# -*- coding: utf-8 -*-
## Description: class ChannelML for loading ChannelML from file or xml element into MOOSE
## Version 1.0 by Aditya Gilra, NCBS, Bangalore, India, 2011 for serial MOOSE
## Version 1.5 by Niraj Dudani, NCBS, Bangalore, India, 2012, ported to parallel MOOSE
## Version 1.6 by Aditya Gilra, NCBS, Bangalore, India, 2012, minor changes for parallel MOOSE
## Version 1.7 by Aditya Gilra, NCBS, Bangalore, India, 2013, further support for NeuroML 1.8.1
## Version 1.8 by Aditya Gilra, NCBS, Bangalore, India, 2013, changes for new IntFire and SynHandler classes
"""
NeuroML.py is the preferred interface. Use this only if NeuroML L1,L2,L3 files are misnamed/scattered.
Instantiate ChannelML class, and thence use method:
readChannelMLFromFile(...) to load a standalone ChannelML file (synapse/channel), OR
readChannelML(...) / readSynapseML to load from an xml.etree xml element (could be part of a larger NeuroML file).
"""
from __future__ import print_function, division
from xml.etree import cElementTree as ET
import string
import os
import sys
import math
import moose
from moose.neuroml import utils
import moose.utils as mu
import moose.print_utils as pu
class ChannelML():
def __init__(self,nml_params):
self.cml='http://morphml.org/channelml/schema'
self.temperature = nml_params['temperature']
def readChannelMLFromFile(self,filename,params={}):
""" specify params as a dict.
Currently none implemented.
For setting channel temperature,
pass in via nml_params during instance creation (see __init__())
or set <instance>.temperature."""
tree = ET.parse(filename)
channelml_element = tree.getroot()
for channel in channelml_element.findall('.//{'+self.cml+'}channel_type'):
## ideally I should read in extra params from within the channel_type element
## and put those in also. Global params should override local ones.
self.readChannelML(channel,params,channelml_element.attrib['units'])
for synapse in channelml_element.findall('.//{'+self.cml+'}synapse_type'):
self.readSynapseML(synapse,channelml_element.attrib['units'])
for ionConc in channelml_element.findall('.//{'+self.cml+'}ion_concentration'):
self.readIonConcML(ionConc,channelml_element.attrib['units'])
def readSynapseML(self,synapseElement,units="SI units"):
if 'Physiological Units' in units: # see pg 219 (sec 13.2) of Book of Genesis
Vfactor = 1e-3 # V from mV
Tfactor = 1e-3 # s from ms
Gfactor = 1e-3 # S from mS
elif 'SI Units' in units:
Vfactor = 1.0
Tfactor = 1.0
Gfactor = 1.0
else:
pu.fatal("Wrong units %s exiting ..." % units)
sys.exit(1)
moose.Neutral('/library') # creates /library in MOOSE tree; elif present, wraps
synname = synapseElement.attrib['name']
if utils.neuroml_debug:
pu.info("Loading synapse : %s into /library" % synname)
moosesynapse = moose.SynChan('/library/'+synname)
doub_exp_syn = synapseElement.find('./{'+self.cml+'}doub_exp_syn')
