Tutorial: Population Level Modeling (with PopNet)#

Option 1: Converting existing networks#

Like BioNet for biophysically detailed modeling and PointNet with point-based networks, PopNet works with networks in the SONATA data format. PopNet supports simulating networks of individual cells at the population level. The first thing you have to do is modify the node-types and edge-types of an existing network to use Population level models (rather than models of individual cells).

See the SONATA users guide for details of a network set to work with BioNet (for an example of the BioNet network, see tutorial 3):

[1]:

import pandas as pd

pd.read_csv('sources/chapter06/converted_network/V1_node_types_bionet.csv', sep=' ')

[1]:

versus the equivalent form for PopNet:

[2]:

pd.read_csv('sources/chapter06/converted_network/V1_node_types_popnet.csv', sep=' ')

[2]:

Some things to note:

• Each node in the network now corresponds to a a population rather than a cell, and therefore model_type is node “population”, rather than individual biophysical/point types

• We have set model_template to dipde:Internal which will tell the simulator to use special DiPDE model types

• We are using new dynamic_params files with parameters that have been adjusted to appropiate range for DiPDE models.

• morophology_file and model_processing, which were used to set and processes individual cell morphologies, are no longer applicable.

See SONATA specification for edges css files here. Finally, when we run the simulation we must tell PopNet to cluster nodes together using the group_by property:

network = popnet.PopNetwork.from_config(configure, group_by='node_type_id')

Option 2: Building a network from scratch#

We will create a network of two populations: one population of excitatory cells and another of inhibitory cells. Then we will save the network into SONATA-formatted data files.

The first step is to use the NetworkBuilder to instantiate a new network with two populations:

[3]:

from bmtk.builder import NetworkBuilder


net = NetworkBuilder('V1')
net.add_nodes(
    pop_name='excitatory',  # name of specific population optional
    ei='e',  # Optional
    location='VisL4',  # Optional
    model_type='population',  # Required, indicates what types of cells are being model
    model_template='dipde:Internal',  # Required, instructs what DiPDE objects will be created
    dynamics_params='exc_model.json'  # Required, contains parameters used by DiPDE during initialization of object
)

net.add_nodes(
    pop_name='inhibitory',
    ei='i',
    model_type='population',
    model_template='dipde:Internal',
    dynamics_params='inh_model.json'
)

Next we will create connections between the two populations:

[4]:

net.add_edges(
    source={'ei': 'e'}, target={'ei': 'i'},
    syn_weight=0.005,
    nsyns=20,
    delay=0.002,
    dynamics_params='ExcToInh.json'
)

net.add_edges(
    source={'ei': 'i'}, target={'ei': 'e'},
    syn_weight=-0.002,
    nsyns=10,
    delay=0.002,
    dynamics_params='InhToExc.json'
)

[4]:

<bmtk.builder.connection_map.ConnectionMap at 0x7f5ea775c940>

and finally we must build and save the network:

[5]:

net.build()
net.save(output_dir='sim_ch06/network')

[6]:

input_net = NetworkBuilder('LGN')
input_net.add_nodes(
    pop_name='tON',
    ei='e',
    model_type='virtual'
)

input_net.add_edges(
    target=net.nodes(ei='e'),
    syn_weight=0.0025,
    nsyns=10,
    delay=0.002,
    dynamics_params='input_ExcToExc.json'
)

input_net.build()
input_net.save(output_dir='sim_ch06/network')

Inputs#

We next need to set the firing rates of the External Population. There are multiple ways to set this value which will be discussed later. The best way is to set the firing rates using a input-rates file for each External Population. We can fetch an input-rates existing file using the command:

$ cd sim_ch06
$ wget https://github.com/AllenInstitute/bmtk/raw/develop/examples/pop_2pops/inputs/external_rates.csv -O lgn_rates.csv

Then we must open the simulation_config.json file with a text editor and add the lgn_rates.csv file as a part of our inputs:

{
  "inputs": {
    "LGN_pop_rates": {
      "input_type": "csv",
      "module": "pop_rates",
      "rates": "${BASE_DIR}/lgn_rates.csv",
      "node_set": "LGN"
    }
  }
}