Neural Simulation Software Demonstration
This brochure was distributed at the 23rd Meeting of the Society for Neuroscience in Washington, DC, November 8 - 11, 1993.
It is being reproduced here because it contains information of general interest.
(C) 1993, Erik De Schutter
23rd Meeting Society for Neuroscience, Washington, DC
November 8 - 11, 1993, booth 1154
Supported by the National Science Foundation
Welcome to the demonstration of simulation software at the Neuroscience Meeting! We show 13 useful and well-supported software packages for different applications and hardware platforms. These packages were peer-reviewed and selected out of a total of 18 applications. We have also two exhibits of high-tech, state-of-the-art uses of computer technology. Most of the demonstrations run continuously, but 5 packages share a single workstation and are on display at the times indicated only. The demonstrations are presented by the authors of the software packages. However, as the authors cannot attend the exhibit continuously, the times during which they will be present are posted. During their absence, the computers may be used to test out the software packages (see posted instructions).
This brochure gives a concise description of all the software that is being demonstrated. We have grouped the packages into 4 groups, however these divisions are somewhat arbitrary as for example 3 of the 4 compartmental modeling packages can also simulate networks. A few months after the meeting a book will be published with complete, detailed descriptions of the software packages. Instructions on how to get on the mailing list for this book are provided at the end of this brochure.
If this demonstration is successful, similar events will be planned for future Society for Neuroscience Meetings.
D.H. Edwards (Georgia State University) was the original promoter of this demonstration during his stay at NSF. The practical organization was by E. De Schutter, J.D. Uhley and J.M. Bower (California Institute of Technology). Selection committee members were J.M. Bower, E. De Schutter, B. Mel (California Institute of Technology) and W. Lytton (University of Wisconsin).
This event is supported by the National Science Foundation. The California Institute of Technology, Sun Computers, Silicon Graphics, Digital Equipment Corporation and Apple Computers provided hardware support.
Compartmental modeling packages
Authors: Matt Wilson and GENESIS Developers
Demonstration by J.M. Bower
Operating system: unix
Prof. J.M. Bower, Division of Biology MS 216-76, Caltech, Pasadena, Ca. 91125
E-mail: genesiscns.caltech.edu, FAX: 818-7952088
Description of GENESIS 1.4:
GENESIS (GEneral NEural SImulation System) was developed to support the simulation of neural systems ranging from complex models of single neurons to simulations of large networks made up of more abstract neuronal components. Neurons are constructed from basic components, such as compartments and variable conductance ionic channels. Compartments are linked to their channels and are then linked together to form multi-compartmental neurons of any desired level of complexity. Neurons may be linked together to form neural circuits. This "object-oriented" approach is central to the generality and flexibility of the system as it allows the user to easily add new features without modification to the base code. Commands may be issued either interactively to a command prompt, by use of simulation scripts, or through the graphical interface, XODUS. Both explicit and implicit integration routines are available.
GENESIS and its graphical front-end XODUS are written in C and run on graphics workstations under unix and on parallel platforms. The current distribution includes full source code for both GENESIS and XODUS as well as several tutorial simulations.
In order to obtain the GENESIS simulator, telnet to genesis.cns.caltech.edu and login as the user "genesis" and answer all the questions. This will create an `ftp' account for you. Then ftp to the same address and download the software.
GENESIS development is supported by NSF, BIR-9017153.
NEMOSYS (Monday 9:00-5:00)
Authors: John Tromp, Frederic Theunissen, Frank Eeckman and John Miller
Demonstration by F.E. Theunissen
Operating system: unix
Prof. J. Miller, MCB, LSA 195., University of California at Berkeley., Berkeley CA
Description of NeMoSys 3.0:
NeMoSys (Neural Modeling System) is a program designed by physiologists for physiologists. It was created for the simulation of current flow in single neurons consisting of one or more connected compartments. A lot of work was devoted to creating an interface that it is easy to use, while at the same time powerful enough to allow one to mimic any possible experimental paradigm. The user interacts with a graphical display of a nerve cell and monitors transmembrane currents and voltages through color-coded animation. Voltage or current traces can be derived from "recording electrodes" positioned at various locations in the cell can be displayed in an oscilloscope-like window. The "recordings" can then be saved on a file for further analysis.
