Large-scale modeling - a tool for conquering the complexity of the brain

Modular-integrative modeling: a new framework for building brain models that blend biological realism and functional performance

This Perspective presents the Modular-Integrative Modeling approach, a novel framework in neuroscience for developing brain models that blend biological realism with functional performance to provide a holistic view on brain function in interaction with the body and environment.

BiCoSS: Toward Large-Scale Cognition Brain With Multigranular Neuromorphic Architecture

The further exploration of the neural mechanisms underlying the biological activities of the human brain depends on the development of large-scale spiking neural networks (SNNs) with different categories at different levels, as well as the corresponding computing platforms. Neuromorphic engineering provides approaches to high-performance biologically plausible computational paradigms inspired by neural systems. In this article, we present a biological-inspired cognitive supercomputing system (BiCoSS) that integrates multiple granules (GRs) of SNNs to realize a hybrid compatible neuromorphic platform. A scalable hierarchical heterogeneous multicore architecture is presented, and a synergistic routing scheme for hybrid neural information is proposed. The BiCoSS system can accommodate different levels of GRs and biological plausibility of SNN models in an efficient and scalable manner. Over four million neurons can be realized on BiCoSS with a power efficiency of 2.8k larger than the GPU platform, and the average latency of BiCoSS is 3.62 and 2.49 times higher than conventional architectures of digital neuromorphic systems. For the verification, BiCoSS is used to replicate various biological cognitive activities, including motor learning, action selection, context-dependent learning, and movement disorders. Comprehensively considering the programmability, biological plausibility, learning capability, computational power, and scalability, BiCoSS is shown to outperform the alternative state-of-the-art works for large-scale SNN, while its real-time computational capability enables a wide range of potential applications.

An efficient automated parameter tuning framework for spiking neural networks

As the desire for biologically realistic spiking neural networks (SNNs) increases, tuning the enormous number of open parameters in these models becomes a difficult challenge. SNNs have been used to successfully model complex neural circuits that explore various neural phenomena such as neural plasticity, vision systems, auditory systems, neural oscillations, and many other important topics of neural function. Additionally, SNNs are particularly well-adapted to run on neuromorphic hardware that will support biological brain-scale architectures. Although the inclusion of realistic plasticity equations, neural dynamics, and recurrent topologies has increased the descriptive power of SNNs, it has also made the task of tuning these biologically realistic SNNs difficult. To meet this challenge, we present an automated parameter tuning framework capable of tuning SNNs quickly and efficiently using evolutionary algorithms (EA) and inexpensive, readily accessible graphics processing units (GPUs). A sample SNN with 4104 neurons was tuned to give V1 simple cell-like tuning curve responses and produce self-organizing receptive fields (SORFs) when presented with a random sequence of counterphase sinusoidal grating stimuli. A performance analysis comparing the GPU-accelerated implementation to a single-threaded central processing unit (CPU) implementation was carried out and showed a speedup of 65× of the GPU implementation over the CPU implementation, or 0.35 h per generation for GPU vs. 23.5 h per generation for CPU. Additionally, the parameter value solutions found in the tuned SNN were studied and found to be stable and repeatable. The automated parameter tuning framework presented here will be of use to both the computational neuroscience and neuromorphic engineering communities, making the process of constructing and tuning large-scale SNNs much quicker and easier.

The connection-set algebra--a novel formalism for the representation of connectivity structure in neuronal network models

The connection-set algebra (CSA) is a novel and general formalism for the description of connectivity in neuronal network models, from small-scale to large-scale structure. The algebra provides operators to form more complex sets of connections from simpler ones and also provides parameterization of such sets. CSA is expressive enough to describe a wide range of connection patterns, including multiple types of random and/or geometrically dependent connectivity, and can serve as a concise notation for network structure in scientific writing. CSA implementations allow for scalable and efficient representation of connectivity in parallel neuronal network simulators and could even allow for avoiding explicit representation of connections in computer memory. The expressiveness of CSA makes prototyping of network structure easy. A C+ + version of the algebra has been implemented and used in a large-scale neuronal network simulation (Djurfeldt et al., IBM J Res Dev 52(1/2):31-42, 2008b) and an implementation in Python has been publicly released.

Creating, documenting and sharing network models

As computational neuroscience matures, many simulation environments are available that are useful for neuronal network modeling. However, methods for successfully documenting models for publication and for exchanging models and model components among these projects are still under development. Here we briefly review existing software and applications for network model creation, documentation and exchange. Then we discuss a few of the larger issues facing the field of computational neuroscience regarding network modeling and suggest solutions to some of these problems, concentrating in particular on standardized network model terminology, notation, and descriptions and explicit documentation of model scaling. We hope this will enable and encourage computational neuroscientists to share their models more systematically in the future.