moosesynapse.Ek = float(doub_exp_syn.attrib['reversal_potential'])*Vfactor
moosesynapse.Gbar = float(doub_exp_syn.attrib['max_conductance'])*Gfactor
moosesynapse.tau1 = float(doub_exp_syn.attrib['rise_time'])*Tfactor # seconds
moosesynapse.tau2 = float(doub_exp_syn.attrib['decay_time'])*Tfactor # seconds
## The delay and weight can be set only after connecting a spike event generator.
## delay and weight are arrays: multiple event messages can be connected to a single synapse
## graded synapses are not supported by neuroml, so set to False here,
## see my Demo/neuroml/lobster_pyloric/STG_net.py for how to still have graded synapses
moosesynapse_graded = moose.Mstring(moosesynapse.path+'/graded')
moosesynapse_graded.value = 'False'
moosesynapse_mgblock = moose.Mstring(moosesynapse.path+'/mgblockStr')
moosesynapse_mgblock.value = 'False'
## check if STDP synapse is present or not
stdp_syn = synapseElement.find('./{'+self.cml+'}stdp_syn')
if stdp_syn is None:
moosesynhandler = moose.SimpleSynHandler('/library/'+synname+'/handler')
else:
moosesynhandler = moose.STDPSynHandler('/library/'+synname+'/handler')
moosesynhandler.aPlus0 = float(stdp_syn.attrib['del_weight_ltp'])
moosesynhandler.aMinus0 = float(stdp_syn.attrib['del_weight_ltd'])
moosesynhandler.tauPlus = float(stdp_syn.attrib['tau_ltp'])
moosesynhandler.tauMinus = float(stdp_syn.attrib['tau_ltd'])
moosesynhandler.weightMax = float(stdp_syn.attrib['max_syn_weight'])
moosesynhandler.weightMin = 0.0
## connect the SimpleSynHandler or the STDPSynHandler to the SynChan (double exp)
moose.connect( moosesynhandler, 'activationOut', moosesynapse, 'activation' )
def readChannelML(self,channelElement,params={},units="SI units"):
## I first calculate all functions assuming a consistent system of units.
## While filling in the A and B tables, I just convert to SI.
## Also convert gmax and Erev.
if 'Physiological Units' in units: # see pg 219 (sec 13.2) of Book of Genesis
Vfactor = 1e-3 # V from mV
Tfactor = 1e-3 # s from ms
Gfactor = 1e1 # S/m^2 from mS/cm^2
concfactor = 1e6 # Mol = mol/m^-3 from mol/cm^-3
elif 'SI Units' in units:
Vfactor = 1.0
Tfactor = 1.0
Gfactor = 1.0
concfactor = 1.0
else:
pu.fatal("Wrong units %s. Existing" % units)
sys.exit(1)
moose.Neutral('/library') # creates /library in MOOSE tree; elif present, wraps
channel_name = channelElement.attrib['name']
if utils.neuroml_debug:
pu.info("Loading channel %s into /library" % channel_name)
IVrelation = channelElement.find('./{'+self.cml+'}current_voltage_relation')
intfire = IVrelation.find('./{'+self.cml+'}integrate_and_fire')
if intfire is not None:
## Below params need to be set while making an LIF compartment
moosechannel = moose.Neutral('/library/'+channel_name)
moosechannelval = moose.Mstring(moosechannel.path+'/vReset')
moosechannelval.value = str(float(intfire.attrib['v_reset'])*Vfactor)
moosechannelval = moose.Mstring(moosechannel.path+'/thresh')
moosechannelval.value = str(float(intfire.attrib['threshold'])*Vfactor)
moosechannelval = moose.Mstring(moosechannel.path+'/refracT')
moosechannelval.value = str(float(intfire.attrib['t_refrac'])*Tfactor)
## refracG is currently not supported by moose.LIF
## Confirm if g_refrac is a conductance density or not?
## assuming g_refrac is a conductance density below
moosechannelval = moose.Mstring(moosechannel.path+'/refracG')
moosechannelval.value = str(float(intfire.attrib['g_refrac'])*Gfactor)
## create an Mstring saying this is an integrate_and_fire mechanism
moosechannelval = moose.Mstring(moosechannel.path+'/integrate_and_fire')
moosechannelval.value = 'True'
return
concdep = IVrelation.find('./{'+self.cml+'}conc_dependence')
if concdep is None:
moosechannel = moose.HHChannel('/library/'+channel_name)
else:
moosechannel = moose.HHChannel2D('/library/'+channel_name)
if IVrelation.attrib['cond_law']=="ohmic":
moosechannel.Gbar = float(IVrelation.attrib['default_gmax']) * Gfactor
moosechannel.Ek = float(IVrelation.attrib['default_erev']) * Vfactor
moosechannelIon = moose.Mstring(moosechannel.path+'/ion')
moosechannelIon.value = IVrelation.attrib['ion']
if concdep is not None:
moosechannelIonDependency = moose.Mstring(moosechannel.path+'/ionDependency')
moosechannelIonDependency.value = concdep.attrib['ion']
nernstnote = IVrelation.find('./{'+utils.meta_ns+'}notes')
if nernstnote is not None:
## the text in nernstnote is "Nernst,Cout=<float>,z=<int>"
nernst_params = nernstnote.text.split(',')
if nernst_params[0] == 'Nernst':
nernstMstring = moose.Mstring(moosechannel.path+'/nernst_str')
nernstMstring.value = str( float(nernst_params[1].split('=')[1]) * concfactor ) + \
',' + str( int(nernst_params[2].split('=')[1]) )
gates = IVrelation.findall('./{'+self.cml+'}gate')
if len(gates)>3:
pu.fatal("Sorry! Maximum x, y, and z (three) gates are possible in MOOSE/Genesis")
sys.exit()
gate_full_name = [ 'gateX', 'gateY', 'gateZ' ] # These are the names that MOOSE uses to create gates.
