This implementation is not compatible with the current version of libNeuroML, it is currently being rewritten by Mike Vella.
See also #18
Compile the PyramidalSimulation object:
g++ -c PyramidalSimulation.cpp -I /usr/include/python2.7/ -I ./ -l python2.7 -o PyramidalSimulation.o
Invoke the linker:
g++ example.cpp PyramidalSimulation.o -I /usr/include/python2.7/ -I ./ -l python2.7 -o simulation
Denis Turutin described the simulations he sent us as follows:
-
For spontaneous 50s, It is spontaneous evoked (current clamp) with holding current 0 pA.
-
For CCramp it is, 0 pA for 100ms, -20 pA 5ms, from -20pA to 40 pA linear within 2 s, finally 0 pA for 100ms
We are using the CCramp data (2) (RATIONALE - it is very difficult to interpret the data obtained under the effect of synaptic input, the current injectino should negate synaptic current influx).
The experimental data recorded in W05042200_1_1_3_1.txt will be used as a target.
Some Calculations - If Denis' protocol description is correct, total time of execution is 0.1+0.05+2+0.1 = 2.205s. There are 11025 data points in the recording, giving a sampling frequency of 5KHz.
Observations:
- (initial) Resting potential 30mV
Note: the initial hyperpolarizing step can be used to elicit information about the passive properties of the cell.
Initial simulation:
The initial simulation will be of reduced complexity:
Take the component of the ramp from 10pA to 40pA and approximate to 25pA. As the current ramp is linear, 10pA should occur at (100+5+1000)ms = 1105, i.e data point 1.1055000=5525.To data point 2.1051000=10525
This is saved as redacted_data.txt
The relevant model is contained in the /pyramidal_implementation folder
This model includes the following currents: - k_fast - k_slow - Ca - leak
To run model install pyramidal and its dependencies as described here: http://pyramidal.readthedocs.org/en/latest/install.html
The relevant model is contained in the /pyramidal_implementation folder, cd to that folder and:
Then execute the command:
python run.py eval_file0 40.041444758152295 0.0 4514.250191560498 35.2 0.4089 0.6 --compile --plotoverlay
This will use the auto-mod file compilation feature of Pyramidal including nrnivmodl compilation and run the model with the best-optimized parameters to-date.
To run model install Optimal Neuron and its dependencies as described here: http://optimal-neuron.readthedocs.org/en/latest/installation.html
Remove any mod files (such as ca.mod) from the pyramidal_implementation repository as this will interfere with the optimizer behaviour, the optimizer is designed to run with the manual_*.mod files which are located in the /mod_file directory and need to be the only mod files present in the /pyramidal directory before execution of the optimization script.
The current optimization is set up to use features extracted from the data file, the feature values are:
'peak_linear_gradient': 0.0126455, 'average_minimum': 32.9139683819512, 'spike_frequency_adaptation': 0.054102950823597951, 'trough_phase_adaptation': -0.032339835206814785, 'mean_spike_frequency': 170.75638755391191, 'average_maximum': 52.484330488178259, 'trough_decay_exponent': 0.082997586003614746, 'interspike_time_covar': 0.67343012507213718, 'min_peak_no': 20, 'spike_width_adaptation': 5.196371093168479e-17, 'max_peak_no': 20, 'first_spike_time': 105.37999999997665, 'peak_decay_exponent': -0.074000673186574759
and corresponding weights:
'peak_linear_gradient': 20,'average_minimum': 5.0, 'spike_frequency_adaptation': 0.0, 'trough_phase_adaptation': 0.0, 'mean_spike_frequency': 1.0, 'average_maximum': 2.0, 'trough_decay_exponent': 0.0, 'interspike_time_covar': 0.0, 'min_peak_no': 1.0, 'spike_width_adaptation': 0.0, 'max_peak_no': 50.0, 'first_spike_time': 1.0, 'peak_decay_exponent': 0.0
Then run:
>>>python optimization.py
The optimizer will execute on 64 threads and display the results. Play with optimization.py to change #threads, #population #max_evaluations etc.