The Vertebrate Brain, Evidence of Its Modular Organization and Operating System: Insights into the Brain's Basic Units of Structure, Function, and Operation and How They Influence Neuronal Signaling and Behavior

The human brain is a complex organ made up of neurons and several other cell types, and whose role is processing information for use in eliciting behaviors. However, the composition of its repeating cellular units for both structure and function are unresolved. Based on recent descriptions of the brain's physiological "operating system", a function of the tri-cellular metabolism of N-acetylaspartate (NAA) and N-acetylaspartylglutamate (NAAG) for supply of energy, and on the nature of "neuronal words and languages" for intercellular communication, insights into the brain's modular structural and functional units have been gained. In this article, it is proposed that the basic structural unit in brain is defined by its physiological operating system, and that it consists of a single neuron, and one or more astrocytes, oligodendrocytes, and vascular system endothelial cells. It is also proposed that the basic functional unit in the brain is defined by how neurons communicate, and consists of two neurons and their interconnecting dendritic-synaptic-dendritic field. Since a functional unit is composed of two neurons, it requires two structural units to form a functional unit. Thus, the brain can be envisioned as being made up of the three-dimensional stacking and intertwining of myriad structural units which results not only in its gross structure, but also in producing a uniform distribution of binary functional units. Since the physiological NAA-NAAG operating system for supply of energy is repeated in every structural unit, it is positioned to control global brain function.

Reconstructing the three-dimensional GABAergic microcircuit of the striatum

A system's wiring constrains its dynamics, yet modelling of neural structures often overlooks the specific networks formed by their neurons. We developed an approach for constructing anatomically realistic networks and reconstructed the GABAergic microcircuit formed by the medium spiny neurons (MSNs) and fast-spiking interneurons (FSIs) of the adult rat striatum. We grew dendrite and axon models for these neurons and extracted probabilities for the presence of these neurites as a function of distance from the soma. From these, we found the probabilities of intersection between the neurites of two neurons given their inter-somatic distance, and used these to construct three-dimensional striatal networks. The MSN dendrite models predicted that half of all dendritic spines are within 100µm of the soma. The constructed networks predict distributions of gap junctions between FSI dendrites, synaptic contacts between MSNs, and synaptic inputs from FSIs to MSNs that are consistent with current estimates. The models predict that to achieve this, FSIs should be at most 1% of the striatal population. They also show that the striatum is sparsely connected: FSI-MSN and MSN-MSN contacts respectively form 7% and 1.7% of all possible connections. The models predict two striking network properties: the dominant GABAergic input to a MSN arises from neurons with somas at the edge of its dendritic field; and FSIs are inter-connected on two different spatial scales: locally by gap junctions and distally by synapses. We show that both properties influence striatal dynamics: the most potent inhibition of a MSN arises from a region of striatum at the edge of its dendritic field; and the combination of local gap junction and distal synaptic networks between FSIs sets a robust input-output regime for the MSN population. Our models thus intimately link striatal micro-anatomy to its dynamics, providing a biologically grounded platform for further study.

The brain has an immensely complex wiring diagram, but few computational models of brain regions attempt accurate renditions of the wiring between neurons. Consequently, these models' dynamics may not accurately reflect those of the region. Key barriers here are the difficulty of reconstructing such networks and the paucity of critical data on neuron morphology. We demonstrate an approach that gets around these problems by using the available data to construct prototype neuron morphologies, and uses these to estimate how the probability of a connection between two neurons changes as we change the distance between them. With these in hand, we constructed artificial three-dimensional networks of the rat striatum and find that the connection distributions agree well with current estimates from anatomical studies. Our networks show features and dynamical implications of striatal wiring that would be difficult to intuit: the dominant input to the striatal projection neuron arises from other neurons just at the edge of its dendrites, and the main inhibitory interneurons are coupled locally by electrical connections and more distally by chemical synapses. Together, these properties set a unique state for the input-output computations of the striatum.

Run-time interoperability between neuronal network simulators based on the MUSIC framework

MUSIC is a standard API allowing large scale neuron simulators to exchange data within a parallel computer during runtime. A pilot implementation of this API has been released as open source. We provide experiences from the implementation of MUSIC interfaces for two neuronal network simulators of different kinds, NEST and MOOSE. A multi-simulation of a cortico-striatal network model involving both simulators is performed, demonstrating how MUSIC can promote inter-operability between models written for different simulators and how these can be re-used to build a larger model system. Benchmarks show that the MUSIC pilot implementation provides efficient data transfer in a cluster computer with good scaling. We conclude that MUSIC fulfills the design goal that it should be simple to adapt existing simulators to use MUSIC. In addition, since the MUSIC API enforces independence of the applications, the multi-simulation could be built from pluggable component modules without adaptation of the components to each other in terms of simulation time-step or topology of connections between the modules.