The electro-physiological characteristics of the cell are also set through the graphical interface. The user can quickly assign global electrophysiological properties and then define regions of the cells which differ from the norm. A graphical curve fitting editor is available to define voltage dependent channels. A parametric mode will run a series of "experiments" allowing for a practical investigation of the effect of variation of one or more parameters on the simulation. NeMoSys3.0 runs on any Unix workstation (Sun, SGI, Risc 6000) running X11. The program is available free of charge from anonymous ftp at pmantis.berkeley.edu. The program has successfully used for educational and research purposes.
Author: Erik De Schutter
Demonstration by E. De Schutter
Operating system: Macintosh
E. De Schutter, Dept Medicine, University of Antwerp, Universiteitsplein 1, B2610 Antwerp,
E-mail: erikdsreks.uia.ac.be, FAX: +32-3-8202541
Description of Nodus 3.2:
Nodus is a software package designed for simulation of the electrical behavior of neurons and small networks. Nodus runs on Apple Macintosh computers (FPU necessary) and makes full use of the user friendly interface. Nodus is slower than some of the other simulation packages, but it is much easier to use. Neurons can have passive or excitable membrane, or both. Several commands support the easy maintenance of large, morphologically detailed, compartmental models. For user specified conductance equations in excitable membrane models, any format using the Hodgkin-Huxley formalism is supported. Concentration pools with buffers, diffusion, pumps are implemented. For network models, graded transmitter release and several types of postsynaptic conductances are supported.
Nodus combines a powerful simulator with sophisticated model database management. Models are defined in separate files: conductance definition files, neuron definition files and network definition files. All files specifying one model are linked together in a hierarchical structure and automatically loaded when the top file is opened.
For simulations two integration methods are available: an accurate Fehlberg method (fifth order Runge-Kutta) and a fast forward Euler method, both with variable time steps. The value of any simulation database parameter can be manipulated by the user during simulations. Experiments like current injections, voltage clamps and pharmacological block of conductances are easy to set up.
An extensive user manual is provided with the software.
There is also a standard poster presentation of NEURON (poster 723.1)
Author: Michael Hines
Demonstration by J.W. Moore
Operating system: unix, MS-DOS
New address (1994): Prof. M. Hines, Dept. of Computer Science, Yale University, PO Box
208285, New Haven, CT 06520-8285
Description of NEURON 4:
NEURON is a nerve simulation program which is designed to solve two kinds of problems: 1) where cable properties of cells play an important role, possibly including extracellular potential close to the membrane, and 2) where cell membrane properties are complex, involving many ion-specific channels and ion accumulation.
NEURON is designed around the notion of continuous cable "sections" which are connected together to form any kind of branched cable and which are endowed with properties which vary continuously with position along the section. The design goal is to keep entirely separate the physical properties of the neuron from the numerical issue of size of spatial compartments.
User defined membrane properties are described by expressing models in terms of kinetic schemes and sets of simultaneous equations. Membrane voltage and states are computed efficiently by compiling these model descriptions and using an implicit integration method optimized for branched structures.
NEURON realizes a tremendous degree of flexibility by using an object oriented interpreter to setup the physical properties of the cables, define the appearance of the graphical interface, control the simulation, and plot the results. The default graphical interface is suitable for initial exploratory simulations, setting parameters, common control of voltage and current stimuli, and graphing variables as a function of time and position.
Continuing development of NEURON is being supported by NIH grant NS11613.
Realistic network modeling packages
NEURONC (Wednesday 9:00-5:00)
There is also a standard poster presentation of simulations using NEURONC (poster 579.1)
Author: Robert G. Smith
Demonstration by R. G. Smith
Operating system: unix
Robert G. Smith, Dept. of Neuroscience, University of PA, Philadelphia, PA 19104-6058
E-mail: robretina.anatomy.upenn.edu, FAX: 215-898-9871
Description of NEURONC 3.4:
NeuronC is a computational language for investigating function of neural circuits with "realistic" biophysical properties, e.g. cell morphology, membrane channels and synaptic connections. Originally designed to perform experiments on vision, NeuronC includes a two-dimensional light stimulus and photoreceptors.