## if impl_prefs tag is present change VMIN, VMAX and NDIVS
impl_prefs = channelElement.find('./{'+self.cml+'}impl_prefs')
if impl_prefs is not None:
table_settings = impl_prefs.find('./{'+self.cml+'}table_settings')
## some problem here... disable
VMIN_here = float(table_settings.attrib['min_v'])
VMAX_here = float(table_settings.attrib['max_v'])
NDIVS_here = int(table_settings.attrib['table_divisions'])
dv_here = (VMAX_here - VMIN_here) / NDIVS_here
else:
## default VMIN, VMAX and dv are in SI
## convert them to current calculation units used by channel definition
## while loading into tables, convert them back to SI
VMIN_here = utils.VMIN/Vfactor
VMAX_here = utils.VMAX/Vfactor
NDIVS_here = utils.NDIVS
dv_here = utils.dv/Vfactor
offset = IVrelation.find('./{'+self.cml+'}offset')
if offset is None: vNegOffset = 0.0
else: vNegOffset = float(offset.attrib['value'])
self.parameters = []
for parameter in channelElement.findall('.//{'+self.cml+'}parameter'):
self.parameters.append( (parameter.attrib['name'],float(parameter.attrib['value'])) )
for num,gate in enumerate(gates):
# if no q10settings tag, the q10factor remains 1.0
# if present but no gate attribute, then set q10factor
# if there is a gate attribute, then set it only if gate attrib matches gate name
self.q10factor = 1.0
self.gate_name = gate.attrib['name']
for q10settings in IVrelation.findall('./{'+self.cml+'}q10_settings'):
## self.temperature from neuro.utils
if 'gate' in q10settings.attrib:
if q10settings.attrib['gate'] == self.gate_name:
self.setQ10(q10settings)
break
else:
self.setQ10(q10settings)
############### HHChannel2D crashing on setting Xpower!
#### temperamental! If you print something before, it gives cannot creategate from copied channel, else crashes
## Setting power first. This is necessary because it also
## initializes the gate's internal data structures as a side
## effect. Alternatively, gates can be initialized explicitly
## by calling HHChannel.createGate().
gate_power = float( gate.get( 'instances' ) )
if num == 0:
moosechannel.Xpower = gate_power
if concdep is not None: moosechannel.Xindex = "VOLT_C1_INDEX"
elif num == 1:
moosechannel.Ypower = gate_power
if concdep is not None: moosechannel.Yindex = "VOLT_C1_INDEX"
elif num == 2:
moosechannel.Zpower = gate_power
if concdep is not None: moosechannel.Zindex = "VOLT_C1_INDEX"
## Getting handle to gate using the gate's path.
gate_path = moosechannel.path + '/' + gate_full_name[ num ]
if concdep is None:
moosegate = moose.HHGate( gate_path )
## set SI values inside MOOSE
moosegate.min = VMIN_here*Vfactor
moosegate.max = VMAX_here*Vfactor
moosegate.divs = NDIVS_here
## V.IMP to get smooth curves, else even with 3000 divisions
## there are sudden transitions.
moosegate.useInterpolation = True
else:
moosegate = moose.HHGate2D( gate_path )
##### If alpha and beta functions exist, make them here
for transition in gate.findall('./{'+self.cml+'}transition'):
## make python functions with names of transitions...
fn_name = transition.attrib['name']
## I assume that transitions if present are called alpha and beta
## for forward and backward transitions...
if fn_name in ['alpha','beta']:
self.make_cml_function(transition, fn_name, concdep)
else:
pu.fatal("Unsupported transition %s" % fn_name)
sys.exit()
time_course = gate.find('./{'+self.cml+'}time_course')
