insight into the physiological significance of long primary dendrites in mitral cells, and provides evidence against the use of reduced single compartmental models to investigate network properties of cortical pyramidal neurons.

Wider dissemination and testing of computational models are crucial to the field of computational neuroscience. Databases are being developed to meet this need. ModelDB is a web-accessible database for convenient entry, retrieval, and running of published models on different platforms. This article provides a guide to entering a new model into ModelDB., (Copyright 2004 Kluwer Academic Plublishers)

The mitral cell primary dendrite plays an important role in transmitting distal olfactory nerve input from olfactory glomerulus to the soma-axon initial segment. To understand how dendritic active properties are involved in this transmission, we have combined dual soma and dendritic patch recordings with computational modeling to analyze action-potential initiation and propagation in the primary dendrite. In response to depolarizing current injection or distal olfactory nerve input, fast Na(+) action potentials were recorded along the entire length of the primary dendritic trunk. With weak-to-moderate olfactory nerve input, an action potential was initiated near the soma and then back-propagated into the primary dendrite. As olfactory nerve input increased, the initiation site suddenly shifted to the distal primary dendrite. Multi-compartmental modeling indicated that this abrupt shift of the spike-initiation site reflected an independent thresholding mechanism in the distal dendrite. When strong olfactory nerve excitation was paired with strong inhibition to the mitral cell basal secondary dendrites, a small fast prepotential was recorded at the soma, which indicated that an action potential was initiated in the distal primary dendrite but failed to propagate to the soma. As the inhibition became weaker, a "double-spike" was often observed at the dendritic recording site, corresponding to a single action potential at the soma. Simulation demonstrated that, in the course of forward propagation of the first dendritic spike, the action potential suddenly jumps from the middle of the dendrite to the axonal spike-initiation site, leaving the proximal part of primary dendrite unexcited by this initial dendritic spike. As Na(+) conductances in the proximal dendrite are not activated, they become available to support the back-propagation of the evoked somatic action potential to produce the second dendritic spike. In summary, the balance of spatially distributed excitatory and inhibitory inputs can dynamically switch the mitral cell firing among four different modes: axo-somatic initiation with back-propagation, dendritic initiation either with no forward propagation, forward propagation alone, or forward propagation followed by back-propagation.

NEURON is a simulation environment for models of individual neurons and networks of neurons that are closely linked to experimental data. NEURON provides tools for conveniently constructing, exercising, and managing models, so that special expertise in numerical methods or programming is not required for its productive use. This article describes two tools that address the problem of how to achieve computational efficiency and accuracy.

Objective: Our aim was to simulate the spread of membrane potential changes in microvascular trees and then make the simulation programs accessible to other researchers. We have applied our simulations to demonstrate the implications of electrical coupling between arteriolar smooth muscle and endothelium., Methods: A two-layered, cable-like model of an arteriole was used, and the assumptions involved in the approach explicitly stated. Several common experimental situations that involve the passive spread of membrane potential changes in microvascular trees were simulated. The calculations were performed using NEURON, a well-established computer simulation program that we have modified for use with vascular trees., Results: Simulated results show that membrane potential changes would probably not spread as far in the endothelium as they would in the smooth muscle of arterioles. Where feed arteries are connected to larger distributing arteries, passive spread alone may not explain the physiologically observed spread of diameter changes., Conclusions: Simulated results suggest that the morphology of an arteriole, in which the muscle layer is much thicker than the endothelium, favors electrical conduction along smooth muscle rather than the endothelium. However, it seems that passive electrical spread is insufficient to explain the apparent spread of membrane potential changes in experimental situations. Active responses involving voltage-dependent conductances may be involved, and these can also be included in our simulation.

Neuronal function involves the interaction of electrical and chemical signals that are distributed in time and space. The mechanisms that generate these signals and regulate their interactions are marked by a rich diversity of properties that precludes a "one size fits all" approach to modeling. This article presents a summary of how the model description language NMODL enables the neuronal simulation environment NEURON to accommodate these differences.

In olfactory mitral cells, dual patch recordings show that the site of action potential initiation can shift between soma and distal primary dendrite and that the shift is dependent on the location and strength of electrode current injection. We have analyzed the mechanisms underlying this shift, using a model of the mitral cell that takes advantage of the constraints available from the two recording sites. Starting with homogeneous Hodgkin-Huxley-like Na(+)-K(+) channel distribution in the soma-dendritic region and much higher sodium channel density in the axonal region, the model's channel kinetics and density were adjusted by a fitting algorithm so that the model response was virtually identical to the experimental data. The combination of loading effects and much higher sodium channel density in the axon relative to the soma-dendritic region results in significantly lower "voltage threshold" for action potential initiation in the axon; the axon therefore fires first unless the voltage gradient in the primary dendrite is steep enough for it to reach its higher threshold. The results thus provide a quantitative explanation for the stimulus strength and position dependence of the site of action potential initiation in the mitral cell.