Title image
Title image2

The 2nd Biosupercomputing Symposium

Top page -> 2ndBSCS -> abstract


Simulations of single neurons and large-scale neural networks

Prof. Shin Ishii
Brain and Neural Systems Team, Integrated Simulation of Living Matter Group,
Computational Science Research Program, RIKEN

The goal of the Brain and Neural Systems Team is to understand how the brain realizes adaptive behaviors by modeling and simulating large-scale neural networks which integrate data available at multiple levels from genes and molecules to neurons and behaviors. To proceed to that direction, we are now constructing a biophysical model of individual neurons which can be used to examine how the structural plasticity of single neurons contributes to emerging the basis for the information processing in the brain, and a model of large-scale neural networks, whose dynamical characters are also important for understanding dynamical information processing in the level of the whole brain. The single neuron model consists of a membrane model, a cytoskeleton model, and a biochemical reaction-diffusion model. This model includes the dynamical membrane-membrane, membrane-cytoskeleton and cytoskeleton-cytoskeleton interaction so as to realize dynamic cell motion such as endocytosis and chemotaxis. By performing simulations of the model, we also expect various phenomena related to the structural plasticity of neurons can be deeply examined, through which the extra-cellular information and intra-cellular information are integrated and processed in the local space of a neuron as to be decoded into the morphology of the neuron. In the large-scale neural network model, each neuron is represented as a point neuron and of the integrate-and-fire-type, but the time-delay in the activities due to the inter-neuron connection is finely represented. To simulate such a network model with a biologically realistic number of neurons with a high performance computing environment, a simulation platform called NEST (Neural Simulation Tool) is employed. A single cortical microcircuit of the volume of about 1mm3, in which there are about 105 neurons and 109 synapses, has been simulated by NEST. The model analysis has revealed that the macroscopic character of layer-specific connectivity within the cortical microcircuit has a crucial role in determining the dynamics of microcircuit networks. Moreover, a larger-scale connection network consisting of 100 microcircuits, in which there are about 107 neurons and 1011 synapses, has also successfully been simulated on the world-class supercomputer system, the Blue Gene architecture. Our current efforts focus on the improvement of the performance for modern supercomputers including the RIKEN next generation supercomputer.

©RIKEN Computational Science Reseach Program