/* Created by Language version: 7.7.0 */
/* VECTORIZED */
#define NRN_VECTORIZED 1
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include "scoplib_ansi.h"
#undef PI
#define nil 0
#include "md1redef.h"
#include "section.h"
#include "nrniv_mf.h"
#include "md2redef.h"
#if METHOD3
extern int _method3;
#endif
#if !NRNGPU
#undef exp
#define exp hoc_Exp
extern double hoc_Exp(double);
#endif
#define nrn_init _nrn_init__dcnCaHVA
#define _nrn_initial _nrn_initial__dcnCaHVA
#define nrn_cur _nrn_cur__dcnCaHVA
#define _nrn_current _nrn_current__dcnCaHVA
#define nrn_jacob _nrn_jacob__dcnCaHVA
#define nrn_state _nrn_state__dcnCaHVA
#define _net_receive _net_receive__dcnCaHVA
#define _f_rate _f_rate__dcnCaHVA
#define rate rate__dcnCaHVA
#define states states__dcnCaHVA
#define _threadargscomma_ _p, _ppvar, _thread, _nt,
#define _threadargsprotocomma_ double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt,
#define _threadargs_ _p, _ppvar, _thread, _nt
#define _threadargsproto_ double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt
/*SUPPRESS 761*/
/*SUPPRESS 762*/
/*SUPPRESS 763*/
/*SUPPRESS 765*/
extern double *getarg();
/* Thread safe. No static _p or _ppvar. */
#define t _nt->_t
#define dt _nt->_dt
#define perm _p[0]
#define ica _p[1]
#define m _p[2]
#define cai _p[3]
#define cao _p[4]
#define minf _p[5]
#define taum _p[6]
#define T _p[7]
#define A _p[8]
#define Dm _p[9]
#define v _p[10]
#define _g _p[11]
#define _ion_cai *_ppvar[0]._pval
#define _ion_cao *_ppvar[1]._pval
#define _ion_ica *_ppvar[2]._pval
#define _ion_dicadv *_ppvar[3]._pval
#if MAC
#if !defined(v)
#define v _mlhv
#endif
#if !defined(h)
#define h _mlhh
#endif
#endif
#if defined(__cplusplus)
extern "C" {
#endif
static int hoc_nrnpointerindex = -1;
static Datum* _extcall_thread;
static Prop* _extcall_prop;
/* external NEURON variables */
extern double celsius;
/* declaration of user functions */
static void _hoc_getGHKexp(void);
static void _hoc_rate(void);
static int _mechtype;
extern void _nrn_cacheloop_reg(int, int);
extern void hoc_register_prop_size(int, int, int);
extern void hoc_register_limits(int, HocParmLimits*);
extern void hoc_register_units(int, HocParmUnits*);
extern void nrn_promote(Prop*, int, int);
extern Memb_func* memb_func;
#define NMODL_TEXT 1
#if NMODL_TEXT
static const char* nmodl_file_text;
static const char* nmodl_filename;
extern void hoc_reg_nmodl_text(int, const char*);
extern void hoc_reg_nmodl_filename(int, const char*);
#endif
extern void _nrn_setdata_reg(int, void(*)(Prop*));
static void _setdata(Prop* _prop) {
_extcall_prop = _prop;
}
static void _hoc_setdata() {
Prop *_prop, *hoc_getdata_range(int);
_prop = hoc_getdata_range(_mechtype);
_setdata(_prop);
hoc_retpushx(1.);
}
/* connect user functions to hoc names */
static VoidFunc hoc_intfunc[] = {
"setdata_dcnCaHVA", _hoc_setdata,
"getGHKexp_dcnCaHVA", _hoc_getGHKexp,
