Document Actions

You are here: Home / Bernstein Conference / 2017 / Satellite Workshops / Multiscale modeling and simulation (Dura-Bernal, Lytton)

Multiscale modeling and simulation (Dura-Bernal, Lytton)


Understanding brain function requires characterizing the interactions occurring across many temporal and spatial scales. Mechanistic multiscale modeling aims to organize and explore these interactions across these scales. Multiscale models can provide insights into how changes at the molecular and cellular levels affect dynamics at local networks and brain areas. At the highest levels, they allow us to connect neural activity to theories of behavior, memory and cognition. The recent introduction of large neuroscience projects in US and EU ­­ Brain Research through Advancing Innovative Neurotechnologies (BRAIN) and Human Brain Project (HBP) respectively ­­ provide an opportunity to rapidly gather new and more accurate data to incorporate into the multiscale models. This workshop will present some of the latest multiscale modeling tools and approaches in neuroscience, as well as representative examples of data­driven multiscale neural simulations. The workshop aims to encourage both computational and experimental neuroscientists to make use of multiscale modeling in their research.

The workshop will place special focus on NEURON, one of two neural simulators, along with MOOSE, that attempt to model the multiscale from molecules to large networks, providing a complement to more specialized tools such as STEPS, MCell, Brian, NEST, etc. NEURON is currently used by multiple modeling groups including HBP or the Allen Brain Institute. Recently, new tools have been added to facilitate set­up, organization and simulation of models across scales. At a lower scale, NRxD (NEURON Reaction­Diffusion) provides a novel Python front­end to add intracellular and extracellular reaction­diffusion to electrophysiology. At a higher scale, NetPyNE facilitates development of large network models that incorporate detailed anatomical and physiological data. The workshop will also feature specific example data­driven multiscale models of several brain regions, including olfactory bulb, hippocampus and primary motor cortex. Additional topics will provide discussion of information processing, synaptic plasticity and disease in the context of interactions across the multiple brain scales; as well as simulation reproducibility, model databasing (ModelDB and NeuroML), and use of high performance computing (HPCs).

Recent growth of computational capacity and resources (supercomputing) now enable large multiscale models. Meanwhile, new experimental tools provide the data needed to build these detailed models. Mechanistic multiscale modeling is now acknowledged by funding agencies as the best technique to map these data and to organize conceptual frameworks for the brain. Application of these models extends to understanding of neural coding and to clinical treatments for brain diseases that damage brain and mind.