Neuroscience Gateway: 

Enabling Developers and Users to Utilize Open High Performance Computing Resources for Large Scale Simulations

 

Thursday, 20 July, 2017 | 14:00 - 18:00 
CNS 2017, Antwerp, Belgium

 

WORKSHOP ORGANIZERS

Amit Majumdar
San Diego Supercomputer Center, University of California San Diego, La Jolla, CA, USA

Subhashini Sivagnanam
San Diego Supercomputer Center, University of California San Diego, La Jolla, CA, USA

Ted Carnevale
Department of Neuroscience, Yale University, New Haven, CT, USA

Abstract of the workshop:

The US National Science Foundation (NSF) funded Neuroscience Gateway (NSG) catalyzes computational neuroscience research by lowering the administrative and technical barriers that make it difficult for researchers to access open supercomputer resources for large scale simulations and data processing. It provides free and open access via a web portal and programmatically to supercomputers and time on the supercomputers is acquired via the peer reviewed process of the Extreme Science and Engineering Discovery Environment in the US. It has been in operation since early 2013, it has over 450 registered users. For the 2017 calendar year alone NSG was awarded 10,000,000 core hours on various supercomputers in the US. NSG already has large number of tools and software (NEURON, PGENESIS, NEST, BRIAN, PyNN, MOOSE, Freesurfer, R, Octave, Matlab etc.), libraries (BluePyOpt, CARLsim, Tensorflow etc.) and pipelines (The Virtual Brain Pipeline etc.). NSG is open to any user from anywhere in the world. Developers (of tools, libraries, and pipelines) and users utilize NSG extensively. This workshop will bring together both the users and the developers of tools/libraries/pipelines associated with the NSG for discussion of tools, software and research where HPC resources are utilized for neuroscience.

Title of Talks and Names of Speakers:

14:00 – 14:30 Introduction to the Neuroscience Gateway, Amit Majumdar, Subhashini Sivagnanam, San Diego Supercomputer Center, UCSD; Ted Carnevale, Department of Neuroscience, Yale University, USA

14:30 – 15:00 Parallel simulation of NEURON-based large scale network models, Salvador Dura-Bernal, Samuel A Neymotin, William W. Lytton, Dept. of Physiology & Pharmacology, SUNY Downstate, Brooklyn, NY, USA

15:00 – 15:30 Interaction of the Neuroscience Gateway with the Brain Simulation Platform of the European Human Brain Project: practical examples, M Migliore, CA Lupascu, LL Bologna, R Migliore, Instiute of Biophysics, National Research Council, Palermo, Italy

15:30 – 16:00 Coffee Break

16:00 – 16:30 Using models from the Open Source Brain repository on the NSG portal infrastructure, Padraig Gleeson, Department of Neuroscience, Physiology and Pharmacology, University of London, UK

16:30 - 17:00 Full-scale detailed modeling of a hippocampal CA1 network using the Neuroscience Gateways, Marianne Bezaire, Department of Psychological and Brain Sciences, Boston University, Boston, USA

17:00 – 17:30 Modeling the Effects of Aging and Neurodegeneration on Cortical and Striatal Neurons, Christina M. Weaver, Department of Mathematics and Computer Science, Franklin and Marshall College, USA

17:30 - 18:00 Towards exascale computing in neuroscience: NEST, NestMC and TVB, Alexandar Peyser, Jülich Supercomputing Centre, Institute for Advanced Simulation, Forschungszentrum Jülich

 

SCHEDULE OF EVENTS

14:00 - 14:30 

Introduction to the Neuroscience Gateway

Amit Majumdar, Subhashini Sivagnanam
San Diego Supercomputer Center, University of California San Diego, La Jolla, CA, USA

Ted Carnevale
Department of Neuroscience, Yale University, New Haven, CT, USA

We will provide a brief overview of the Neuroscience Gateway, including all the tools, software and pipelines that are provided via NSG on supercomputing resources. We will describe how users are using the NSG, how it has expanded since its inception, and how developers interact with the NSG and the NSG team to implement their tools, software and pipelines. From the NSG perspective we will discuss the opportunities, ideas and interest to interact with other projects in a collaborative way to serve the broader neuroscience community world-wide.

14:30 - 15:00 

Parallel simulation of NEURON-based large scale network models 

Salvador Dura-Bernal, Samuel A Neymotin, William W Lytton
Department Physiology & Pharmacology, SUNY Downstate, Brooklyn, NY 11203, USA

