import os import numpy as np from bmtk.builder import NetworkBuilder from bmtk.builder.auxi.node_params import positions_columinar, xiter_random #build_recurrent_edges = True # List of non-virtual cell models bio_models = [ { 'model_name': 'Scnn1a', 'ei': 'e', 'model_template': 'nest:iaf_psc_alpha', 'dynamics_params': '472363762_point.json' }, { 'model_name': 'Rorb', 'ei': 'e', 'model_template': 'nest:iaf_psc_alpha', 'dynamics_params': '473863510_point.json' }, { 'model_name': 'Nr5a1', 'ei': 'e', 'model_template': 'nest:iaf_psc_alpha', 'dynamics_params': '473863035_point.json' }, { 'model_name': 'PV1', 'ei': 'i', 'model_template': 'nest:iaf_psc_alpha', 'dynamics_params': '472912177_point.json' }, { 'model_name': 'PV2', 'ei': 'i', 'model_template': 'nest:iaf_psc_alpha', 'dynamics_params': '473862421_point.json' } ] point_models = [ { 'model_name': 'LIF_exc', 'ei': 'e', 'model_template': 'nest:iaf_psc_alpha', 'dynamics_params': 'IntFire1_exc_point.json' }, { 'model_name': 'LIF_inh', 'ei': 'i', 'model_template': 'nest:iaf_psc_alpha', 'dynamics_params': 'IntFire1_inh_point.json' } ] # Build a network of 300 biophysical cells to simulate internal = NetworkBuilder("internal") for i, model_props in enumerate(bio_models): n_cells = 80 if model_props['ei'] == 'e' else 30 # 80% excitatory, 20% inhib # Randomly get positions uniformly distributed in a column positions = positions_columinar(N=n_cells, center=[0, 10.0, 0], max_radius=50.0, height=200.0) internal.add_nodes(N=n_cells, x=positions[:, 0], y=positions[:, 1], z=positions[:, 2], rotation_angle_yaxis=xiter_random(N=n_cells, min_x=0.0, max_x=2 * np.pi), # randomly rotate y axis rotation_angle_zaxis=xiter_random(N=n_cells, min_x=0.0, max_x=2 * np.pi), # model_type='point_process', orig_model='biophysical', **model_props) # Build intfire type cells for model_props in point_models: n_cells = 75 # Just assume 75 cells for both point inhibitory and point excitatory positions = positions_columinar(N=n_cells, center=[0, 10.0, 0], max_radius=50.0, height=200.0) internal.add_nodes(N=n_cells, x=positions[:, 0], y=positions[:, 1], z=positions[:, 2], model_type='point_process', orig_model='intfire', **model_props) def n_connections(src, trg, prob=0.1, min_syns=1, max_syns=5): """Referenced by add_edges() and called by build() for every source/target pair. For every given target/source pair will connect the two with a probability prob (excludes self-connections)""" if src.node_id == trg.node_id: return 0 return 0 if np.random.uniform() > prob else np.random.randint(min_syns, max_syns) # Connections onto biophysical components, use the connection map to save section and position of every synapse # exc --> exc connections internal.add_edges(source={'ei': 'e'}, target={'ei': 'e', 'orig_model': 'biophysical'}, connection_rule=n_connections, connection_params={'prob': 0.2}, dynamics_params='ExcToExc.json', model_template='static_synapse', syn_weight=2.5, delay=2.0) # exc --> inh connections internal.add_edges(source={'ei': 'e'}, target={'ei': 'i', 'orig_model': 'biophysical'}, connection_rule=n_connections, dynamics_params='ExcToInh.json', model_template='static_synapse', syn_weight=5.0, delay=2.0) # inh --> exc connections internal.add_edges(source={'ei': 'i'}, target={'ei': 'e', 'orig_model': 'biophysical'}, connection_rule=n_connections, dynamics_params='InhToExc.json', model_template='static_synapse', syn_weight=-6.5, delay=2.0) # inh --> inh connections internal.add_edges(source={'ei': 'i'}, target={'ei': 'i', 'orig_model': 'biophysical'}, connection_rule=n_connections, connection_params={'prob': 0.2}, dynamics_params='InhToInh.json', model_template='static_synapse', syn_weight=-3.0, delay=2.0) # For connections on point neurons it doesn't make sense to save syanpatic location internal.add_edges(source={'ei': 'e'}, target={'orig_model': 'intfire'}, connection_rule=n_connections, dynamics_params='instanteneousExc.json', model_template='static_synapse', syn_weight=3.0, delay=2.0) internal.add_edges(source={'ei': 'i'}, target={'orig_model': 'intfire'}, connection_rule=n_connections, dynamics_params='instanteneousInh.json', model_template='static_synapse', syn_weight=-4.0, delay=2.0) # Build and save internal network internal.build() print('Saving internal network') internal.save(output_dir='network') # Build a network of 100 virtual cells that will connect to and drive the simulation of the internal network print('Building external connections') external = NetworkBuilder("external") external.add_nodes(N=100, model_type='virtual', ei='e') # Targets all biophysical excitatory cells external.add_edges(target=internal.nodes(ei='e', orig_model='biophysical'), source=external.nodes(), connection_rule=lambda *_: np.random.randint(0, 5), dynamics_params='ExcToExc.json', model_template='static_synapse', delay=2.0, syn_weight=11.0) # Targets all biophysical inhibitory cells external.add_edges(target=internal.nodes(ei='i', orig_model='biophysical'), source=external.nodes(), connection_rule=lambda *_: np.random.randint(0, 5), dynamics_params='ExcToInh.json', model_template='static_synapse', delay=2.0, syn_weight=14.0) # Targets all intfire1 cells (exc and inh) external.add_edges(target=internal.nodes(orig_model='intfire'), source=external.nodes(), connection_rule=lambda *_: np.random.randint(0, 5), dynamics_params='instanteneousExc.json', model_template='static_synapse', delay=2.0, syn_weight=13.0) external.build() print('Saving external') external.save(output_dir='network')