"rate_dcnCaHVA", _hoc_rate,
0, 0
};
#define _f_getGHKexp _f_getGHKexp_dcnCaHVA
#define getGHKexp getGHKexp_dcnCaHVA
extern double _f_getGHKexp( _threadargsprotocomma_ double );
extern double getGHKexp( _threadargsprotocomma_ double );
static void _check_rate(double*, Datum*, Datum*, _NrnThread*);
static void _check_getGHKexp(double*, Datum*, Datum*, _NrnThread*);
static void _check_table_thread(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt, int _type) {
_check_rate(_p, _ppvar, _thread, _nt);
_check_getGHKexp(_p, _ppvar, _thread, _nt);
}
/* declare global and static user variables */
#define qdeltat qdeltat_dcnCaHVA
double qdeltat = 1;
#define usetable usetable_dcnCaHVA
double usetable = 1;
/* some parameters have upper and lower limits */
static HocParmLimits _hoc_parm_limits[] = {
"usetable_dcnCaHVA", 0, 1,
0,0,0
};
static HocParmUnits _hoc_parm_units[] = {
"perm_dcnCaHVA", "cm/seconds",
"ica_dcnCaHVA", "mA/cm2",
0,0
};
static double delta_t = 0.01;
static double m0 = 0;
/* connect global user variables to hoc */
static DoubScal hoc_scdoub[] = {
"qdeltat_dcnCaHVA", &qdeltat_dcnCaHVA,
"usetable_dcnCaHVA", &usetable_dcnCaHVA,
0,0
};
static DoubVec hoc_vdoub[] = {
0,0,0
};
static double _sav_indep;
static void nrn_alloc(Prop*);
static void nrn_init(_NrnThread*, _Memb_list*, int);
static void nrn_state(_NrnThread*, _Memb_list*, int);
static void nrn_cur(_NrnThread*, _Memb_list*, int);
static void nrn_jacob(_NrnThread*, _Memb_list*, int);
static int _ode_count(int);
static void _ode_map(int, double**, double**, double*, Datum*, double*, int);
static void _ode_spec(_NrnThread*, _Memb_list*, int);
static void _ode_matsol(_NrnThread*, _Memb_list*, int);
#define _cvode_ieq _ppvar[4]._i
static void _ode_matsol_instance1(_threadargsproto_);
/* connect range variables in _p that hoc is supposed to know about */
static const char *_mechanism[] = {
"7.7.0",
"dcnCaHVA",
"perm_dcnCaHVA",
0,
"ica_dcnCaHVA",
0,
"m_dcnCaHVA",
0,
0};
static Symbol* _ca_sym;
extern Prop* need_memb(Symbol*);
static void nrn_alloc(Prop* _prop) {
Prop *prop_ion;
double *_p; Datum *_ppvar;
_p = nrn_prop_data_alloc(_mechtype, 12, _prop);
/*initialize range parameters*/
perm = 7.5e-006;
_prop->param = _p;
_prop->param_size = 12;
_ppvar = nrn_prop_datum_alloc(_mechtype, 5, _prop);
_prop->dparam = _ppvar;
/*connect ionic variables to this model*/
prop_ion = need_memb(_ca_sym);
nrn_promote(prop_ion, 1, 0);
_ppvar[0]._pval = &prop_ion->param[1]; /* cai */
_ppvar[1]._pval = &prop_ion->param[2]; /* cao */
_ppvar[2]._pval = &prop_ion->param[3]; /* ica */
_ppvar[3]._pval = &prop_ion->param[4]; /* _ion_dicadv */
}
static void _initlists();
/* some states have an absolute tolerance */
static Symbol** _atollist;
static HocStateTolerance _hoc_state_tol[] = {
0,0
};
static void _update_ion_pointer(Datum*);
extern Symbol* hoc_lookup(const char*);
