Skip to content

Latest commit

 

History

History
410 lines (307 loc) · 14.4 KB

README.md

File metadata and controls

410 lines (307 loc) · 14.4 KB
Raw

Amygdala Theta

Code by Tyler Banks. In partnership with Unal Lab (Tuna and Unal)

Modeling Basal Forebrain GABAergic Neuromodulation of the Amygdala Theta Rhythm

Quickstart

Generate the network and inputs:

python build_network.py homogenous

Run the experiments (using MPI):

RUNS=10 ./h99_run_multiple.sh

Analyze:

cd cases_paper
python analysis_hom.py && python analysis_hom.py --final
python plot_signal_phase.py --all
cd ..
python whole_picture.py

Raw data output is stored in cases_paper/figures_data.

Core Neuron Update

cd ./components/mechanisms
nrnivmodl -coreneuron modfiles
cd -
./components/mechanisms/x86_64/special run_network.py simulation_configECP_base.json

# Running with MPI
mpiexec -n 8 ./components/mechanisms/x86_64/special -mpi run_network.py simulation_configECP_base.json

Installing the environment

## See https://github.com/BlueBrain/CoreNeuron
module load mpi/mpich-x86_64 # also add to your .bashrc
git config --global core.autocrlf input
cd
python3.9 venv neuro-venv
source neuro-venv/bin/activate #place in your .bashrc
mkdir -p core_nrn_install/install
cd core_nrn_install
git clone https://github.com/neuronsimulator/nrn
cd nrn
mkdir build
cd build
cmake ..  -DNRN_ENABLE_CORENEURON=ON -DCORENRN_ENABLE_NMODL=ON  -DNRN_ENABLE_INTERVIEWS=OFF  -DNRN_ENABLE_MPI=ON  -DNRN_ENABLE_RX3D=OFF  -DCMAKE_INSTALL_PREFIX=${HOME}/core_nrn_install/install -DPYTHON_EXECUTABLE=$(which python) -DCMAKE_C_COMPILER=$(which cc) -DCMAKE_CXX_COMPILER=$(which c++)
cmake --build . --parallel 8 --target install

Add the following to your bashrc
export PYTHONPATH=$HOME/core_nrn_install/install/lib/python:$PYTHONPATH
export PATH=$HOME/core_nrn_install/install/bin:$PATH

Running the Model

1. Build network input files :

Input file Purpose Generating code
thalamus_pyr_spikes.h5 Thalamic 2Hz poisson input to Pyramidal A and C cells build_input.py
thalamus_som_spikes.h5 Thalamic 2Hz poisson input to SOM cells build_input.py
thalamus_cr_spikes.h5 Thalamic 2Hz poisson input CR+ cells build_input.py
vpsi_inh_spikes_nonrhythmic.h5 VPSI poisson input to PN A and C for non-rhythmic inhibition build_input.py
shell_spikes.h5 Edge effects shell spikes build_input_shell.py
vpsi_inh_spikes.h5 VPSI 8Hz Rhythmic 3Hz poisson input to PN A and C matlab/generatethetainputs.m & matlab/convert_spikesmatrix.py

(primary files bold)

To generate the above files create Thalamic and VSPI inputs using the following. ** build_network.py will run this automatically with the appropriate parameters **

python build_input.py

To generate 8Hz rhythmic inputs (based on Fink et al 2015). Repository comes with this file so, following these steps isn't completely necessary. If you want to edit the parameters or rate at which the rhythmic inhibition is presented, run the following: If you change the scale you will need to re-run this with the correct scale variable set!

matlab &
generatethetainputs

The files will then need to be converted to the Sonata format BMTK understands.

python matlab/convert_spikematrix.py

Once these steps have been completed, input files WILL NOT need to be regenerated again.

2. Build network configuration files

Building of the network can be customized by altering build_network.py configuration at the beginning of the script.

To generate necessary network specification files in the [network](./network) directory, execute

python build_network.py

If you have a shell of cells to deal with edge effects then the shell_spikes.h5 file will be re-generated.

