SomaRestingVmTest

Validate Resting membrane voltage from soma of Purkinje cell

class cerebunit.validation_tests.cells.Purkinje.test_for_soma_restVm.SomaRestingVmTest(*args: Any, **kwargs: Any)

This test compares the measured resting Vm observed in real animal (in-vitro or in-vivo, depending on the data) generated from neuron against those by the model.

The test class has three levels of mechanisms.

Level-1 validate_observation()

Given that the experimental/observed data has the following: mean, SD (or SE), sample_size, units, and raw_data, validate_observation() checks for them. The method then checks the data condition by asking NecessaryForHTMeans. Depending on the data condition the appropriate score_type is assigned and corresponding necessary parameter; for t-Test, the parameter observation["standard_error"] and for sign-Test, the parameter observation["median"].

Level-2 generate_prediction()

The model is executed to get the model prediction. The prediction is a the resting Vm from the soma of a PurkinjeCell returned as a quantities.Quantity object.

Level-3 compute_score()

The prediction made by the model is then used as the null value for the compatible score_type based on the data condition (normality and skewness) determined by validate_observation(). The level ends by returning the compatible test-statistic (t or z-statistic) as a score.

How to use:

from cerebunit.validation_tests.cells.Purkinje import SomaRestingVmTest
data = json.load(open("/home/main-dev/cerebdata/expdata/cells/PurkinjeCell/Llinas_Sugimori_1980_soma_restVm.json"))
test = SomaRestingVmTest( data )
s = test.judge(chosenmodel, deep_error=True)

Then to get the test score s.score and test report call print(s.description). If one is interested in getting the computed statistics call s.statistics.

Further notes on the test.

• The experimental observation data (as json file) must have the element protocol_parameters, which in turn has the nests the elements temperature and initial_resting_Vm.

• One should consider whether the model is compared against in vitro or in vivo experimental data (in addition to the species under study). For example,

  • Consider the Llinas and Sugimori (1980, 10.1113/jphysiol.1980.sp013357) experimental data (Llinas_Sugimori_1980_soma_restVm.json)

  • The reported experimental data only includes those with initial resting levels for \(\geq -50 mV\) discarding those for \(< -50 mV\).

  • The observed resting potential are claimed by the authors to be more negative than those observed in vivo.

  • The authors infer that this could be due to in vitro which is done on slices. The slicing removes background synaptic input generated by parallel fibre synapses.

  • For more details see Llinas_Sugimori_1980_soma_restVm.json

compute_score(observation, prediction, verbose=False)

This function like generate_pediction is called automatically by sciunit which RestingVmTest is a child of. This function must be named compute_score The prediction processed from “vm_soma” is compared against the experimental_data to get the binary score; 0 if the prediction correspond with experiment, else 1.

generate_prediction(model, verbose=False)

Generates resting Vm from soma. The function is automatically called by sciunit.Test which this test is a child of. Therefore as part of sciunit generate_prediction is mandatory.

validate_observation(observation, first_try=True)

This function is called automatically by sciunit and clones it into self.observation This checks if the experimental_data is of some desired form or magnitude. Not exactly this function but a version of this is already performed by the ValidationTestLibrary.get_validation_test