extern void _nrn_thread_reg(int, int, void(*)(Datum*));
extern void _nrn_thread_table_reg(int, void(*)(double*, Datum*, Datum*, _NrnThread*, int));
extern void hoc_register_tolerance(int, HocStateTolerance*, Symbol***);
extern void _cvode_abstol( Symbol**, double*, int);
void _dcn_CaHVA_reg() {
int _vectorized = 1;
_initlists();
ion_reg("ca", -10000.);
_ca_sym = hoc_lookup("ca_ion");
register_mech(_mechanism, nrn_alloc,nrn_cur, nrn_jacob, nrn_state, nrn_init, hoc_nrnpointerindex, 1);
_mechtype = nrn_get_mechtype(_mechanism[1]);
_nrn_setdata_reg(_mechtype, _setdata);
_nrn_thread_reg(_mechtype, 2, _update_ion_pointer);
_nrn_thread_table_reg(_mechtype, _check_table_thread);
#if NMODL_TEXT
hoc_reg_nmodl_text(_mechtype, nmodl_file_text);
hoc_reg_nmodl_filename(_mechtype, nmodl_filename);
#endif
hoc_register_prop_size(_mechtype, 12, 5);
hoc_register_dparam_semantics(_mechtype, 0, "ca_ion");
hoc_register_dparam_semantics(_mechtype, 1, "ca_ion");
hoc_register_dparam_semantics(_mechtype, 2, "ca_ion");
hoc_register_dparam_semantics(_mechtype, 3, "ca_ion");
hoc_register_dparam_semantics(_mechtype, 4, "cvodeieq");
hoc_register_cvode(_mechtype, _ode_count, _ode_map, _ode_spec, _ode_matsol);
hoc_register_tolerance(_mechtype, _hoc_state_tol, &_atollist);
hoc_register_var(hoc_scdoub, hoc_vdoub, hoc_intfunc);
ivoc_help("help ?1 dcnCaHVA D:/Projects/SchreglmannEtAl2020/CCTC_model/modfiles/dcn_CaHVA.mod\n");
hoc_register_limits(_mechtype, _hoc_parm_limits);
hoc_register_units(_mechtype, _hoc_parm_units);
}
static double *_t_minf;
static double *_t_taum;
static double *_t_getGHKexp;
static int _reset;
static char *modelname = "High voltage activated calcium current (CaHVA) of deep cerebellar nucleus (DCN) neuron";
static int error;
static int _ninits = 0;
static int _match_recurse=1;
static void _modl_cleanup(){ _match_recurse=1;}
static int _f_rate(_threadargsprotocomma_ double);
static int rate(_threadargsprotocomma_ double);
static int _ode_spec1(_threadargsproto_);
/*static int _ode_matsol1(_threadargsproto_);*/
static double _n_getGHKexp(_threadargsprotocomma_ double _lv);
static void _n_rate(_threadargsprotocomma_ double _lv);
static int _slist1[1], _dlist1[1];
static int states(_threadargsproto_);
/*CVODE*/
static int _ode_spec1 (double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt) {int _reset = 0; {
rate ( _threadargscomma_ v ) ;
Dm = ( minf - m ) / taum ;
}
return _reset;
}
static int _ode_matsol1 (double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt) {
rate ( _threadargscomma_ v ) ;
Dm = Dm / (1. - dt*( ( ( ( - 1.0 ) ) ) / taum )) ;
return 0;
}
/*END CVODE*/
static int states (double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt) { {
rate ( _threadargscomma_ v ) ;
m = m + (1. - exp(dt*(( ( ( - 1.0 ) ) ) / taum)))*(- ( ( ( minf ) ) / taum ) / ( ( ( ( - 1.0 ) ) ) / taum ) - m) ;
}
return 0;
}
static double _mfac_rate, _tmin_rate;
