Simulate with EBRAINS

How to use EBRAINS research infrastructure and Fenix resources for simulations on a scale from molecules via small and large networks of point or structured spiking neurons to simplified whole brain activity and virtual environments.

.

Abstract

This online workshop aims at giving an overview of how to simulate with EBRAINS, from the infrastructure level to simulations at multiple scales. It covers a large variety of topics, e.g. computing and data services of or related to the Fenix infrastructure, Neuromorphic Computing, and simulations on a scale from molecules via small and large networks of point or structured spiking neurons to simplified whole brain activity and virtual environments Participants can choose from a large number of sessions what is relevant for them. The first day will be a short workshop about the Fenix User and Resource Management Service (FURMS) and the Fenix Authentication & Authorisation Infrastructure (AAI). The second day will start with the official welcome session, followed by introductions to the different topics of this workshop and to put tools and simulators into relation with each other. These introductions also serve the purpose to help the participants to update their choice of sessions for the following days. This day will conclude with a hands-on introduction to the EBRAINS Collaboratory environment, a web-portal through which many services and tool are available, and which will be used by several sessions on the following days. The last two days are organised with multiple sessions taking place in parallel, in which the tools and services will be presented in more detail, with lectures and/or as hands-on tutorials.

Registration

The event is free of charge, but registration is necessary.

• An EBRAINS account (free of charge as well) is needed to register. To get an EBAINS account, please use ebrains.eu/register
• Then please register here
• The dial-in information for the zoom video conference will be sent to the registered participants shortly before the event

Venue: Online

The training event will happen as an online zoom video conference = can be attended without the need to travel.

Agenda structure:

• 7 November: Fenix User and Resource Management Service (FURMS) and Fenix AAI (level: for specialists)
• 8 November: introductions to the tools and a hands-on introduction to the EBRAINS Collaboratory environment (level: suitable for first time EBRAINS users)
• 9&10 November: parallel tracks showing different tools available in the EBRAINS research infrastructure, some of them as hands-on tutorials

For capacity planning we are going to ask the registered attendants, which sessions they plan to attend.

Time zone

Times in the agenda are by default in CET (Europe, Berlin) and UTC.

To show different time zones: Africa (UTC, Western-, Central- and Eastern-African Time zones), America (Pacific-, Mountain-, Central- and Eastern-Time), Australia, China, India, . Japan, .. or only CET ...

This workshop will provide attendees with the latest updates about both the Fenix Authentication and Authorization Infrastructure (Fenix AAI) and Fenix User and Resource Management Service (FURMS). Together, Fenix AAI and FURMS enable a federated identity and access management service including accounting and reporting of resources across the Fenix sites. The workshop will provide an overview of the current status and developments of Fenix AAI and FURMS, and will present demos of how to use both services.

This day starts with brief introductions to the tools in the morning, followed by a tour from beginner to expert of the EBRAINS "Collaboratory". This is the interface for collaboratively working in teams with documents and accessing services (e.g. using Jupyter notebooks).

(Very) short introductions into all the tools and services (simulations tools, emulations, computing, analysis, ...) covered during this training workshop.

Please see the agenda below for the covered tools and some additional information on the respective session (please use the agenda in the view with all abstracts unfolded)

• Macromolecular flexibility databases; Setting up & running Molecular Dynamics (MD) simulations
• SSB toolkit: from molecular structure to subcellular networks
• τ-RAMD: Estimation of residence time and determination of unbinding pathways
• MiMiC: Mutiscale QM/MM simulations of ligand-enzyme complex with MiMiC

• Electrical model building toolset
• NEURON / CoreNEURON: building and using computational models of neurons and networks of neurons
• Arbor: portable, high-performance library for computational neuroscience simulations

• PyNN: Using the PyNN library to build and simulate spiking neural network models on EBRAINS
• NEST Desktop: web-based graphical user interface (GUI) for the NEST Simulator.
• NESTML: language for neuron and synapse models.

• CINECA HPC
  • Scalable Compute on Galileo100 platform
  • HPC Cloud Computing infrastructure at CINECA
  • BURST buffering
• Central Data Transfer Services

Introduction to the EBRAINS Collaboratory

We will present the services included : identifying users, permissions, wiki pages, storage in Drive and Bucket, Office collaborative editing, sharing files, and a very short intro to the Lab for running Jupyter notebooks. This will include a small hands-on for users to try out the basic features.

This will be an introduction to Jupyter Lab (=execution of python programs in the browser, no additional software installation on the client needed), and Python notebooks.

We will look at:

• a few examples of basic Python notebooks,
• a real notebook using EBRAINS tools,
• how the Lab can be used with notebooks in a git repository,
• how to run supercomputing jobs with Unicore from the Lab This will include some time for hands-on for users to try out notebooks.

This session is more specifically targeted at developers of EBRAINS services and tools.

We will present how to:

• use IAM for user authentication (for services),
• use the quota manager (for services),
• deploy tools to the Lab environment,
• using the INTegration environment for testing This will include some time for open discussion with attendees.

