demo in plot_simulate_gamma.py

examples/plot_simulate_gamma.py

@@ -82,7 +82,7 @@
 dpls[trial_idx].plot_tfr_morlet(freqs, n_cycles=7, tmin=tmin, ax=axes[1])
 
 ###############################################################################
-# As a final exercise, let us try to re-run the simulation with a tonic bias
+# Now, let us try to re-run the simulation with a tonic bias
 # applied to the L5 Pyramidal cells. Notice that the oscillation waveform is
 # more regular, with less noise due to the fact that the tonic depolarization
 # dominates over the influence of the Poisson drive. By default, a tonic bias
@@ -108,6 +108,21 @@
 from hnn_core.viz import plot_psd
 plot_psd(dpls[trial_idx], fmin=20., fmax=100., tmin=tmin)
 
+###############################################################################
+# Finally, we demonstrate the mechanistic link between PING and the GABAA decay
+# time constant (`tau2`). Using the same network/drive configuration as before,
+# we decrease `tau2` from 5 to 2 ms. This will shorten the effective
+# refactory period between L5 pyramidal cell spikes and increase the PING
+# frequency from ~50 to ~65 Hz.
+net.cell_properties['L5_pyramidal']['synapses']['gabaa']['tau2'] = 2
+dpls = simulate_dipole(net, n_trials=1)
+
+fig, axes = plt.subplots(3, 1, sharex=True, figsize=(6, 6),
+                         constrained_layout=True)
+dpls[trial_idx].plot(ax=axes[0], show=False)
+net.cell_response.plot_spikes_raster(ax=axes[1], show=False)
+dpls[trial_idx].plot_tfr_morlet(freqs, n_cycles=7, tmin=tmin, ax=axes[2])
+
 ###############################################################################
 # References
 # ----------

hnn_core/cell.py

@@ -160,6 +160,7 @@ def gid(self, gid):
     def _set_biophysics(self, p_secs):
         "Set the biophysics for the cell."
 
+        from .cells_default import _set_variable_mech
         # neuron syntax is used to set values for mechanisms
         # sec.gbar_mech = x sets value of gbar for mech to x for all segs
         # in a section. This method is significantly faster than using
@@ -177,10 +178,11 @@ def _set_biophysics(self, p_secs):
             for mech_name, p_mech in p_sec['mechs'].items():
                 sec.insert(mech_name)
                 for attr, val in p_mech.items():
-                    if hasattr(val, '__call__'):
+                    if val == 'variable':
                         sec.push()
                         for seg in sec:
-                            setattr(seg, attr, val(h.distance(seg.x)))
+                            val = _set_variable_mech(h.distance(seg.x))
+                            setattr(seg, attr, val)
                         h.pop_section()
                     else:
                         setattr(sec, attr, val)

hnn_core/cells_default.py

@@ -146,6 +146,11 @@ def _get_mechanisms(p_all, cell_type, section_names, mechanisms):
     return mech_props
 
 
+def _set_variable_mech(x):
+    # this is a temporary hack
+    return 1e-6 * np.exp(3e-3 * x)
+
+
 def basket(cell_name='L2Basket', gid=None):
     """Get layer 2 / layer 5 basket cells.
 
@@ -251,7 +256,7 @@ def pyramidal(cell_name, override_params=None, gid=None):
             syns = list(p_syn.keys())
             if cell_name == 'L5Pyr':
                 p_secs[key]['mechs'][
-                    'ar']['gbar_ar'] = lambda x: 1e-6 * np.exp(3e-3 * x)
+                    'ar']['gbar_ar'] = 'variable'
         p_secs[key]['syns'] = syns
 
     for sec_name in p_secs:

hnn_core/network.py

@@ -230,20 +230,22 @@ def copy(self):
         return net_copy
 
     def _set_cell_properties(self):
-        '''Populate the network with cell objects'''
-
-        for cell_type in self.pos_dict.keys():
-            if cell_type in self.cellname_list:
-                if cell_type in ('L2_pyramidal', 'L5_pyramidal'):
-                    p_secs, p_syn, topology, sect_loc = pyramidal(
-                        cell_name=_short_name(cell_type))
-                else:
-                    p_secs, p_syn, topology, sect_loc = basket(
-                        cell_name=_short_name(cell_type))
-                self.cell_properties[cell_type] = {'sections': p_secs,
-                                                   'synapses': p_syn,
-                                                   'topology': topology,
-                                                   'sect_loc': sect_loc}
+        '''Set cell properties by cell type
+
+        This relies on self.cellname_list and must be updated when introducting
+        new cell types to HNN.
+        '''
+        for cell_type in self.cellname_list:
+            if cell_type in ('L2_pyramidal', 'L5_pyramidal'):
+                p_secs, p_syn, topology, sect_loc = pyramidal(
+                    cell_name=_short_name(cell_type))
+            else:
+                p_secs, p_syn, topology, sect_loc = basket(
+                    cell_name=_short_name(cell_type))
+            self.cell_properties[cell_type] = {'sections': p_secs,
+                                               'synapses': p_syn,
+                                               'topology': topology,
+                                               'sect_loc': sect_loc}
 
     def add_evoked_drive(self, name, *, mu, sigma, numspikes,
                          sync_within_trial=False, location,

hnn_core/network_builder.py

@@ -13,7 +13,6 @@
     h.nrnunit_use_legacy(1)
 
 from .cell import Cell, _ArtificialCell
-from .cells_default import pyramidal, basket
 from .params import _long_name, _short_name
 from copy import deepcopy
 

hnn_core/tests/test_cell.py

@@ -61,7 +61,7 @@ def test_cell():
                     'gbar_km': 60
                 },
                 'ca': {
-                    'gbar_ca': lambda x: 3e-3 * x
+                    'gbar_ca': 'variable'
                 }
             }
         }