static void _check_rate(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt) {
static int _maktable=1; int _i, _j, _ix = 0;
double _xi, _tmax;
if (!usetable) {return;}
if (_maktable) { double _x, _dx; _maktable=0;
_tmin_rate = - 150.0 ;
_tmax = 100.0 ;
_dx = (_tmax - _tmin_rate)/300.; _mfac_rate = 1./_dx;
for (_i=0, _x=_tmin_rate; _i < 301; _x += _dx, _i++) {
_f_rate(_p, _ppvar, _thread, _nt, _x);
_t_minf[_i] = minf;
_t_taum[_i] = taum;
}
}
}
static int rate(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt, double _lv) {
#if 0
_check_rate(_p, _ppvar, _thread, _nt);
#endif
_n_rate(_p, _ppvar, _thread, _nt, _lv);
return 0;
}
static void _n_rate(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt, double _lv){ int _i, _j;
double _xi, _theta;
if (!usetable) {
_f_rate(_p, _ppvar, _thread, _nt, _lv); return;
}
_xi = _mfac_rate * (_lv - _tmin_rate);
if (isnan(_xi)) {
minf = _xi;
taum = _xi;
return;
}
if (_xi <= 0.) {
minf = _t_minf[0];
taum = _t_taum[0];
return; }
if (_xi >= 300.) {
minf = _t_minf[300];
taum = _t_taum[300];
return; }
_i = (int) _xi;
_theta = _xi - (double)_i;
minf = _t_minf[_i] + _theta*(_t_minf[_i+1] - _t_minf[_i]);
taum = _t_taum[_i] + _theta*(_t_taum[_i+1] - _t_taum[_i]);
}
static int _f_rate ( _threadargsprotocomma_ double _lv ) {
if ( fabs ( _lv + 34.5 ) < 1e-6 ) {
minf = 1.0 / ( 1.0 + exp ( ( _lv + 34.50001 ) / - 9.0 ) ) ;
}
else {
minf = 1.0 / ( 1.0 + exp ( ( _lv + 34.5 ) / - 9.0 ) ) ;
}
taum = 1.0 / ( ( 31.746 * ( pow( ( exp ( ( _lv - 5.0 ) / - 13.89 ) + 1.0 ) , - 1.0 ) ) ) + ( 3.97e-4 * ( _lv + 8.9 ) ) * ( pow( ( exp ( ( _lv + 8.9 ) / 5.0 ) - 1.0 ) , - 1.0 ) ) ) ;
taum = taum / qdeltat ;
return 0; }
static void _hoc_rate(void) {
double _r;
double* _p; Datum* _ppvar; Datum* _thread; _NrnThread* _nt;
if (_extcall_prop) {_p = _extcall_prop->param; _ppvar = _extcall_prop->dparam;}else{ _p = (double*)0; _ppvar = (Datum*)0; }
_thread = _extcall_thread;
_nt = nrn_threads;
#if 1
_check_rate(_p, _ppvar, _thread, _nt);
#endif
_r = 1.;
rate ( _p, _ppvar, _thread, _nt, *getarg(1) );
hoc_retpushx(_r);
}
static double _mfac_getGHKexp, _tmin_getGHKexp;
static void _check_getGHKexp(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt) {
static int _maktable=1; int _i, _j, _ix = 0;
double _xi, _tmax;
static double _sav_T;
if (!usetable) {return;}
if (_sav_T != T) { _maktable = 1;}
if (_maktable) { double _x, _dx; _maktable=0;
_tmin_getGHKexp = - 150.0 ;
_tmax = 100.0 ;
_dx = (_tmax - _tmin_getGHKexp)/300.; _mfac_getGHKexp = 1./_dx;
for (_i=0, _x=_tmin_getGHKexp; _i < 301; _x += _dx, _i++) {
_t_getGHKexp[_i] = _f_getGHKexp(_p, _ppvar, _thread, _nt, _x);
}
_sav_T = T;
}
}
double getGHKexp(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt, double _lv) {
#if 0
_check_getGHKexp(_p, _ppvar, _thread, _nt);
#endif
return _n_getGHKexp(_p, _ppvar, _thread, _nt, _lv);
}