Capturing dopaminergic modulation and bimodal membrane behaviour of striatal medium spiny neurons in accurate, reduced models

Loss of dopamine from the striatum can cause both profound motor deficits, as in Parkinson's disease, and disrupt learning. Yet the effect of dopamine on striatal neurons remains a complex and controversial topic, and is in need of a comprehensive framework. We extend a reduced model of the striatal medium spiny neuron (MSN) to account for dopaminergic modulation of its intrinsic ion channels and synaptic inputs. We tune our D1 and D2 receptor MSN models using data from a recent large-scale compartmental model. The new models capture the input-output relationships for both current injection and spiking input with remarkable accuracy, despite the order of magnitude decrease in system size. They also capture the paired pulse facilitation shown by MSNs. Our dopamine models predict that synaptic effects dominate intrinsic effects for all levels of D1 and D2 receptor activation. We analytically derive a full set of equilibrium points and their stability for the original and dopamine modulated forms of the MSN model. We find that the stability types are not changed by dopamine activation, and our models predict that the MSN is never bistable. Nonetheless, the MSN models can produce a spontaneously bimodal membrane potential similar to that recently observed in vitro following application of NMDA agonists. We demonstrate that this bimodality is created by modelling the agonist effects as slow, irregular and massive jumps in NMDA conductance and, rather than a form of bistability, is due to the voltage-dependent blockade of NMDA receptors. Our models also predict a more pronounced membrane potential bimodality following D1 receptor activation. This work thus establishes reduced yet accurate dopamine-modulated models of MSNs, suitable for use in large-scale models of the striatum. More importantly, these provide a tractable framework for further study of dopamine's effects on computation by individual neurons.

Simple cellular and network control principles govern complex patterns of motor behavior

The vertebrate central nervous system is organized in modules that independently execute sophisticated tasks. Such modules are flexibly controlled and operate with a considerable degree of autonomy. One example is locomotion generated by spinal central pattern generator networks (CPGs) that shape the detailed motor output. The level of activity is controlled from brainstem locomotor command centers, which in turn, are under the control of the basal ganglia. By using a biophysically detailed, full-scale computational model of the lamprey CPG (10,000 neurons) and its brainstem/forebrain control, we demonstrate general control principles that can adapt the network to different demands. Forward or backward locomotion and steering can be flexibly controlled by local synaptic effects limited to only the very rostral part of the network. Variability in response properties within each neuronal population is an essential feature and assures a constant phase delay along the cord for different locomotor speeds.

Dopamine-modulated dynamic cell assemblies generated by the GABAergic striatal microcircuit

The striatum, the principal input structure of the basal ganglia, is crucial to both motor control and learning. It receives convergent input from all over the neocortex, hippocampal formation, amygdala and thalamus, and is the primary recipient of dopamine in the brain. Within the striatum is a GABAergic microcircuit that acts upon these inputs, formed by the dominant medium-spiny projection neurons (MSNs) and fast-spiking interneurons (FSIs). There has been little progress in understanding the computations it performs, hampered by the non-laminar structure that prevents identification of a repeating canonical microcircuit. We here begin the identification of potential dynamically-defined computational elements within the striatum. We construct a new three-dimensional model of the striatal microcircuit's connectivity, and instantiate this with our dopamine-modulated neuron models of the MSNs and FSIs. A new model of gap junctions between the FSIs is introduced and tuned to experimental data. We introduce a novel multiple spike-train analysis method, and apply this to the outputs of the model to find groups of synchronised neurons at multiple time-scales. We find that, with realistic in vivo background input, small assemblies of synchronised MSNs spontaneously appear, consistent with experimental observations, and that the number of assemblies and the time-scale of synchronisation is strongly dependent on the simulated concentration of dopamine. We also show that feed-forward inhibition from the FSIs counter-intuitively increases the firing rate of the MSNs. Such small cell assemblies forming spontaneously only in the absence of dopamine may contribute to motor control problems seen in humans and animals following a loss of dopamine cells.

Simplicity and efficiency of integrate-and-fire neuron models

Lovelace and Cios (2008) recently proposed a very simple spiking neuron (VSSN) model for simulations of large neuronal networks as an efficient replacement for the integrate-and-fire neuron model. We argue that the VSSN model falls behind key advances in neuronal network modeling over the past 20 years, in particular, techniques that permit simulators to compute the state of the neuron without repeated summation over the history of input spikes and to integrate the subthreshold dynamics exactly. State-of-the-art solvers for networks of integrate-and-fire model neurons are substantially more efficient than the VSSN simulator and allow routine simulations of networks of some 10(5) neurons and 10(9) connections on moderate computer clusters.