Fronto-parietal visuospatial WM model with HH cells (Edin et al 2007)

 Download zip file 
Help downloading and running models
Accession:98017
1) J Cogn Neurosci: 3 structural mechanisms that had been hypothesized to underlie vsWM development during childhood were evaluated by simulating the model and comparing results to fMRI. It was concluded that inter-regional synaptic connection strength cause vsWM development. 2) J Integr Neurosci: Given the importance of fronto-parietal connections, we tested whether connection asymmetry affected resistance to distraction. We drew the conclusion that stronger frontal connections are beneficial. By comparing model results to EEG, we concluded that the brain indeed has stronger frontal-to-parietal connections than vice versa.
References:
1 . Edin F, Macoveanu J, Olesen P, Tegnér J, Klingberg T (2007) Stronger synaptic connectivity as a mechanism behind development of working memory-related brain activity during childhood. J Cogn Neurosci 19:750-60 [PubMed]
2 . Edin F, Klingberg T, Stödberg T, Tegnér J (2007) Fronto-parietal connection asymmetry regulates working memory distractibility. J Integr Neurosci 6:567-96 [PubMed]
Model Information (Click on a link to find other models with that property)
Model Type: Realistic Network;
Brain Region(s)/Organism: Neocortex;
Cell Type(s): Neocortex U1 L2/6 pyramidal intratelencephalic GLU cell; Abstract Wang-Buzsaki neuron;
Channel(s):
Gap Junctions: Gap junctions;
Receptor(s):
Gene(s):
Transmitter(s):
Simulation Environment: NEURON;
Model Concept(s): Working memory; Attractor Neural Network;
Implementer(s):
Search NeuronDB for information about:  Neocortex U1 L2/6 pyramidal intratelencephalic GLU cell;
/* MultiModuleWMNet.hoc creates a network consisting of several modules of
* inhibitory and excitatory networks.
* Every module is based on the network in 
* "The dynamical stability of reverberatory neural circuits" by 
* Tegnér, Biol. Cybern. 2002 and can recreate the results of that paper
*
* Network parameters are loaded from a file named SERIES by loadSettings().
* Then, simulation() starts the simulation.
*
* Author: Fredrik Edin, 2003.
* Address: freedin@nada.kth.se
*
*/

load_file( "nrngui.hoc" )

objref results
load_file( "Results.hoc" )

objref simulator
load_file( "Simulator.hoc" )

objref net
load_file( "Net.hoc" )

objref view
load_file( "View.hoc" )


/* Simulation creates a network with the parameters contained in PARAM,
* CUE and EXPv, and performs a simulation.
*
*(Arg 1, pl: "plot" = Plot simulation. Another or no argument = don't plot.) 
*/
objref fi         // file object
objref PARAM      // Vector with parameters loaded from file
objref CUE        // Vector with cue parameters
objref EXPv       // Vector with experimental parameters
objref netborder  // borders between nets
objref trajs      // vector for cells whose trajectories are to be plotted
objref tmp        // temporary object
strdef pl         // plot string
strdef pr         // print string
strdef str        // temporary string
strdef fstr	  // temporary string

proc simulation() {
    
    /* Plot and save? */
    if( numarg() > 0 ) {
	pl = $s1
    } else {
	pl = "no plot"
    }
    if( numarg() > 1 ) {
	pr = $s2
    } else {
	pr = "print"
    }
    if( numarg() > 2 ) {
	saveV = $3 
    } else {
	saveV = 0
    }
    
    /* Times */
    tStart = 0
    tSyn = PARAM.object(0).x[3]
    tStop = PARAM.object(0).x[4]
    DT = PARAM.object(0).x[5] // 0 = variable time step. 
                              // Values above 0 = dt, fixed time step
			      
    /* Data about the nets */
    nmod = PARAM.object(0).x[1]
    netborder = new Vector(2*nmod+1)
    for i = 0, nmod-1 {
	netborder.x[2*i+1] = PARAM.object(0).x[8+4*i] + netborder.x[2*i]
	netborder.x[2*i+2] = PARAM.object(0).x[9+4*i] + netborder.x[2*i+1]
    }
    nCell = netborder.x[ netborder.size() - 1 ]
    
    if( strcmp( pl, "plot" ) == 0 ) {
	/* View only gives good results for 1-module networks */
	view = new View( 4, n2-n1+1, tStop, nCell, tStart )
	view.lInit(tStart)
	view.rInit(tStart)
	view.hInit(tStart) 
	n = 10
	trajs = new Vector(n) /* Plot only some voltage trajectories */
	trajs.indgen( netborder.x[ 1 ], int((netborder.x[ 2 ]-netborder.x[ 1 ]) / n ) )
	print "Cells to be monitored"
	for i = 0, n-1 {
	    print trajs.x[i]
	}
	view.someTrajectories(trajs)
    } else {
	view = new View(1, 1) 
    }
    