## tau is divided by self.q10factor in make_function()
## thus, it gets divided irrespective of <time_course> tag present or not.
if time_course is not None:
self.make_cml_function(time_course, 'tau', concdep)
steady_state = gate.find('./{'+self.cml+'}steady_state')
if steady_state is not None:
self.make_cml_function(steady_state, 'inf', concdep)
if concdep is None: ca_name = '' # no Ca dependence
else: ca_name = ','+concdep.attrib['variable_name'] # Ca dependence
## Create tau() and inf() if not present, from alpha() and beta()
for fn_element,fn_name,fn_expr in [(time_course,'tau',"1/(alpha+beta)"),\
(steady_state,'inf',"alpha/(alpha+beta)")]:
## put in args for alpha and beta, could be v and Ca dep.
expr_string = fn_expr.replace('alpha', 'self.alpha(v'+ca_name+')')
expr_string = expr_string.replace('beta', 'self.beta(v'+ca_name+')')
## if time_course/steady_state are not present,
## then alpha annd beta transition elements should be present, and fns created.
if fn_element is None:
self.make_function(fn_name,'generic',expr_string=expr_string,concdep=concdep)
## non Ca dependent channel
if concdep is None:
## while calculating, use the units used in xml defn,
## while filling in table, I convert to SI units.
v0 = VMIN_here - vNegOffset
n_entries = NDIVS_here+1
tableA = [ 0.0 ] * n_entries
tableB = [ 0.0 ] * n_entries
for i in range(n_entries):
v = v0 + i * dv_here
inf = self.inf(v)
tau = self.tau(v)
## convert to SI before writing to table
## qfactor is already in inf and tau
tableA[i] = inf/tau / Tfactor
tableB[i] = 1.0/tau / Tfactor
moosegate.tableA = tableA
moosegate.tableB = tableB
## Ca dependent channel
else:
## UNITS: while calculating, use the units used in xml defn,
## while filling in table, I convert to SI units.
## Note here Ca units do not enter, but
## units of CaMIN, CaMAX and ca_conc in fn expr should match.
v = VMIN_here - vNegOffset
CaMIN = float(concdep.attrib['min_conc'])
CaMAX = float(concdep.attrib['max_conc'])
CaNDIVS = 100
dCa = (CaMAX-CaMIN)/CaNDIVS
## CAREFUL!: tableA = [[0.0]*(CaNDIVS+1)]*(NDIVS_here+1) will not work!
## * does a shallow copy, same list will get repeated 200 times!
## Thus setting tableA[35][1] = 5.0 will set all rows, 1st col to 5.0!!!!
tableA = [[0.0]*(CaNDIVS+1) for i in range(NDIVS_here+1)]
tableB = [[0.0]*(CaNDIVS+1) for i in range(NDIVS_here+1)]
for i in range(NDIVS_here+1):
Ca = CaMIN
for j in range(CaNDIVS+1):
inf = self.inf(v,Ca)
tau = self.tau(v,Ca)
## convert to SI (Tfactor) before writing to table
## qfactor is already in inf and tau
tableA[i][j] = inf/tau / Tfactor
tableB[i][j] = 1.0/tau / Tfactor
Ca += dCa
v += dv_here
## Presently HHGate2D doesn't allow the setting of tables as 2D vectors directly
moosegate.tableA = tableA
moosegate.tableB = tableB
## set SI values inside MOOSE
moosegate.xminA = VMIN_here*Vfactor
moosegate.xmaxA = VMAX_here*Vfactor
moosegate.xdivsA = NDIVS_here
#moosegate.dxA = dv_here*Vfactor
moosegate.yminA = CaMIN*concfactor
moosegate.ymaxA = CaMAX*concfactor
moosegate.ydivsA = CaNDIVS
#moosegate.dyB = dCa*concfactor
## set SI values inside MOOSE
moosegate.xminB = VMIN_here*Vfactor
moosegate.xmaxB = VMAX_here*Vfactor
moosegate.xdivsB = NDIVS_here
#moosegate.dxB = dv_here*Vfactor
moosegate.yminB = CaMIN*concfactor
moosegate.ymaxB = CaMAX*concfactor
moosegate.ydivsB = CaNDIVS
#moosegate.dyB = dCa*concfactor
def setQ10(self,q10settings):
if 'q10_factor' in q10settings.attrib:
self.q10factor = float(q10settings.attrib['q10_factor'])\