• NEST Desktop: a web-based graphical user interface (GUI) for the NEST Simulator
• TVB: demonstrating the TVB-EBRAINS integrated end-to-end personalized brain simulation workflows in the Cloud
• Electrical model building toolset: build a biologically detailed neuron model using the NEURON simulator and the open source python packages: eFEL, BluePyEfe and BluePyOpt.

NEST Desktop

NEST Desktop is a web-based graphical user interface (GUI) for the NEST Simulator. The GUI guides users to understand the script code for NEST Simulator. The aim of NEST Desktop is an application which should be intuitive and easy to learn for newcomers, but also attractive for experienced users to perform sophisticated simulations.

In this session, we will learn typical steps of a virtual experiment with spiking network model:

• Construct a network of neurons and I/O devices
• Modify models and parameters
• Learn the simulation code
• Analyze spiking activity or analog signals, e.g. membrane potentials
• Manage projects and event data.

By the end of the session we will demonstrate the concept of the client-server architecture in which NEST Desktop is the client and NEST Simulator is the backend server.

TVB

We demonstrate the TVB-EBRAINS integrated end-to-end personalized brain simulation workflows in the Cloud (Schirner et al. 2022) that can be reached via EBRAINS The Virtual Brain Cloud: https://ebrains.eu/service/the-virtual-brain/. Attendees can work jointly in a dedicated collaboratory during the hands-on-session.

Electrical model building toolset

In this session, you will build a biologically detailed neuron model using the NEURON simulator and three open source python packages: eFEL, BluePyEfe and BluePyOpt. First, a short lecture will describe the process of extraction of electrical features from experimental patch-clamp recordings as well as the algorithms used for the creation of the numerical models. Then, during a hands-on session, biological data will be made available and will be used as target for the creation of one or several electrical neuron models.

• NESTML: a domain-specific language for neuron and synapse models
• PyNN: Using the PyNN library to build and simulate spiking neural network models on EBRAINS

NESTML

NESTML is a domain-specific language for neuron and synapse models. These dynamical models can be used in simulations of brain networks on several platforms, in particular NEST Simulator. NESTML combines an easy to understand, yet powerful syntax with a flexible processing toolchain, written in Python, and good simulation performance by means of code generation (C++ for NEST Simulator).

In this tutorial, you will get hands-on experience creating new neuron and synapse models, instantiating them in a NEST simulation, and doing analysis on the recorded data. We will cover creating a compartmental neuron model with an active dendrite, adding stochasticity to a model, and several synaptic plasticity models such as spike-timing dependent plasticity (STDP) and third-factor plasticity rules.

Using the PyNN library to build and simulate spiking neural network models on EBRAINS

PyNN is a simulator-independent Python library for building neuronal network models. Models built with the PyNN API can be simulated without any code modifications on any simulator that PyNN supports (currently NEURON, NEST, and Brian), and on the SpiNNaker and BrainScaleS neuromorphic hardware systems. This allows users to:

• Perform simulations on laptops, supercomputers or neuromorphic hardware with the same code
• Build complex models with minimal code
• More easily migrate models between different simulators and from simulators to neuromorphic computing systems
• Cross-check your simulation results on different simulators In this hands-on session, we will introduce the concepts behind PyNN and together build and simulate spiking network models on the EBRAINS Jupyter Lab service.

Participants will learn how to access different simulation tools in the EBRAINS Lab; how to build spiking network models with PyNN, how to run simulation experiments with NEST, NEURON, Brian and SpiNNaker, and how to visualize the results.

• Health Data Cloud: an introduction to EBRAINS Service for Sensitive Data - the Health Data Cloud (HDC).
• Arbor: a portable, high-performance library for computational neuroscience simulations with multi-compartment, morphologically-detailed cells
• MUSIC: a tool for co-simulation in the computational neuroscience domain

Health Data Cloud

This session will provide an introduction to EBRAINS Service for Sensitive Data - the Health Data Cloud (HDC). We will demonstrate the technical and organizational measures to provide data protection by design and by default according to the EU general data protection regulations (GDPR). Features such as the GUI, command line interface, gateway to HPC, workbench tools, etc. will be demonstrated. We also plan a hands on session where attendees can experience HDC functionality on their own.

Arbor

Arbor is a portable, high-performance library for computational neuroscience simulations with multi-compartment, morphologically-detailed cells, ranging from single cell models to very large networks. Optimisations make Arbor an order of magnitude faster than the most widely-used comparable simulation software. Download Arbor as a C++ library and integrate it in your own program, or install it as a Python library (through pip) and import in any Python script.

MUSIC

MUSIC is a tool for co-simulation in the computational neuroscience domain. It supports simulations where different parts of a model is simulated by different instances of a simulator or different simulators and/or other software or hardware. MUSIC supports large-scale simulations on clusters but can be run on a desktop or laptop as well.

In this session, participants will be introduced to the PyNN/MUSIC interface which allows MUSIC co-simulations to be specified by a single PyNN script. Co-simulations involving NEST and NEURON as well as SpiNNaker and a virtual environment will be demonstrated.

