Building 3D Network Models with neuroConstruct

1
Building 3D Network Models with neuroConstruct

• Padraig Gleeson
• University College London
• p.gleeson_at_ucl.ac.uk
• WAM-BAMM05
• 31 March 2005

2
Overview

• Scope of application
• Main features
• Visualisation, Packing in 3D, validation/editing
  of morphologies
• Simulator interaction
• Import and export of GENESIS/NEURON code
• Network functionality
• Simulation management
• Extensibility
• Future plans

3
Issues

• Aim is to build biologically realistic 3D models,
  including complex connectivity seen in many
  networks
• Difficult to build networks in 3D in current
  simulators
• Existing cell models difficult to transfer
  between simulators

4
Scope of Application

• Compliments existing simulation environments
• Code produced is native GENESIS/NEURON
• Adds functionality
• Graphical interface
• Checks on morphologies
• Network building capabilities
• Storage/replay/analysis of simulations
• Built with Java runs on any platform
• Reuses existing base of models/modellers

5
Visualization

• Single Cells can be viewed in 3D
• Information on morphology/groups/distribution of
  channels, etc.
• Checks on consistency of cell structure
• Segments can be edited
• Info on basic electrophysiology

6
Screenshot Cell Visualization
7
Packing in 3D

• Cell Groups are packed in 3D Regions
• Rectangular Box
• Spherical
• Various Packing Patterns
• Random
• Cubic close packed
• Hexagonal
• Single positioned
• Evenly spaced in 1D

8
Screenshot Packing in 3D
9
Simulator Interaction(1)

• Morphology files can be imported from
• GENESIS (.p readcell format files)
• NEURON (.nrn like files, limited to create,
  pt3dadd, connect, etc.)
• Cvapp (SWC format files)
• MorphML
• Imported cells are checked for validity i.e.
  errors which may cause problems on some platforms
• zero length segments
• all except root segment have parents
• unique names, etc.

10
Simulator Interaction(2)

• Files can currently be exported to
• NEURON, for simulation
• GENESIS, for simulation
• MorphML, for publishing/use by other simulators
• Cell info held in simulator independent format
• Can be mapped to other/future simulators

11
Screenshot Imported Morphology
12
Cell Processes (1)

• Generic way of handling
• Passive membrane conductances
• Voltage dependent channels
• Synaptic mechanisms
• etc.
• 2 options
• Direct use of native file (e.g. GENESIS script,
  NEURON mod file)
• Generic model of Cell Process, mapped to native
  code

13
Cell Processes (2)

• Option 1 Reuse of existing files
• Advantages
• Quick and easy solution
• Existing files tried and tested
• Disadvantages
• Only possible to use Cell Process (and so
  cell/network model) on one simulator
• Still opaque to someone not familiar with
  scripting language
• Doesn't break the process down to fundamental
  model and experimental parameters

14
Cell Processes (3)

• Option 2 Create generic model of Cell Process
• Model separated from experimentally measured
  parameters
• Reuse of tried and tested template files
• Can be mapped on to any simulator
• Automatic handling of units in neuroConstruct

15
Modularity of Cell Processes (1)
Pre-existing well tested
XML template of model of Cell Process, e.g.
Double Exp Synapse, HH Channel
Experimentally determined parameter set gmax,
Tau Rise/Decay, etc.
Published model of Cell Process (XML file)
Mapping of templates to existing simulators
16
(No Transcript)
17
Screenshot Cell Processes
18
Morphology mapping (1)

• neuroConstruct Concepts
• Section unbranched part of axon/dendrite with
  the same biophysical properties (similar to
  section concept in NEURON)
• Segment Specifies one 3D point along Section,
  shaped like conical frustum
• Section object specifies start point, Segment
  objects specify 3D points along the Section

19
Morphology mapping (2)

• Going from GENESIS -gt NEURON
• Simple mapping Compartments in GENESIS mapped to
  Sections (with one Segment) in neuroConstruct
• Adv every point simulated in GENESIS is
  simulated in NEURON
• Disadv May not be need for so many Compartments
• Complex mapping Optimization of compartments in
  unbranched parts into multi Segment Sections
• Need to keep electrotonic length in mind
• Not recording simulation at same point in
  GENESIS/NEURON

20
Morphology mapping (3)

• Going from NEURON -gt GENESIS
• Simple mapping Each Segment to Compartment with
  equivalent surface area
• Many more Compartments than Sections in NEURON
• If axial resistance is high can give incorrect
  results
• Complex mapping Sections with nseg points mapped
  to 3 x nseg Compartments
• Ensures same total membrane area (and so total
  capacitance, conductance) and same axial
  resistance to each simulated point

21
Screenshot Morphology
22
Network features

• Groups of Axons/dendrites on cells can be
  specified as Synaptic Connection Locations
• Factors to specify for how cells are connected
• Source and Target Cell Group
• Number of connections (fixed, random, Gaussian
  distribution)
• Weight and delay can be given fixed
  value/distribution
• Max and min lengths
• Random connection, closest of n cells, the
  closest
• Generation direction (from source or from target)

23
Screenshot Network Connections
24
Simulation Management

• Storage of GENESIS/NEURON sim data in same format
• Browsing of stored simulations
• Replay of data in neuroConstruct
• Visualization/plotting of network activity
• Data Sets (stored simulation data or new plots)
• All of these can be zipped up and distributed in
  one file

25
Screenshot Simulation Browser
26
Network behaviour analysis

• Functions for displaying firing rates of
  individual cells/Cell Groups
• Average interspike intervals
• Cell Group firing histograms
• Example
• Maex DeSchutter model of synchronization in
  Granule Cell Layer

27
Screenshot Network analysis
28
Design philosophy

• Transparency
• Information on morphologies/positions/connections
  easily accessible
• Extensibility
• Additions can be made to the code for new network
  scenarios/ simulation formats
• Modularity
• Ease of unit testing/reuse of network elements
• Ease of access and distribution
• The more researchers who can test models the
  better

29
(No Transcript)
30
Functionality in development

• Models of Granule Cell layer/whole cerebellum in
  3D
• Diffusion in 3D signalling, oxygen supply, etc.
• Command line interface
• Simplified way to alter project settings (e.g.
  for parameter searching)
• Keeps the GUI phobic happy
• Closer integration with NeuroML/MorphML
• Integration with Condor (Uni. Wisconsin)
• Distribution of jobs to a pool of workstations
  (High Throughput Computing)
• Completed simulations appear in Simulation Browser

31
Current status

• Version 0.8
• Beta testers welcome (Note not authorized for
  research until formally released)
• Want to help with testing? Contact via
• p.gleeson_at_ucl.ac.uk

32
Collaborators Funding

• University College London
• Angus Silver, Volker Steuber
• University of Edinburgh
• Fred Howell, Nigel Goddard, David Willshaw
• Work is funded by the Medical Research Council