static double _n_getGHKexp(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt, double _lv){ int _i, _j;
double _xi, _theta;
if (!usetable) {
return _f_getGHKexp(_p, _ppvar, _thread, _nt, _lv);
}
_xi = _mfac_getGHKexp * (_lv - _tmin_getGHKexp);
if (isnan(_xi)) {
return _xi; }
if (_xi <= 0.) {
return _t_getGHKexp[0];
}
if (_xi >= 300.) {
return _t_getGHKexp[300];
}
_i = (int) _xi;
return _t_getGHKexp[_i] + (_xi - (double)_i)*(_t_getGHKexp[_i+1] - _t_getGHKexp[_i]);
}
double _f_getGHKexp ( _threadargsprotocomma_ double _lv ) {
double _lgetGHKexp;
_lgetGHKexp = exp ( - 23.20764929 * _lv / T ) ;
return _lgetGHKexp;
}
static void _hoc_getGHKexp(void) {
double _r;
double* _p; Datum* _ppvar; Datum* _thread; _NrnThread* _nt;
if (_extcall_prop) {_p = _extcall_prop->param; _ppvar = _extcall_prop->dparam;}else{ _p = (double*)0; _ppvar = (Datum*)0; }
_thread = _extcall_thread;
_nt = nrn_threads;
#if 1
_check_getGHKexp(_p, _ppvar, _thread, _nt);
#endif
_r = getGHKexp ( _p, _ppvar, _thread, _nt, *getarg(1) );
hoc_retpushx(_r);
}
static int _ode_count(int _type){ return 1;}
static void _ode_spec(_NrnThread* _nt, _Memb_list* _ml, int _type) {
double* _p; Datum* _ppvar; Datum* _thread;
Node* _nd; double _v; int _iml, _cntml;
_cntml = _ml->_nodecount;
_thread = _ml->_thread;
for (_iml = 0; _iml < _cntml; ++_iml) {
_p = _ml->_data[_iml]; _ppvar = _ml->_pdata[_iml];
_nd = _ml->_nodelist[_iml];
v = NODEV(_nd);
cai = _ion_cai;
cao = _ion_cao;
_ode_spec1 (_p, _ppvar, _thread, _nt);
}}
static void _ode_map(int _ieq, double** _pv, double** _pvdot, double* _pp, Datum* _ppd, double* _atol, int _type) {
double* _p; Datum* _ppvar;
int _i; _p = _pp; _ppvar = _ppd;
_cvode_ieq = _ieq;
for (_i=0; _i < 1; ++_i) {
_pv[_i] = _pp + _slist1[_i]; _pvdot[_i] = _pp + _dlist1[_i];
_cvode_abstol(_atollist, _atol, _i);
}
}
static void _ode_matsol_instance1(_threadargsproto_) {
_ode_matsol1 (_p, _ppvar, _thread, _nt);
}
static void _ode_matsol(_NrnThread* _nt, _Memb_list* _ml, int _type) {
double* _p; Datum* _ppvar; Datum* _thread;
Node* _nd; double _v; int _iml, _cntml;
_cntml = _ml->_nodecount;
_thread = _ml->_thread;
for (_iml = 0; _iml < _cntml; ++_iml) {
_p = _ml->_data[_iml]; _ppvar = _ml->_pdata[_iml];
_nd = _ml->_nodelist[_iml];
v = NODEV(_nd);
cai = _ion_cai;
cao = _ion_cao;
_ode_matsol_instance1(_threadargs_);
}}
extern void nrn_update_ion_pointer(Symbol*, Datum*, int, int);
static void _update_ion_pointer(Datum* _ppvar) {
nrn_update_ion_pointer(_ca_sym, _ppvar, 0, 1);
nrn_update_ion_pointer(_ca_sym, _ppvar, 1, 2);
nrn_update_ion_pointer(_ca_sym, _ppvar, 2, 3);
nrn_update_ion_pointer(_ca_sym, _ppvar, 3, 4);
}
static void initmodel(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt) {
int _i; double _save;{
m = m0;
{
T = 273.15 + celsius ;
rate ( _threadargscomma_ v ) ;
m = minf ;
}
}
}
static void nrn_init(_NrnThread* _nt, _Memb_list* _ml, int _type){