NEURON is a widely used neuronal simulator, with over 1600 published models. As part of our NIH U01 grant we developed NetPyNE, a Python package to facilitate the development of biological neuronal networks in the NEURON simulator, with an emphasis on the incorporation of multiscale anatomical and physiological data. NetPyNE seamlessly converts a set of a simple, standardized high-level specifications in a declarative format, into a NEURON model. NetPyNE makes it easy to run parallel simulations by managing the burdensome task of manually distributing the workload and gathering data across computing nodes. It also provides a powerful set of analysis methods so the user can plot spike raster plots, LFP power spectra, information transfer measures, connectivity matrices, or intrinsic time-varying variables (eg. voltage) of any subset of cells. To facilitate data sharing, the package saves and loads the specifications, network, and simulation results using common file formats (Pickle, Matlab, JSON or HDF5), and can convert to and from NeuroML, a standard data format for exchanging models in computational neuroscience. This tool is available on NSG and is being used to run efficient parallel NEURON simulations on the XSEDE supercomputers Comet and Stampede. We employed NetPyNE to develop a detailed multiscale computational model of mouse primary motor cortex (M1) microcircuits, based on novel data provided by experimentalist collaborators. Our M1 model includes over 10,000 neurons distributed across the cortical layers, simulating in full scale a cylindrical volume of diameter of 300 μm and cortical depth 1350 μm. Layer 5 corticospinal cell (Betz cell) models accurately reproduce the electrophysiology and morphology of real neurons. More than 80 million synaptic connections reproduce cell type- and sublayer-specific connectivity patterns derived from optogenetic studies. This multiscale model enabled us to analyze neural dynamics and information flow in M1 microcircuits, to better understand how input activity from different regions is propagated and transformed across cortical layers. The model will help decipher the neural code underlying the brain circuits responsible for producing movement, help understand motor disorders, and evaluate novel pharmacological or neurostimulation treatments. 

Acknowledgments: Research supported by NIH grant U01EB017695, NIH R01EB022903 and NIH R01MH086638. 

15:00 - 15:30 

Interaction of the Neuroscience Gateway with the Brain Simulation Platform of the European Human Brain Project: practical examples 

M Migliore, CA Lupascu, LL Bologna, R Migliore
Institute of Biophysics, National Research Council, Palermo, Italy 

The Brain Simulation Platform of the Human Brain Project provides tools and test suites for modelling and simulation at different scales. They are publicly released on a regular basis in the form of use cases, allowing the user communities a relatively easy access and use of platform. In this presentation, we will demonstrate a few practical representative examples of how the Neuroscience Gateway plays a fundamental role in providing the Computational Neuroscience community with HPC resources to run a variety of data-driven models and tools available in the Brain Simulation Platform. We will discuss strengths, limitations, and future user-oriented directions to exploit the NSG capabilities. 

15:30 - 16:00 

Coffee Break

 

16:00 - 16:30 

Using models from the Open Source Brain repository on the NSG portal infrastructure 

Padraig Gleeson
Department of Neuroscience, Physiology and Pharmacology, University College London, UK

This presentation will give an overview of the Open Source Brain initiative (OSB), which aims to create a repository of open source, collaboratively developed models in computational neuroscience. The models are developed in simulator independent NeuroML format, which can then be mapped to run on multiple simulation platforms. This also facilitates extraction of the physiological and anatomical properties of the models for presentation in more accessible forms through the 3D model explorer on the OSB website. We have recently added the ability to simulate models from OSB directly on the NSG infrastructure, and visualise the results in your web browser. 

16:30 - 17:00 

Full-scale detailed modeling of a hippocampal CA1 network using the Neuroscience Gateway 

Marianne Bezaire
Department of Psychological and Brain Sciences, Boston University, Boston, USA 

Advances in available biological detail and computational resources enable the development of large, detailed neural network models. However, managing the simulations, including documenting the parameters and software code used for each simulation and using supercomputers to execute the code, can be tedious. Using the MATLAB-based SimTracker in combination with the Neuroscience Gateway, the process can be streamlined and enable greater efficiency and ability to recreate past simulations. 

17:00 - 17:30 

Modeling the Effects of Aging and Neurodegeneration on Cortical and Striatal Neurons 

Christina M. Weaver, Ph.D.
Associate Professor of Mathematics (Sabbatical 2016-17), Department of Mathematics and Computer Science, Franklin & Marshall College, USA 

We describe our use of parameter optimization of compartmental neuron models related to two projects. The first, focused on the cellular substrates of normal aging, investigates intracellular properties that may underlie the increased excitability observed in pyramidal neurons of the prefrontal cortex of aged vs. young rhesus monkeys in vitro. The second project is focused on Huntington’s Disease (HD), a neurodegenerative disorder characterized by movement, cognitive, and psychiatric dysfunction. Specifically we modeled spiny projection neurons (SPNs) of the striatum, both in wildtype mice and in a commonly studied transgenic mouse model of HD (Q175). The Neuroscience Gateway has been an essential resource in the development of our evolutionary algorithm-based optimization protocols, and for fitting models to empirical data from several neurons in both projects. We describe the optimization procedure, the results from each study, and how these results lead to empirically testable hypotheses. 

17:30 - 18:00 

Towards exascale computing in neuroscience: NEST, NestMC and TVB 

Alexandar Peyser
Jülich Supercomputing Centre, Institute for Advanced Simulation, Forschungszentrum Jülich

Software packages for describing neuronal networks at various scales of resolution are being developed, adopted and improved for the current and oncoming generation of large scale computing. This work will allow the simulation of larger systems and longer timescales while varying more parameters and collecting finer resolution data. In particular, I will be presenting the current HPC status of projects from the morphologically detailed scale (NestMC) to point neurons (NEST) up to neural mass models (The Virtual Brain). Significant improvements are currently under development for new hardware architectures as well as greatly improved scaling which will allow neuroscientists to pose new questions regarding networks, models and data sets.