Distorted Ex/In dynamics and pattern classification in a spiking network model of Fmr1-KO cortical layer IV

---

This MATLAB code accompanies the following paper:

Domanski APF, Booker S, Wyllie DJA, Isaac JTR, Kind PC (2019) "Cellular and Synaptic Compensations Limit Circuit Disruption in Fmr1-KO Mouse but Fail to Prevent Deficits in Information Processing"

Nature Communications 2019 Oct 23;10(1):4814.

DOI: 10.1038/s41467-019-12736-y

Author: Aleksander PF Domanski 2015-2019 University of Bristol, UK aleks.domanski@bristol.ac.uk

Copyright: (C) Aleksander PF Domanski 2019 University of Bristol, UK

License:

GNU General Public License version 2 This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA http://www.gnu.org/copyleft/gpl.html

Usage:

This code builds a spiking network model of early postnatal Fmr1-KO cortical layer 4, provides stimulation and analyses single cell and population output.

Some features of this model:

• Leaky Integrate-and-Fire model neurons with Conductance-based synapses and realistic intrinsic properties:

• Sparse, random connectivity between Ex and In neurons:

• Realistic short-term depression at all synapses:

• All parameters tuned by experimental results from the above paper:

---

Core functions:

L4sim_DesignNetwork.m

This function builds intrinsic parameters and a synaptic connectivity matrix for a recurrent spiking neural network with external stimulation and synapse-specific short-term plasticity.

L4sim_RunModel.m

This function runs a conductance-based spiking network simulation using predefined parameters for network connectivity and synapses. Choice between leaky I&F and Izhikevich model neurons can be selected, short-term plasticity can be in/excluded and in-the-loop plotting can be configured based on input arguement switches.

L4sim_MakePulseInput.m

This function specifies the simulation parameters and builds the external pulse input structure for the thalamocortical pulse-response simulation. Stimulation parameters are specified as independently tunable rhythmic Dirac deltas to each of the Ex and In pools.

L4sim_Analyse.m

This function analyses sigle-trial simulation results and extracts statistics on Excitation/Inhibition balance in the network pool:

• Population Ex/In balance: Fraction of firing cells and smoothed spike density functions for each pool.
• Single cell Ex/In balance: Balance in input currents.
• Correlation between In and Ex firing pools
• Power spectral density LFP is simulated by taking multi-taper estimates of low-pass filtered summed input currents.
• Summary stats on spike time and rate

L4sim_PlotResults.m

This function plots results of a single model trial:

• Spike rasters and smoothed spike density plots
• Simulated membrane potential and power spectra/spectrograms
• Synaptic currents
• Measures of Excitation/Inhibition
• Single cell firing summary stats

N.B. frequency-domain analysis requires the Chronux toolbox, redistributed under the GNU General Public License ver.2 http://chronux.org/forum/ and "Observed Brain Dynamics", Partha Mitra and Hemant Bokil, Oxford University Press, New York, 2008.

---

Experiment 1) Test intrinsic tuning and connections

• Recapitulates Figure S6

Example_CurrentSteps.m

Plots a simulated single-cell response to injected current (F-I plot):

Example_STP.m

Plots example short-term plastic behaviour of synapses in response to repetitive stimulation:

Experiment 2) Simulated Frequency response of Layer 4 network to patterned thalamocortical activity

• Recapitulates Figure 9

• Run with L4sim_RunFullModelLoop.m and L4sim_RunMultiTrialLoop.m

• Analyse with L4sim_metaAnalyseGroupData_FiringRate.m and L4sim_metaAnalyseGroupData_ExInBalance.m

---

Experiment 3) Examine thalamocortical pattern classification performance

• Recapitulates Figure 10

• Generate input pattern distractor sequences with L4sim_MakeDistractorInput.m

Single cell classification performance

• Run with L4Sim_ExamplePatternSeparation.m

This script recapitulates the simulation analysis results for Fig 10A-C in the linked paper. This code plots example spike trains from single units in response to inputs with/without added distractors. Quantification of spike train modulation (change in rate, first spike time) is provided, as well as through a spike train metric (van Rossum MC, Neural computation 2001 Apr;13(4):751-63)

This code uses an optimised implementation of the van Rossum distance (Haughton C, Kreuz T, Network. 2012;23(1-2):48-58.). This dependency function is available online: provided (MATLAB) by Thomas Kreuz at: http://wwwold.fi.isc.cnr.it/users/thomas.kreuz/Source-Code/VanRossum.html and (C++) by Conor Haughton at: https://sourceforge.net/p/spikemetrics/code/ci/default/tree/

Population level classification performance

• Run with L4sim_RunMultiPatternLoop.m

• Analyse with metaAnalyseDistractors_ClassificationAccuracyVsNoUnits.m

• Spike trains from distractor stimulation trials are compared againt those from the reference condition using a multivariate population decoder.

• Trials are iteratively withheld and the performance of classifiers trainied on the remaining N-1 trials is evaluated in successfully determining the presence/absence of the distractor in the witheld trial.
• Population information content is then compared between genotypes as a function of the size of randomly drawn pools (ensembles) of neurons used for classification.

---