double* _p; Datum* _ppvar; Datum* _thread;
Node *_nd; double _v; int* _ni; int _iml, _cntml;
#if CACHEVEC
_ni = _ml->_nodeindices;
#endif
_cntml = _ml->_nodecount;
_thread = _ml->_thread;
for (_iml = 0; _iml < _cntml; ++_iml) {
_p = _ml->_data[_iml]; _ppvar = _ml->_pdata[_iml];
#if 0
_check_rate(_p, _ppvar, _thread, _nt);
_check_getGHKexp(_p, _ppvar, _thread, _nt);
#endif
#if CACHEVEC
if (use_cachevec) {
_v = VEC_V(_ni[_iml]);
}else
#endif
{
_nd = _ml->_nodelist[_iml];
_v = NODEV(_nd);
}
v = _v;
cai = _ion_cai;
cao = _ion_cao;
initmodel(_p, _ppvar, _thread, _nt);
}
}
static double _nrn_current(double* _p, Datum* _ppvar, Datum* _thread, _NrnThread* _nt, double _v){double _current=0.;v=_v;{ {
A = getGHKexp ( _threadargscomma_ v ) ;
ica = perm * m * m * m * ( 4.47814e6 * v / T ) * ( ( cai / 1000.0 ) - ( cao / 1000.0 ) * A ) / ( 1.0 - A ) ;
}
_current += ica;
} return _current;
}
static void nrn_cur(_NrnThread* _nt, _Memb_list* _ml, int _type) {
double* _p; Datum* _ppvar; Datum* _thread;
Node *_nd; int* _ni; double _rhs, _v; int _iml, _cntml;
#if CACHEVEC
_ni = _ml->_nodeindices;
#endif
_cntml = _ml->_nodecount;
_thread = _ml->_thread;
for (_iml = 0; _iml < _cntml; ++_iml) {
_p = _ml->_data[_iml]; _ppvar = _ml->_pdata[_iml];
#if CACHEVEC
if (use_cachevec) {
_v = VEC_V(_ni[_iml]);
}else
#endif
{
_nd = _ml->_nodelist[_iml];
_v = NODEV(_nd);
}
cai = _ion_cai;
cao = _ion_cao;
_g = _nrn_current(_p, _ppvar, _thread, _nt, _v + .001);
{ double _dica;
_dica = ica;
_rhs = _nrn_current(_p, _ppvar, _thread, _nt, _v);
_ion_dicadv += (_dica - ica)/.001 ;
}
_g = (_g - _rhs)/.001;
_ion_ica += ica ;
#if CACHEVEC
if (use_cachevec) {
VEC_RHS(_ni[_iml]) -= _rhs;
}else
#endif
{
NODERHS(_nd) -= _rhs;
}
}
}
static void nrn_jacob(_NrnThread* _nt, _Memb_list* _ml, int _type) {
double* _p; Datum* _ppvar; Datum* _thread;
Node *_nd; int* _ni; int _iml, _cntml;
#if CACHEVEC
_ni = _ml->_nodeindices;
#endif
_cntml = _ml->_nodecount;
_thread = _ml->_thread;
for (_iml = 0; _iml < _cntml; ++_iml) {
_p = _ml->_data[_iml];
#if CACHEVEC
if (use_cachevec) {
VEC_D(_ni[_iml]) += _g;
}else
#endif
{
_nd = _ml->_nodelist[_iml];
NODED(_nd) += _g;
}
}
}
static void nrn_state(_NrnThread* _nt, _Memb_list* _ml, int _type) {
double* _p; Datum* _ppvar; Datum* _thread;
Node *_nd; double _v = 0.0; int* _ni; int _iml, _cntml;
#if CACHEVEC
_ni = _ml->_nodeindices;
#endif
_cntml = _ml->_nodecount;
_thread = _ml->_thread;
for (_iml = 0; _iml < _cntml; ++_iml) {
_p = _ml->_data[_iml]; _ppvar = _ml->_pdata[_iml];
_nd = _ml->_nodelist[_iml];
#if CACHEVEC
if (use_cachevec) {
_v = VEC_V(_ni[_iml]);
}else
#endif
{
_nd = _ml->_nodelist[_iml];
_v = NODEV(_nd);
}
v=_v;
{
cai = _ion_cai;
cao = _ion_cao;
{ states(_p, _ppvar, _thread, _nt);
} }}
}
static void terminal(){}
static void _initlists(){
double _x; double* _p = &_x;
int _i; static int _first = 1;
if (!_first) return;
_slist1[0] = &(m) - _p; _dlist1[0] = &(Dm) - _p;