To generate a homogenous network then run

python build_network.py homogenous

To generate Feng's original homogenous network

python build_network.py feng_homogenous

3. Execute run script

The network can be tested using any one of the simulation configuration files listed below (python run_network.py [configuration file])

Configuration file Input Details Notes
simulation_config.json Thalamic 2Hz to PN, INT, SOM, CR, VPSI 2Hz exc and 8Hz inh to PN/INT deprecated example
simulation_configECP.json Thalamic 2Hz to PN, SOM, CR, VPSI 2Hz exc and 8Hz inh to PN/INT deprecated example with LFP recording electrodes linear_electrode.csv
simulation_configECP_base.json Thalamic 2Hz to PN, SOM, CR Quiet-waking state, baseline configuration file
simulation_configECP_base_edge_effects.json Thalamic 2Hz to PN, SOM, CR Quiet-waking state, baseline configuration file WITH EDGE EFFECT spikes presented
simulation_configECP_base_vclamp.json Thalamic 2Hz to PN, SOM, CR Same as simulation_configECP_base.json, 11 voltage clamped PN cells, igaba from SOM+ synapses are recorded and placed in outputECP/syn_report.h5 - Analysis completed using ipsc_analysis.m notes below
simulation_configECP_gamma.json Gamma testing for original model replicated from Feng et al 2019
simulation_configECP_vpsi.json Thalamic 2Hz to PN, SOM, CR, VPSI 8Hz rhythmic inh to PN/INT Primary VPSP input test
simulation_configECP_vpsi_edge_effects.json Thalamic 2Hz to PN, SOM, CR, VPSI 8Hz rhythmic inh to PN/INT Primary VPSP input test WITH EDGE EFFECT spikes presented
simulation_configECP_vpsi_vclamp.json Thalamic 2Hz to PN, SOM, CR, VPSI 8Hz rhythmic inh to PN/INT Same as simulation_configECP_vpsi.json, 11 voltage clamped PN cells, igaba from SOM+ synapses are recorded and placed in outputECP/syn_report.h5 - Analysis completed using ipsc_analysis.m notes below
simulation_configECP_vpsi_vclamp_nonrhythmic.json Thalamic 2Hz to PN, SOM, CR, VPSI 3Hz Poisson inh to PN/INT Testing Non-rhythmic inhibition to PN/PV
simulation_configECP_base_homogenous.json Thalamic 2Hz to PN, SOM, CR Quiet-waking state, baseline configuration file for HOMOGENOUS network
simulation_configECP_vpsi_homogenous.json Thalamic 2Hz to PN, SOM, CR, VPSI 8Hz rhythmic inh to PN/INT Primary VPSP input test for HOMOGENOUS network
simulation_configECP_base_feng_homogenous.json Thalamic 2Hz to PN Quiet-waking state, baseline configuration file for FENG'S ORIGINAL HOMOGENOUS NETWORK network

Primary tests bold

Single Core Mode

1000 Cell models typically run for 4-6 hours on a single core.

python run_network.py simulation_configECP_base_edge_effects.json

Parallel Mode

1000 Cell models run for 5-6 minutes on ~50 cores.

mpirun -n 50 nrniv -mpi -python run_network.py simulation_configECP_base_edge_effects.json

OR

sbatch batch_run.sh

Theta Test

mpirun -n 50 nrniv -mpi -python run_network.py simulation_configECP_vpsi_edge_effects.json

OR

sbatch batch_vpsi_run.sh

Homogenous Tests (Base and VPSI)

mpirun -n 50 nrniv -mpi -python run_network.py simulation_configECP_base_homogenous.json
mpirun -n 50 nrniv -mpi -python run_network.py simulation_configECP_vpsi_homogenous.json

OR

sbatch batch_run_hom.sh
sbatch batch_vpsi_run_hom.sh

Feng Homogenous Tests (Base)

mpirun -n 50 nrniv -mpi -python run_network.py simulation_configECP_base_feng_homogenous.json

OR

sbatch batch_run_feng_hom.sh

Analysis of the model

Analysis of the model is primarily comprised of a spike raster, mean firing rates, raw LFP, and LFP PSD. Launch using:

python analysis.py

To simply get firing rates for quick analysis run

python analysis.py --no-plots

loading outputECP/spikes.h5
done
Type : mean (std)
PN : 0.84 (0.93)
PV : 46.33 (45.58)
SOM : 8.67 (6.07)
CR : 3.67 (1.37)

analysis.m

NOT RECOMMENDED (slow on remote servers)

Plots can also be generated in MATLAB. A spike raster, mean firing rates, LFP and LFP PSD on 4 separate graphs.

analysis('../outputECP/ecp.h5','../outputECP/spikes.h5');

ipsc_analysis.m

Used in conjuction with simulation_configECP_base_vclamp.json and simulation_configECP_vpsi_vclamp.json to sum igaba currents and perform a PSD to produce the raw signal and PSD plots.

ipsc_analysis('../outputECP/syn_report.h5';

Note: use get_som2pn_targets.py to determine the correct PN cells to set in node_sets.json prior to running the simulation.

connection_info.py

Used to print the connectivity between cell types.

