If you are using the Surf-Hippo software, please send mail to firstname.lastname@example.org to get on the mailing list or for other questions. Periodically updated versions of the User Manual and Reference Manual may be found here. The home page is here.
Motivations, and Why Lisp?
Retrieving, Installing and Running Surf-Hippo
Acknowledgements and References
The Surf-Hippo neuron simulator is used to investigate morphologically and biophysically detailed compartmental models of single neurons and networks of neurons. Surf-Hippo allows ready construction of cells and networks using built-in functions and various anatomical file formats (Neurolucida, NTS and others). Surf-Hippo is a public domain package, written in Lisp, and runs under Unix and Linux.
Cell models may have complicated 3-dimensional dendritic trees with distributed non-linearities and synaptic inputs driven by arbitrary inputs or other cells. The number of circuit nodes (cell compartments) is in practice limited by the memory in the machine; simulations with thousands of nodes are routine. The retention of XYZ coordinates for each circuit node may be exploited, for example, by simulations which use spatially-coded input, or with network simulations where synaptic connectivity is defined by spatial proximity. Channels (Hodgkin-Huxley, an extended Hodgkin-Huxley model (Borg-Graham, 1991, 1999, Markovian (Borg-Graham, 1999), and others), synapses (autonomous, voltage-dependent, light-dependent, and others), current and voltage sources may be added at arbitrary locations within a neuron. Stationary or moving two dimensional input (e.g. light, for retinal (Borg-Graham and Grzywacz, 1992; Borg-Graham, 2001) or visual cortex (Gazeres etal, 1998) simulations) is also provided. Multi-compartment concentration systems are provided for modelling mechanisms such as calcium-dependent processes.
Surf-Hippo has an extensive graphical user interface, including menus, 3D graphics of dendritic trees (with point-and-click measurement, examination and modification of circuit elements), and data plotting (reformatable, with mouse-oriented measurement). Publication quality postscript files of all graphical output are easily generated. Data files may also be saved for analysis with external tools. Surf-Hippo allows automatic saving of edited model parameters into files which are both loadable and human readable, a feature which helps to avoid errors in saving results during a simulation session.
For integrating the circuit equations, Surf-Hippo uses a variant of the Crank-Nicholson method described by Hines (1984; Borg-Graham, 2000). A major difference in the method used by Surf-Hippo is a variable time step option, where step size is adjusted according to an estimate of the linear truncation error for all state variables (e.g. node voltages, channel particles). The adaptive time step can give much faster run times for typical simulations, with the option of verifying selected results using the more conservative fixed time step integration. Another difference is that Surf-Hippo allows for either ideal or non-ideal voltage clamp at arbitrary nodes in the circuit.
In general, the application of Surf-Hippo overlaps packages such as GENESIS and NEURON, as well as several other simulators designed for the computational neuroscience community. The fact that Surf-Hippo is the only complete system written in Lisp has at least the following advantages.
To begin with, the necessity for the numerical analysis of the nonlinear dynamical systems that characterize compartmental neuron models means that formal verification of even moderately complicated neuron models is not possible. A practical approach to verification is by the cross-validation of models using independent tools, e.g. simulators written in different languages and with major or minor differences in algorithms. Thus, in an ideal world (someday!) neuron models will be evaluated with more than one software package.
A traditional strength of Lisp is its ability to handle both symbolic as well as numeric representations. Formulation of complex models of physical systems - of which perhaps the brain is the canonical example - is by necessity a symbolic task. Thus, access to model components in Surf-Hippo is quasi object-oriented. For example, a large family of functions are based on the concept of circuit elements, which apply to different types of physical models (somas, dendritic segments, channels, particles, channel types, synapses, etc.). Coupled with the natural language-like syntax of Lisp, this makes for efficient writing of model construction and analysis, without demanding a highly sophisticated knowledge of Lisp by the user. Simulation scripts written by the user, a necessity for serious parameter searching, are also in Lisp, whose flexible and powerful syntax is arguably more transparent than most other languages.
Evaluation of those models, on the other hand, is a numerical task, and the performance of Lisp can be similar to C or Fortran. Our results using the Rallpack benchmark suite (Bhalla etal, 1992) show that the speed and accuracy of Surf-Hippo is comparable to GENESIS and NEURON.