To simulate a large neural circuit, arrays of simulated neurons must be constructed and synaptically interconnected. NeuronC was developed as a "simulation language" to simplify construction and running such simulations. It contains a subset of the C language, but in addition contains statements that:
- construct a simulated neural circuit.
- set up a 2-dimensional stimulus.
- set up a recording and "real-time display" paradigm.
- run the simulation.
- analyze and display results.
Because it is a language, NeuronC is not "interactive". However sets of simulations differing in parameter values are easy to run.
PREPARATION (Tuesday 1:00-5:00)
There is also a standard poster presentation of a simulation using PREPARATION (poster
Authors: Peter Rowat and I-Teh Hsieh
Demonstration by I-T. Hsieh
Operating system: unix
Peter Rowat, Biology Department 0357, University of California San Diego, La Jolla, CA
E-mail: ihsiehucsd.edu, FAX: 619-5340301
Description of PREPARATION 1:
The PREPARATION: a computer workbench for investigating the properties of small network models. This software provides the neurobiologist with a framework for investigating the properties of models of small networks, especially oscillatory ones. The system is being developed with the idea of replicating the dynamic properties of living biological systems, for instance the lobster stomatogastric ganglion, Clione, or lamprey spinal cord preparations.
A simulation, once started, continues to run with its output traces scrolling across the screen. A trace can be any model quantity. While a simulation is running, all parameters of the model can be changed. Thus, when membrane conductances or synaptic reversal potentials or any other model quantities are changed, the effects can be seen immediately. Hence one can predict the functional changes caused by neuromodulatory substances. Injected current can be applied in single pulses or in trains, so it is easy to study the phase-reset and entrainment properties of oscillatory networks and to simulate electrophysiological experiments. The time to set up a new model is little more than the time to type the describing equations.
The next version of the software will have multiple on-line phase-portraits available.
SNNAP (Thursday 9:00-5:00)
There is also a standard poster presentation of SNNAP (poster 86.13)
Authors: Israel Ziv, Douglas A. Baxter and John H. Byrne
Demonstration by D.A. Baxter
Operating system: unix
John H. Byrne, Dept Neurobiology, Univ. Texas Med. School, PO Box 20708, Houston, TX 77225
E-mail: jbyrnenbal19.med.uth.tmc.edu, FAX: 713-7924818
Description of SNNAP 1.0:
SNNAP: Simulator for Neural Networks and Action Potentials. SNNAP was designed as a tool for the rapid development and simulation of realistic models of single neurons and small neural networks.
The electrical properties of individual neurons are described with Hodgkin-Huxley type voltage- and time-dependent ionic currents. The connections among neurons can be made by either electrical or chemical synapses. The chemical synaptic connections are capable of expressing several forms of plasticity, such as homo- and hetero-synaptic depression and facilitation. SNNAP also includes mathematical descriptions of intracellular second messengers and ions. The synthesis of second messengers can be driven by either synaptic inputs or by externally applied transmitters. The accumulation of an ion can be driven by any specified voltage- and time-dependent ionic current(s). The intracellular concentrations of ions and/or second messengers, in turn, can be linked to one or more ionic conductances and/or mechanisms contributing to chemical synaptic transmission. Thus, SNNAP can simulate the modulation, either enhancement or inhibition, of membrane currents and synaptic transmission. Moreover, SNNAP can simulate current flow in multicompartment models of neurons by using the equations describing electrical coupling.
Some advantages of SNNAP are its graphical user interface, its ability to simulate common experimental manipulations, and the modular organization of its input files, which allows for the development of a library of distinctive types of neuronal and synaptic properties, thereby facilitating the creation of new neurons or networks from preexisting elements.
Authors: Orjan Ekeberg, Magnus Stensmo and Per Hammarlund
Demonstration by O. Ekeberg
Operating system: unix
O. Ekeberg, SANS, NADA, Royal Institute of Technology, S-100 44 Stockholm, Sweden
Description of SWIM 1.5:
SWIM is primarily intended for simulating large networks of neurons without loosing the ability to manipulate channel and other membrane characteristics including morphology. Each neuron is modeled using a set of interconnected compartments (typically less than 100) equipped with different ion-channels described by Hodgkin-Huxley type parameters. Calcium-dependent potassium channels and NMDA synapses are included.