> python connection_info.py
        total uni   bi
PN->PN	2.0	  1.96	0.04
PN->PV	26.82	11.24	15.58
PN->SOM	31.19	29.17	2.01
PN->CR	18.43	16.41	2.02
PV->PN	52.0	36.42	15.58
PV->PV	22.92	17.41	5.5
PV->SOM	55.8	 55.8	  0.0
PV->CR	0.0	  0.0	  0.0
SOM->PN	6.57	4.55	2.01
SOM->PV	0.0	  0.0	  0.0
SOM->SOM	0.0	0.0	  0.0
SOM->CR	0.0   0.0   0.0
CR->PN	11.59	9.57	2.02
CR->PV	29.7	29.7	0.0
CR->SOM	75.25	75.25	0.0
CR->CR	0.0	  0.0	  0.0

Other important files

File/Directory Description
connectors.py This is where you should define all connection rules - keeps build_network.py clean
synapses.py When adding new synapse types to be used by build_network.py they must be defined here
components/templates/feng.hoc Contains the PN A, PN C and INT hoc cell templates
components/templates/SOM.hoc Contains the SOM and CR hoc cell templates
components/synaptic_models Contains parameter configuration files for synapses used
components/mechanisms/modfiles .mod definition files for cells and synapses
tuning/current_clamp Directory of easy to use hoc-based testers/tuners to simulate current injection into cells in the model

Connectivity Matrix

  • + : Excitatory
  • - : Inhibitory
  • +/- : Both
Source/Target Pyr PV+ SOM+ CR+
Pyr + + + +
PV+ - - -
SOM+ -
CR+ - - -
Thalamic Input + + +
VP/SI Input - -

Note: current model has no connection between PV to SOM.

Tuning synaptic weights

For each connection, edit the mean and standard deviation of weights to adjust firing properties.

Source Target Synapse JSON
PN PN PN2PN_feng_min.json
PN PV PN2INT_feng_min.json
PN SOM PN2SOM_tyler.json
PN CR PN2CR_tyler.json
PV PN INT2PN_feng_min.json
PV PV INT2INT_feng_min.json
PV SOM INT2SOM_tyler.json
SOM PN SOM2PN_tyler.json
CR PN CR2PN_tyler.json
CR PV CR2INT_tyler.json
CR SOM CR2SOM_tyler.json
Thalamic PN BG2PNe_thalamus_min.json
Thalamic SOM BG2SOM_thalamus_min.json
Thalamic CR BG2CR_thalamus_min.json

Single Cell Profiling

Cell templates can be found in:

./components/templates/templates.hoc
./components/templates/SOM.hoc

Templates used are:

Cell_A
Cell_C
PVCell
SOM_Cell
CR_Cell

To tune individual cells, run

cd /tuning/current_clamp
nrngui Cell_A.hoc
nrngui Cell_C.hoc
nrngui PVCell.hoc
nrngui SOM_Cell.hoc
nrngui CR_Cell.hoc

To get fi curves for individual cells, run

bmtool util cell --template Cell_Af fi
bmtool util cell --template Cell_Cf fi
bmtool util cell --template InterneuronCellf fi
bmtool util cell --template SOM_Cell fi
bmtool util cell --template CR_Cell fi

Network Profiling

Plot Connection Totals

bmtool plot connection total

Plot the connections for a single cell

This is useful for verifying that your connection rules are functioning propertly, especially if they're space dependent

python plot_conns.py

BMTOOLS

Version 0.2.1+

Install BMTools by running

pip install bmtool

Help Examples:

bmtool --help
bmtool util --help
bmtool util cell --help
bmtool util cell fi --help

BMTK

Built using BMTK checkout 52fee and later. Parallel netcon recording was added 6/29/21.

git clone https://github.com/AllenInstitute/bmtk
cd bmtk
python setup.py develop

Appendix

Installing NEURON and BMTK

See https://github.com/tjbanks/easy-nrn-install

Tuning tutorial

To tune the model, several things must be in done in order.

  1. Tune individual cells
  2. Build the network
  3. Ensure connectivity is correct
  4. Run the network
  5. Analyze the network

We want connectivity and base firing rates to match literature.