_t_minf = makevector(301*sizeof(double));
_t_taum = makevector(301*sizeof(double));
_t_getGHKexp = makevector(301*sizeof(double));
_first = 0;
}
#if defined(__cplusplus)
} /* extern "C" */
#endif
#if NMODL_TEXT
static const char* nmodl_filename = "dcn_CaHVA.mod";
static const char* nmodl_file_text =
"TITLE High voltage activated calcium current (CaHVA) of deep cerebellar nucleus (DCN) neuron\n"
"COMMENT\n"
" Translated from GENESIS by Johannes Luthman and Volker Steuber. \n"
" This mechanism and the other calcium channel (CaLVA.mod) are the only channel\n"
" mechanisms of the DCN model that use the GHK mechanism to calculate reversal\n"
" potential. Thus, extracellular Ca concentration is of importance and shall be \n"
" set from hoc.\n"
"\n"
" Calcium entering the neuron through this channel is kept track of by CaConc.mod\n"
" and affects the conductance of the SK.mod channel mechanism.\n"
"ENDCOMMENT \n"
"\n"
"NEURON { \n"
" SUFFIX dcnCaHVA \n"
" USEION ca READ cai, cao WRITE ica\n"
" RANGE perm, ica, m, cai\n"
" GLOBAL qdeltat\n"
"} \n"
" \n"
"UNITS { \n"
" (mA) = (milliamp) \n"
" (mV) = (millivolt) \n"
" (molar) = (1/liter)\n"
" (mM) = (millimolar)\n"
"} \n"
" \n"
"PARAMETER { \n"
" qdeltat = 1\n"
" perm = 7.5e-6 (cm/seconds)\n"
"} \n"
"\n"
"ASSIGNED {\n"
" v (mV)\n"
" cai (mM)\n"
" cao (mM) \n"
" ica (mA/cm2) \n"
" minf (1)\n"
" taum (ms) \n"
" celsius (degC)\n"
" T (kelvin)\n"
" A (1)\n"
"} \n"
"\n"
"STATE {\n"
" m\n"
"} \n"
"\n"
"INITIAL { \n"
" T = 273.15 + celsius\n"
" rate(v)\n"
" m = minf \n"
"} \n"
" \n"
"BREAKPOINT { \n"
" SOLVE states METHOD cnexp \n"
" A = getGHKexp(v)\n"
" : \"4.47814e6 * v / T\" in the following is the simplification of the GHK\n"
" : current equation's (z^2 * F^2 * (0.001) * v) / (R * T). [*(0.001) is to get \n"
" : volt from NEURON's mV.] Together with the simplification in getGHKexp() \n"
" : (below), this speeds up the whole DCN simulation (without synapses) by 8%.\n"
" : The division of the calcium concentrations (mM) by 1000 gives molar as \n"
" : required by the GHK current equation.\n"
" ica = perm * m*m*m * (4.47814e6 * v / T) * ((cai/1000) - (cao/1000) * A) / (1 - A)\n"
"} \n"
" \n"
"DERIVATIVE states { \n"
" rate(v) \n"
" m' = (minf - m)/taum \n"
"} \n"
"\n"
"PROCEDURE rate(v(mV)) {\n"
" TABLE minf, taum FROM -150 TO 100 WITH 300\n"
" if(fabs(v + 34.5) < 1e-6) {\n"
" minf = 1 / (1 + exp((v + 34.50001) / -9))\n"
" } else {\n"
" minf = 1 / (1 + exp((v + 34.5) / -9))\n"
" }\n"
" taum = 1 / ((31.746 * ((exp((v - 5) / -13.89) + 1) ^ -1)) + (3.97e-4 * (v + 8.9)) * ((exp((v + 8.9) / 5) - 1) ^ -1))\n"
" taum = taum / qdeltat\n"
"} \n"
"\n"
"FUNCTION getGHKexp(v(mV)) {\n"
" TABLE DEPEND T FROM -150 TO 100 WITH 300 \n"
" getGHKexp = exp(-23.20764929 * v / T): =the calculated values of\n"
" : getGHKexp = exp((-z * F * (0.001) * v) / (R * T)).\n"
"}\n"
;
#endif