**((self.temperature-float(q10settings.attrib['experimental_temp']))/10.0)
elif 'fixed_q10' in q10settings.attrib:
self.q10factor = float(q10settings.attrib['fixed_q10'])
def readIonConcML(self, ionConcElement, units="SI units"):
if units == 'Physiological Units': # see pg 219 (sec 13.2) of Book of Genesis
Vfactor = 1e-3 # V from mV
Tfactor = 1e-3 # s from ms
Gfactor = 1e1 # S/m^2 from mS/cm^2
concfactor = 1e6 # mol/m^3 from mol/cm^3
Lfactor = 1e-2 # m from cm
Ifactor = 1e-6 # A from microA
else:
Vfactor = 1.0
Tfactor = 1.0
Gfactor = 1.0
concfactor = 1.0
Lfactor = 1.0
Ifactor = 1.0
moose.Neutral('/library') # creates /library in MOOSE tree; elif present, wraps
ionSpecies = ionConcElement.find('./{'+self.cml+'}ion_species')
if ionSpecies is not None:
if not 'ca' in ionSpecies.attrib['name']:
pu.fatal("Sorry, I cannot handle non-Ca-ion pools. Exiting ...")
sys.exit(1)
capoolName = ionConcElement.attrib['name']
pu.info("Loading Ca pool %s into /library ." % capoolName)
caPool = moose.CaConc('/library/'+capoolName)
poolModel = ionConcElement.find('./{'+self.cml+'}decaying_pool_model')
caPool.CaBasal = float(poolModel.attrib['resting_conc']) * concfactor
caPool.Ca_base = float(poolModel.attrib['resting_conc']) * concfactor
if 'decay_constant' in poolModel.attrib:
caPool.tau = float(poolModel.attrib['decay_constant']) * Tfactor
elif 'inv_decay_constant' in poolModel.attrib:
caPool.tau = 1.0/float(poolModel.attrib['inv_decay_constant']) * Tfactor
## Only one of pool_volume_info or fixed_pool_info should be present, but not checking
volInfo = poolModel.find('./{'+self.cml+'}pool_volume_info')
if volInfo is not None:
caPool.thick = float(volInfo.attrib['shell_thickness']) * Lfactor
fixedPoolInfo = poolModel.find('./{'+self.cml+'}fixed_pool_info')
if fixedPoolInfo is not None:
## Put in phi under the caPool, so that it can
## be used instead of thickness to set B (see section 19.2 in Book of Genesis)
caPool_phi = moose.Mstring(caPool.path+'/phi')
caPool_phi.value = str( float(fixedPoolInfo.attrib['phi']) * concfactor/Ifactor/Tfactor )
## Put a high ceiling and floor for the Ca conc
#~ self.context.runG('setfield ' + caPool.path + ' ceiling 1e6')
#~ self.context.runG('setfield ' + caPool.path + ' floor 0.0')
def make_cml_function(self, element, fn_name, concdep=None):
fn_type = element.attrib['expr_form']
if fn_type in ['exponential','sigmoid','exp_linear']:
self.make_function( fn_name, fn_type, rate=float(element.attrib['rate']),\
midpoint=float(element.attrib['midpoint']), scale=float(element.attrib['scale'] ) )
elif fn_type == 'generic':
## OOPS! These expressions should be in SI units, since I converted to SI
## Ideally I should not convert to SI and have the user consistently use one or the other
## Or convert constants in these expressions to SI, only voltage and ca_conc appear!
## NO! SEE ABOVE, for calculating values for tables, I use units specified in xml file,
## then I convert to SI while writing to MOOSE internal tables.
expr_string = element.attrib['expr']
if concdep is None: ca_name = '' # no Ca dependence
else: ca_name = ','+concdep.attrib['variable_name'] # Ca dependence
expr_string = expr_string.replace( 'alpha', 'self.alpha(v'+ca_name+')')
expr_string = expr_string.replace( 'beta', 'self.beta(v'+ca_name+')')
self.make_function( fn_name, fn_type, expr_string=expr_string, concdep=concdep )
else:
pu.fatal("Unsupported function type %s "% fn_type)
sys.exit()
def make_function(self, fn_name, fn_type, **kwargs):