Lisp has the advantage that the user communicates directly with the Lisp interpreter environment, and thus has complete access to all components of a simulation. In practice, it is often convenient to move in and out of the Surf-Hippo GUI loop and access simulation data directly from the Lisp interpreter. The data analysis and graphical capablities of Surf-Hippo are sufficiently sophisticated so that we use this simulator for working with real physiological data (and for directly comparing that data with simulation data).
Another advantage of Lisp is that all functions defined in the system may be executed from the interpreter either individually or within scripts, which makes for a very flexible working environment. New code (including bug fixes) may be (incrementally) compiled and used as needed, without recompiling the entire executable. These features, plus those such as the integrated documentation of Lisp and flexibility regarding types, makes the development and maintenance of a large system like Surf-Hippo much more efficient.
We have endeavored to make this package flexible enough and the user interface powerful enough so that Surf-Hippo will be useful in the public domain. User input is encouraged, and we will do do our best to assist in the integration of Surf-Hippo (i.e. debug) into outside projects in computational neuroscience. We would greatly appreciate reprints or pointers to work that uses Surf-Hippo. Please use the address found in the surf-hippo/lib/SNAIL-MAIL file.
Surf-Hippo runs under Carnegie Mellon University Common Lisp (CMUCL), using the CMU Garnet GUI package - both packages, as well as Surf-Hippo, are in the public domain. It is not absolutely necessary to know Lisp to use Surf-Hippo, but a minimal knowledge (easily acquired) helps considerably.
The supplied executable is for x86 boxes under Linux; otherwise, Surf-Hippo should work with any platform that can run CMUCL (see the installation documentation and the CMUCL and Garnet README files for more information).
The Surf-Hippo software and documentation takes about 500 megabytes of disk space, and is 100% complete. Using SLIME, an interface from GNU Emacs to an inferior Lisp, is a very convenient environment from which to run Surf-Hippo, and is strongly recommended.
For the impatient and/or efficient, refer to the Quick Start document.
The complete source and executable tar files are found at the Surf-Hippo home download directory. Untar the file appropriate for your machine and operating system under a top level directory called "surf-hippo", for example. For the rest of the installation, setup and running, including the necessary edits of your shell startup and emacs initialization files, see the installation.doc.txt and other files in the surf-hippo/doc/installation directory.
The initial motivation for Surf-Hippo came from the neuron simulation packages BULLFROG and NEURON (not the same as that of M. Hines), developed by Christof Koch and Patrick O'Donnell on Symbolics Lisp Machines at the MIT AI Laboratory. A new simulator, HIPPO (as in hippocampus), was developed on the Lisp Machine, and later Surf-Hippo was written with some features based on the SURF (as in circuit node WAVEforms) circuit simulator, written by Don Webber, then of the VLSI CAD Group at the University of California at Berkeley. The name "Surf-Hippo" has the advantage of not being already used in other common contexts, e.g. biblical or biological, nor is it an attempt at a tortured acronym. The PLOT-HACK name and many of the ideas in the plot utilities that I wrote for Surf-Hippo were inspired by Patrick. Surf-Hippo was developed first within the former Center for Biological Information Processing (Tomaso Poggio and Ellen Hildreth, directors), Department of Brain and Cognitive Sciences, MIT.
Again, we would greatly appreciate reprints or pointers to work that uses Surf-Hippo. Please use the address found in the surf-hippo/SNAIL-MAIL file or email@example.com
To cite this package, please use:
Graham, L., The Surf-Hippo Neuron Simulation System, v4, 2017 (surf-hippo.neurophysics.eu)
With thanks to Leonardo Topa at MIT and to the Garnet group at CMU - the system organization for Surf-Hippo is derived directly from Garnet, and to the CMUCL group at CMU. The work of both groups contributed enormously to the realization of this project. Also a tip 'o the hat to the CMUCL net community and their ongoing help.
Bhalla, U. S., Bilitch, D. H. and Bower, J. M., "Rallpacks: a set of benchmarks for neuronal simulators" Trends In Neurosciences, v. 15, n. 11, 1992. Available from ftp://genesis.bbb.caltech.edu/pub/genesis
Borg-Graham, L., "Modelling the Electrical Behavior of Cortical Neurons -- Simulations of Hippocampal Pyramidal Cells". Chapter in "Computer Simulation in Brain Science", edited by R.M.J. Cotterill, Cambridge University Press, 1987.
Borg-Graham, L., ``Simulations Suggest Information Processing Roles for the Diverse Currents Found in Hippocampal Neurons''. Neural Information Processing Systems (Proceedings of the Denver NIPS Conference), edited by D.Z. Anderson, American Institute of Physics Press, 1987.