• SSB toolkit: from molecular structure to subcellular networks
• SpiNNaker: Hands-on the SpiNNaker Neuromorphic platform, writing basic neural networks using PyNN via Jupyter.

SSB toolkit

from molecular structure to subcellular networks

The attendees will learn how to use structural biological data to simulate mathematical models of signal-transductions pathways of GPCRs to predict individual drug response phenotypes by correlating molecular interactions with cellular functions.

SpiNNaker

Hands-on session on the SpiNNaker neuromorphic compute system

• Central Data Transfer Services: data transfer in Fenix
• Macromolecular flexibility databases; Setting up & running Molecular Dynamics (MD) simulations
• The Neurorobotics Platform (NRP) is a set if software tools to enable its users to connect brain models to embodied agents interacting with a simulated environment.

Central Data Transfer Services

This session is dedicated to introduce the Data Transfer service in FENIX and explain its usage in different scenarios. It will include a short presentation and demos of Swift-to-Swift data transfers using the DT service, with a possibility to provide hands-on experience to users.

Macromolecular flexibility databases; Setting up & running Molecular Dynamics (MD) simulations

The session will present and describe macromolecular flexibility databases integrated in the EBRAINS infrastructure (MoDEL-CNS, CNS ligands, BioExcel-CV19), built from information extracted from Molecular Dynamics (MD) simulations. The hands-on session will be focused on setting up and running a MD simulation.

Neurorobotics

The Neurorobotics Platform (NRP) is a set if software tools that was specifically developed to enable its users to connect brain models to embodied agents interacting with a simulated environment. V4.0 of the NRP emphasizes its function as an integrative ecosystem capable of orchestrating multiple simulation engines and heterogeneous control components in a distributed manner. This session will introduce NRP v4.0 and showcase how it can be used for multi-engine simulations, learning experiments, and soft real-time simulations.

• Scalable Compute on Galileo100 platform: information on how to get a budget and use the system
• τ-RAMD: Estimation of residence time and determination of unbinding pathways through τ-Random Acceleration Molecular Dynamics (τ-RAMD)
• Co-Sim: co-simulation, connecting several different simulation, analysis and visualization engines

Scalable Compute on Galileo100 platform

During this session we will provide an overview of Galileo100 cluster available at CINECA and information for its usage. Following general details on how to request a budget, the access mode, the on-line documentation and the support provided, we will describe the main harwdware caracteristics of this cluster and provide information on the available storage spaces. We will also describe the programming environment, focusing on main compilers and libraries available, and we will provide an overview of the production environment, focusing on main applications available (modules environment), the scheduling of jobs and the budget consumption. At last, benchmarks results performed on Cascadelake architecture with usual HBP codes will be shown.

τ-RAMD: Estimation of residence time and determination of unbinding pathways through τ-Random Acceleration Molecular Dynamics (τ-RAMD)

The determination of binding kinetic parameters, and particularly of dissociation rates, is of high interest for compound optimization in structure-based drug discovery as well as for modelling complex signaling cascades. τ-RAMD is a computationally efficient tool that enables the estimation of residence time of molecular partners. In this session, lecture and hands-on session on the use of RAMD simulations for estimating dissociation rates of protein-ligand and protein-protein complexes by the τRAMD procedure will be given.

Co-Sim

(details pending) The co-simulation session description is pending

• HPC Cloud Computing infrastructure at CINECA; hands on configuring an using a virtual machine in the CINECA cloud infrastructure
• MiMiC: Mutiscale QM/MM simulations of ligand-enzyme complex with MiMiC
• BrainScaleS: hands-on the physical emulation neuromorphic compute system BrainScaleS

HPC Cloud Computing infrastructure at CINECA

After a brief overview of the HPC Cloud Computing infrastructure available at CINECA, it will be shown how to access and use it. All the participants will be provided with the possibility to test the infrastructure by configuring and launching their own virtual machine also.

MiMiC: Mutiscale QM/MM simulations of ligand-enzyme complex with MiMiC

This hands-on session will provide participants with theoretical background on multiscale QM/MM simulations and practical experience using the highly scalable MiMiC code to investigate neurobiologically relevant systems runnning on high-end HPC facilities.

BrainScaleS

Hands-on tutorial with information on the first steps to run a PyNN described network interactively on the BrainScaleS systems up to defining and using structured neurons.

• BURST buffering: BURST buffering: how to benefit from a fast data tier
• Elephant: a data analysis frmework for electrophysiological activity data

BURST buffering

A brief introduction to CINECA's burst buffered storage service. During the hands-on session participants will have the opportunity to see how to benefit from a fast data tier for I/O-intensive applications.

Elephant

In this workshop, we will explore how to represent and work with electrophysiological activity data consisting of spiking activity and local field potentials using the Neo data model, and how to perform a statistical analysis of such data using the Elephant framework. We will work with a publicly available electrophysiological dataset to walk-through a potential data analysis scenario step-by-step from loading the data to obtaining results on the statistical characterization of spiking activity using a series of prepared Jupyter notebooks.