Networks of up to about 1000 neurons can be handled on a high-end desktop unix workstation. To permit execution of even larger systems, support has been added for CRAY and Connection Machine supercomputers.
Neuron and synapse types are hierarchically described in a special specification language. This enables experimental parameters to be easily modified between runs. The output is processed by separate programs, either to get ordinary time-voltage graphs or to compute other measures of importance.
Abstract network modeling packages
Authors: J.-F. Vibert, D. Lambolez and K. Pakdama
Demonstration by Jean-François Vibert
Operating system: unix, VMS
Dr J. F. Vibert , B3E, Fac. Medecine Saint-Antoine, 27 rue Chaligny, 75571 PARIS Cedex 12,
E-mail: vibertb3e.jussieu.fr, FAX: +33-1-43073957
Description of NBC 6.3:
Neuro_bio_clusters (NBC) is a software package created to simulate interacting biological neural networks. It is designed for neuroscientists who know little about computer science. NBC provides two neuron simulation levels: a simple phenomenological one in which membrane potential and threshold are included, and a more sophisticated one, based on conductance variations in different ionic channels. NBC enables the simulation of biologically plausible networks formed by several interconnected neural clusters connected through pathways of variable length, which can receive external inputs and whose connection matrix is specified. It also allows the input and other properties of individual units and synapses to be modified during the simulation, thereby taking into account the changes in environmental conditions due to learning or pathological conditions. NBC can accept inputs from neurons not simulated during the current simulation (either from a previous simulation or living neurons).
The behavior of the simulated network can be visualized using dynamic color representations. NBC provides analysis tools highly useful in the study of network behavior, at both global and unitary levels, and in both the frequency and the temporal domain. NBC is menu driven with a user-friendly XWindow/Motif interface and produces PostScript graphic outputs.
Authors: Peter Wilke et. al.
Demonstration by P. Wilke
Operating system: unix, VMS, MS-DOS
Prof. P. Wilke, Univ. Erlangen-Nuernberg, Martensstr. 3, D-8520 Erlangen, Germany
E-mail: wilkeinformatik.uni-erlangen.de, FAX: +49-9131-39388
Description of NEUROGRAPH 2.2:
NeuroGraph is a simulator for artificial neural networks, which consist of layers, pools and neurons, which themselves consist of several functions (e.g. activation, transfer, etc.). NeuroGraph has an easy to use graphical interface and several visualization tools, e.g. Hinton diagrams, charts and event driven protocols.
All common network models (e.g. backpropagation, interactive competition, Boltzman, Hopfield, etc.) are implemented and can be customized. In addition custom models can be created and integrated.
NeuroGraph runs on PCs and workstations under unix. To speed up time consuming tasks like learning NeuroGraph can take advantage of workstation clusters and/or parallel computers.
NeuroGraph was developed at the Chair for Programming Languages at the University of Erlangen-Nuernberg, Germany. It is written in C, using the X and Motif libraries.
Author: Alfredo Weitzenfeld
Demonstration by J. Liaw and P. Dominey
Operating system: unix, VMS
A. Weitzenfeld, Center Neural Engineering, Univ. Southern California, Los Angeles, CA
E-mail: alfredocs.usc.edu, FAX: 213-7405687
Description of NSL 2.1:
NSL is a general purpose neural network simulation language built on top of C++. NSL class libraries include basic objects such as vectors and matrices, to the more sophisticated numerical and neural networks learning libraries (some of which are in development). The design philosophy of NSL is that the user creates a model using NSL libraries in C++, which is then linked with the rest of the NSL system. At run time the user has access to an interpreter for setting up parameters and graphics specifications, which include temporal, spatial, 2D and 3D displays.
The main difference between NSL and other systems, is that we put emphasis on flexibility and extendibility in the models which may be created, while at the same time we give an evolved set of object-oriented components including graphics tools for analyzing the behavior of those models.
The latest version runs on Suns through X windows (using the XView toolkit), and we are in the process of porting the interface to Motif and other workstation environments. NSL requires a C++ compiler.