Borg-Graham, L., ``Modelling the Somatic Electrical Behavior of Hippocampal Pyramidal Neurons'' Master's Thesis, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, 1987 (Also appears as MIT AI Lab TR 1161).
Borg-Graham, L., ``Modelling the Non-Linear Conductances of Excitable Membranes'' Chapter in ``Cellular Neurobiology: A Practical Approach'', edited by J. Chad and H. Wheal, IRL Press at Oxford University Press, 1991.
Borg-Graham, L., ``On Directional Selectivity in Vertebrate Retina: An Experimental and Computational Study'' PhD Thesis, Harvard-MIT Division of Health Sciences and Technology, Massachusetts Institute of Technology, 1991 (Also appears as MIT AI Lab TR 1350).
Borg-Graham, L., Interpretations of Data and Mechanisms for Hippocampal Pyramidal Cell Models Chapter in "Cerebral Cortex, Volume 13, Cortical Models", edited by P.S. Ulinski, E.G. Jones, and A. Peters, Plenum Press, 1999.
Borg-Graham, L. Facilitating the science in computational neuroscience, Nature Neuroscience 3 supp., page 1191, 2000.
Borg-Graham, L., Additional Efficient Computation of Branched Nerve Equations: Adaptive Time Step and Ideal Voltage Clamp, Journal of Computational Neuroscience (v8:3 pp209-226, 2000)
Borg-Graham, L., The Computation of Directional Selectivity in the Retina Occurs Presynaptic to the Ganglion Cell Nature Neuroscience v4(2):176-183, 2001
Borg-Graham, L. and Grzywacz, N. M., ``A Model of the Direction Selectivity Circuit in Retina: Transformations by Neurons Singly and in Concert''. Chapter in ``Single Neuron Computation'', edited by T. McKenna, J. Davis, and S. F. Zornetzer, Academic Press, 1992.
Gazeres, N., Borg-Graham, L., and Fregnac, Y., A phenomenological model of visually evoked spike trains in cat geniculate non-lagged X-cells, Visual Neuroscience, v15:1157-1174, 1998
Graham, L. Modelling neuronal biophysics In The Handbook for Brain Theory and Neural Networks, 2nd edition pp164-170, ed. M. Arbib, MIT Press, 2002.
Graham, L. ModelDB, Neuroinformatics, and Computational Neuroscience, Neuroinformatics, v1:141-144, 2002.
Grzywacz, N. M., Amthor, F. R., and Borg-Graham, L. Does Synaptic Facilitation Mediate Motion Facilitation in the Retina? Proceedings of the Computation in Neural Systems Meeting (CNS92), San Francisco, 1992.
Hines, M.,"Efficient computation of branched nerve equations" International Journal Bio-Medical Computing, v. 15, pp. 69-76, 1984.
Hu, H., Vervaeke, K., Graham, L. J. and Storm, J. F., “Dual, complementary theta resonance filtering in hippocampal pyramidal neurons”, J. Neuroscience 29(46):14472-14483, 2009
Shao, L.R., Halvorsrud, Borg-Graham, L. and Storm, J. F., "The role of BK-type Ca++-dependent K+ channels in spike broadening during repetitive firing in rat hippocampal pyramidal cells" Journal of Physiology, v521.1, pp135-146, 1999
Storm, J., Borg-Graham, L., and Adams, P. ``A passive component of the afterdepolarization (ADP) in rat hippocampal cells'' Biophysical Society Abstracts, 1987
Vervaeke, K., Hu, H., Graham, L.J. and Storm, J.F., "Contrasting effects of the persistent Na+ current on neuronal excitability and spike timing" Neuron, v37:663-680, 2006
The Surf-Hippo Neuron Simulator System
This code was written as part of the Surf-Hippo Project, originally at the Center for Biological Information Processing, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, and currently at the Laboratory of Neurophotonics, CNRS UMR 8250, UFR Biomédicale de l'Université Paris Descartes, Paris.
Permission to use, copy, modify, and distribute this software and its documentation for any purpose and without fee is hereby granted, provided that this software is cited in derived published work, and the copyright notice appears in all copies and in supporting documentation. The Surf-Hippo Project makes no representations about the suitability of this software for any purpose. It is provided "as is" without express or implied warranty.
Copyright (c) 1989 - 2016, Lyle J. Graham