    /* Core of the function */
    results = new Results( nCell, netborder, EXPv, tStop, tStart )
    net = new Net(results, netborder, PARAM, CUE, EXPv )
    results.setNet(net)
    if( EXPv.size() > 1 ) {
	simulator = new Simulator( net, results, view, DT, tStop, tSyn, 0.01, 1, saveV, EXPv )
    } else {
	simulator = new Simulator( net, results, view, DT, tStop, tSyn, 0.01, 1, saveV )
    }
    if( strcmp( pl, "plot" ) == 0 ) {
	view.reinit( net, results )
    }
    calcCh = 1 /* For calculating relative NMDA */ 
    simulator.run( pl, pr, calcCh ) //, "no print" ) /* Start the simulation */
    net.saveConn()
    net.saveParams()
    
    if( strcmp( pl, "plot" ) == 0 ) {
	view.plotAPs()
	view.plotHist()
    } 
    
    /* Save results and parameters */
    fi = new File()
    fi.aopen( "Parameters.txt" )
    fi.printf( "Stimulation till puckel:\n" )
    fi.printf( "Nat:\tTid:\tLangd:\tPos:\tBredd:\tAmp:\n" )
    for i = 0, CUE.size()/5-1 {
	fi.printf( "%d\t%d\t%d\t%d\t%g\n", CUE.x[5*i], CUE.x[5*i+1], CUE.x[5*i+2], CUE.x[5*i+3], CUE.x[5*i+4] )
    }
    fi.close()
    
    fi.wopen( "Q.txt" )
    for i = 0, ( CUE.size() )/5-1 {
	fi.printf( "%d\t%d\t%d\t%d\t%g\n", CUE.x[5*i], CUE.x[5*i+1], CUE.x[5*i+2], CUE.x[5*i+3], CUE.x[5*i+4] )
    }
    fi.close()
    
    fi.wopen( "QKey.txt" )
    fi.printf( "Tid:\n" )
    fi.printf( "Langd:\n" )
    fi.printf( "Position:\n" )
    fi.printf( "Bredd:\n" )
    fi.printf( "Amplitud:\n" )
    fi.close()
}


/* Loads settings for the simulations
*
* return: 1 if settings were successfully loaded, 0 otherwise
*/
func loadSettings() {
    fi = new File()
    fi.ropen("SERIES.txt")
    tmp = new StringFunctions()
    pl = "#"
    
    /* Read position of current row */
    readComment()
    fi.gets( str )
    readComment()
    curr = fi.scanvar()
    fi.gets(str)  // Reads last of curr-line
    readComment()
    /* read until current row */
    for i = 0, curr - 1 {
	fi.gets(str)
	readComment()
    }
	
    
    /* make sure there is something left to read */
    start = fi.tell()
    fi.gets( str )
    if( tmp.len( str ) < 2 ) {
	fi.close()
	return 0
    }
    
    /* read parameters */
    fi.seek( start )
    type = fi.scanvar()
    nmod = fi.scanvar()
    nrow = fi.scanvar()
    fi.seek( start )
    PARAM = new List()
    CUE = new Vector()
    EXPv = new Vector()
    for j = 0, nrow-1 {
	PARAM.append( new Vector() )
	start = fi.tell()
	fi.gets( str )
	end = fi.tell()
	fi.seek( start )
	while( start < end-2 ) {
	    PARAM.object( j ).append( fi.scanvar() )
	    start = fi.tell()
	}
	fi.gets( str ) // just to finish the line
	readComment()
    }
    fi.close()
    if( PARAM.object(1).size()%5 == 0 ) {
	CUE = PARAM.object(1).c
    }
    EXPv = PARAM.object(2).c
    PARAM.remove(1)
    PARAM.remove(1)
    /*print "PARAM"
    for i = 0, PARAM.count()-1 {
	PARAM.object(i).printf()
    }
    CUE.printf()
    EXPv.printf()*/
    return 1
}

/* Auxiliary function reading comments until next non-comment is found */
proc readComment() { local i
    pos = fi.tell()
    i = 0
    while( fi.gets(fstr) && tmp.substr( fstr, pl ) == 0 ) {
	pos = fi.tell()
	i = i + 1
	//print i, " ", fstr
    }
    fi.seek( pos )
}

if( loadSettings() ) {
    simulation("no plot", "no print", 0 )
}
quit()

Loading data, please wait...