You can get the system via "anonymous ftp" from: yorick.usc.edu (126.96.36.199).
Analysis and database tools
Authors: John Guckenheimer, Mark Myers, Rick Wicklin and Patrick Worfolk
Demonstration by J. Guckenheimer
Operating system: unix
Prof. J. Guckenheimer, Center for Applied Mathematics 504ETC, Cornell Univ., Ithaca, NY
E-mail: guckencam.cornell.edu, FAX: 607-2559860
Description of DSTOOL 1.1:
DsTool (short for Dynamical System TOOLkit) is designed to aid researchers in the investigation of dynamical systems. Toward this goal, our ultimate objective is to construct a comprehensive, open "package" that provides an interactive interface for all computations involving dynamical systems. This objective is ambitious and is certainly not yet fully realized.
Scientific libraries of numerical algorithms (LINPACK, EISPACK,...) are extraordinarily valuable resources for building complex programs out of modular components. But the value of these individual units is greatly enhanced when they are integrated into a larger software package which uses shared data structures to connect the algorithms with a sophisticated user interface that provides interactive control. Such a package can increase our ability to use computers as tools for gaining insight by minimizing time wasted in repetitious programming and using clumsy schemes for manipulating data. DsTool is a program that can serve as the core of an evolving environment that effectively integrates a comprehensive computational library with a user interface, visualization tools, and utilities for examining and printing data.
GAD (Tuesday 9:00-12:30)
There is also a standard poster presentation of GAD (poster 138.8)
Authors: Bruno Olshausen and William Press
Demonstration by B. Olshausen
Operating system: unix
W. Press, Dept. of Anatomy, Box 8108, Washington Univ. School of Med., 660 S. Euclid Ave.,
St. Louis, MO 63110
E-mail: pressv1.wustl.edu, brunolgn.wustl.edu, FAX: 314-3623446
Description of GAD:
For over four decades, neuroanatomists have used lesions and anatomical tracers to reveal the connections between various parts of the brain. While many such studies exist, there is currently no method for quantitatively unifying these data to present a summary of our current state of knowledge. The Graphical Anatomical Database represents an initial attempt to bring such a quantitative approach to bear on anatomical tracer and lesion studies.
Each study is entered into the database by drawing the injection (or lesion) and corresponding label (or degeneration) sites onto digitized images of the appropriate brain region. After many such studies have been entered, one can then query the database to display a summary of what is known about the connections between any two regions. These results are displayed as a "heat map", where color represents the probability of any two areas being connected. At present, we are using the database to study the connections between pulvinar and cortex.
Demonstrations of state of the art computing:
Collaborators: D. Deerfield, N.H. Goddard (Pittsburgh Supercomputing Center), G.T. Kenyon (University of Texas)
The Metacenter poster and demonstration is an opportunity for two-way communication between neuroscientists and the NSF Supercomputing Centers. We will describe and demonstrate the use of packages such as GENESIS and NEURON, as well as project-specific programs, on the vector and parallel platforms that the Centers make available to the scientific community. The major use of our machines is for work on problems which require intensive data processing far beyond the capabilities of a single workstation. We are actively seeking input from neuroscientists as to the software that would be most useful for large-scale computational problems in neuroscience. This could include simulation, databases, visualization and data analysis software, but is not limited to those problems. Please stop by, see what the Centers can currently provide for you, and let us know what you would like to see the Centers do to be of greater utility to the neuroscience community.
Collaborators: J. Leigh, T. A. DeFanti (University of Illinois at Chicago), C. Assad, U.S.
Bhalla, E. De Schutter, A. Protopappas, B. Rasnow, J.M. Bower (Caltech)
Demonstration by J. Leigh (spiffbert.eecs.uic.edu).
The application of virtual reality (VR) in scientific visualization is still a relatively new field and its benefits over conventional computer graphics are still being investigated. This is a demonstration of the experimental use of virtual reality in visualizing computational neuroscience data. The suite of demos includes visualizations for electric organ discharge from the weak electric fish. Additional demonstrations include the visualization of a simulation model of the piriform cortex, and the cerebellar Purkinje cell. Finally, data gathered from recordings made in near and distant electrodes in the olfactory bulb of a rat are also visualized.