EBRAINS Seminar B: Simulation Capabilities

28th January, 2025, 17:00-18:30 JST

Susanne Kunkel
The NEST ecosystem: A key enabler of efficient brain-scale spiking network simulation and sustainable neuroscience research

Thorsten Hater
Multiscale Simulations of Full Brain Models using Arbor and TVB

Flashlight talks of developing integrated EBRAINS-RI workflows

The NEST ecosystem: A key enabler of efficient brain-scale spiking network simulation and sustainable neuroscience research

Susanne Kunkel

NEST [2] is a powerful simulation engine that has evolved with the neuroscience community over a quarter-century. The simulator has been continuously advanced and extended to tackle new scientific questions and to push the boundaries of large-scale brain simulation at the resolution of single neurons and synapses. The code is extremely scalable, where recent technological advancements also target GPU-based systems [3]. The graphical frontend NEST Desktop [4] and the modelling language NESTML [5] make the NEST ecosystem easily accessible to new users. The NEST ecosystem is open source and developed under the umbrella of the NEST Initiative [1]. This community organization promotes the use of standard high-quality tools in neuroscience and related fields such as artificial intelligence and neurorobotics, thus fostering sustainable research. In the EBRAINS research infrastructure, NEST provides key functionality for research into the dynamics, structure, and function of brain-scale spiking networks [6].

  1. https://www.nest-initiative.org
  2. https://nest-simulator.readthedocs.io
  3. https://nest-gpu.readthedocs.io
  4. https://nest-desktop.readthedocs.io
  5. https://nestml.readthedocs.io
  6. https://www.ebrains.eu/tools/nest

Multiscale Simulations of Full Brain Models using Arbor and TVB

J. Courson, T. Manos (ETIS Lab, ENSEA, CNRS, UMR8051, CY Cergy-Paris University)

S. Diaz, T. Hater, H. Lu, M. v. d. Vlag (Forschungszentrum Jülich)

Simulating full-brain dynamics at neuron resolution is a challenge still far out of the reach of computational neuroscience. Neural mass models, as implemented in The Virtual Brain (TVB) [^tvb], capture whole brain dynamics with a coarse or fine grid at the brain area level and have made personalized medicine and clinical interventions feasible [^tvb-epilepsy]. At the other end of the spectrum, computational tools like Arbor [^arb] enable us to model neurons with full morphological details but with only fractions of the required counts of neurons and synapses. In this study, we present an attempt to combine TVB and Arbor in a practical approach — simultaneously simulating a given brain region at the microscopic level and neural mass models to provide a salient approximation of the neural activity of the remaining brain areas. We build a scalable co-simulation workflow by leveraging the interfaces provided by both Arbor and TVB to communicate with other external simulators. The resulting framework is flexible as it employs existing models available in both simulators respectively with only minor changes. Additionally, synaptic connections bridging models between TVB and Arbor are specified. This feature allows for swapping dynamics on a per-region (TVB) or per-cell (Arbor) basis without redefining the workflow. Thus, different parts of the brain can be modeled and simulated in detail without sweeping changes to the overall model. We show here some preliminary applications of this Arbor-TVB framework and workflow to simulate the dynamics of seizure propagation.