""" This dynamically creates a function called fn_name
If fn_type is exponential, sigmoid or exp_linear,
**kwargs is a dict having keys rate, midpoint and scale.
If fin_type is generic, **kwargs is a dict having key expr_string """
if fn_type == 'exponential':
def fn(self,v):
return kwargs['rate']*math.exp((v-kwargs['midpoint'])/kwargs['scale'])
elif fn_type == 'sigmoid':
def fn(self,v):
return kwargs['rate'] / ( 1 + math.exp((v-kwargs['midpoint'])/kwargs['scale']) )
elif fn_type == 'exp_linear':
def fn(self,v):
if v-kwargs['midpoint'] == 0.0: return kwargs['rate']
else:
return kwargs['rate'] * ((v-kwargs['midpoint'])/kwargs['scale']) \
/ ( 1 - math.exp((kwargs['midpoint']-v)/kwargs['scale']) )
elif fn_type == 'generic':
## python cannot evaluate the ternary operator ?:, so evaluate explicitly
## for security purposes eval() is not allowed any __builtins__
## but only safe functions in utils.safe_dict and
## only the local variables/functions v, self
allowed_locals = {'self':self}
allowed_locals.update(utils.safe_dict)
def fn(self,v,ca=None):
expr_str = kwargs['expr_string']
allowed_locals['v'] = v
allowed_locals['celsius'] = self.temperature
allowed_locals['temp_adj_'+self.gate_name] = self.q10factor
for i,parameter in enumerate(self.parameters):
allowed_locals[parameter[0]] = self.parameters[i][1]
if 'concdep' in kwargs:
concdep = kwargs['concdep']
## ca should appear as neuroML defined 'variable_name' to eval()
if concdep is not None:
allowed_locals[concdep.attrib['variable_name']] = ca
if '?' in expr_str:
condition, alternatives = expr_str.split('?',1)
alternativeTrue, alternativeFalse = alternatives.split(':',1)
if eval(condition,{"__builtins__":None},allowed_locals):
val = eval(alternativeTrue,{"__builtins__":None},allowed_locals)
else:
val = eval(alternativeFalse,{"__builtins__":None},allowed_locals)
else:
val = eval(expr_str,{"__builtins__" : None},allowed_locals)
if fn_name == 'tau':
return val/self.q10factor
else:
return val
fn.__name__ = fn_name
setattr(self.__class__, fn.__name__, fn)
return None
def make_new_synapse(syn_name, postcomp, syn_name_full, nml_params):
## if channel does not exist in library load it from xml file
if not moose.exists('/library/'+syn_name):
cmlR = ChannelML(nml_params)
model_filename = syn_name+'.xml'
model_path = utils.find_first_file(model_filename,nml_params['model_dir'])
if model_path is not None:
cmlR.readChannelMLFromFile(model_path)
else:
raise IOError(
'For mechanism {0}: files {1} not found under {2}.'.format(
syn_name, model_filename, nml_params['model_dir']
)
)
## deep copies the library SynChan and SynHandler
## to instances under postcomp named as <arg3>
synid = moose.copy(moose.element('/library/'+syn_name),postcomp,syn_name_full)
#synhandlerid = moose.copy(moose.element('/library/'+syn_name+'/handler'), postcomp,syn_name_full+'/handler') This line was a bug: double handler
synhandler = moose.element( synid.path + '/handler' )
syn = moose.SynChan(synid)
synhandler = moose.element(synid.path + '/handler') # returns SimpleSynHandler or STDPSynHandler
## connect the SimpleSynHandler or the STDPSynHandler to the SynChan (double exp)
moose.connect( synhandler, 'activationOut', syn, 'activation' )
# mgblock connections if required
childmgblock = mu.get_child_Mstring(syn,'mgblockStr')
#### connect the post compartment to the synapse
if childmgblock.value=='True': # If NMDA synapse based on mgblock, connect to mgblock
mgblock = moose.Mg_block(syn.path+'/mgblock')
moose.connect(postcomp,"channel", mgblock, "channel")
else: # if SynChan or even NMDAChan, connect normally
moose.connect(postcomp,"channel", syn, "channel")