RuntimeError when attempting to run an already setup cell

[lungsi]

Arbor Version: 0.4.1.dev0

1. RuntimeError Type 1

Running my cable cell cell_model.run(100) returns a RuntimeError

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-41323ba3c1b4> in <module>
----> 1 cell_model.run(100) # 100ms

RuntimeError: no mechanism passive in catalogue

The mechanisms and their decorations on the cell are as follows

# Set mechanisms
leaky_chnl = arbor.mechanism('passive/el=-54.3', {'g': 0.0003})
hh_na_k_chnl = arbor.mechanism("hh", {"ena": 50., "gnabar": 0.120,
                                      "ek": -77., "gkbar": 0.036,
                                      "gl": 0.0003})
# Attach mechanism
cell.paint('"soma"', hh_na_k_chnl)
cell.paint('"axon"', hh_na_k_chnl)
cell.paint('"dend"', hh_na_k_chnl)
cell.paint('"everywhere"', leaky_chnl)

I get the error (RuntimeError above) only when I try to run the model: I don't get any error during the setup.

2. RuntimeError Type 2

If try to run the same model but without cell.paint('"everywhere"', leaky_chnl) I get

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-41323ba3c1b4> in <module>
----> 1 cell_model.run(100) # 100ms

RuntimeError: mechanism hh has no parameter ena

3. RuntimeError Type 3

If try to run the same model but without cell.paint('"soma"', hh_na_k_chnl); cell.paint('"axon"', hh_na_k_chnl); cell.paint('"dend"', hh_na_k_chnl) I get

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-41323ba3c1b4> in <module>
----> 1 cell_model.run(100) # 100ms

RuntimeError: no mechanism passive in catalogue

How can I fix the cause for all the three errors?

[noraabiakar]

Hi,

The passive mechanism is called pas in arbor's default mechanism catalogue. If you use pas that should fix the first and third errors.
As for the second error: in arbor's version of the hh mechanism, ena and ek are not an assigned parameters. They refer to the reversal potentials of the sodium channel and potassium channel respectively; and you can paint the initial reversal potentials of ions (as well as their initial concentrations) directly on the cell, similar to how you painted mechanisms:

# Set mechanisms
leaky_chnl = arbor.mechanism('pas/e=-54.3', {'g': 0.0003})
hh_na_k_chnl = arbor.mechanism("hh", {"gl": 0.0003})

# Attach mechanism and paint ion parameters
cell.paint('"soma"', "na", rev_pot=50)
cell.paint('"soma"', "k", rev_pot=-77)
cell.paint('"soma"', hh_na_k_chnl)

cell.paint('"axon"', "na", rev_pot=50)
cell.paint('"axon"', "k", rev_pot=-77)
cell.paint('"axon"', hh_na_k_chnl)

cell.paint('"dend"', "na", rev_pot=50)
cell.paint('"dend"', "k", rev_pot=-77)
cell.paint('"dend"', hh_na_k_chnl)

cell.paint('"everywhere"', leaky_chnl)

You can also join regions to reduce the lines of code:

leaky_chnl = arbor.mechanism('pas/el=-54.3', {'g': 0.0003})
hh_na_k_chnl = arbor.mechanism("hh", {"gl": 0.0003})

# Attach mechanism and paint ion parameters
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "na", rev_pot=50)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "k", rev_pot=-77)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', hh_na_k_chnl)

cell.paint('"everywhere"', leaky_chnl)

Note that pas has no el parameter in arbor, it is called e instead.

Hope this fixes your issues.

[lungsi]

Thank you so much @noraabiakar (setting the mechanism has certainly changed from the previous version)

BTW In the previous version I had cell.paint("everywhere", leaky_chnl) how is cell.paint('"everywhere"', leaky_chnl) different?

PS: With the latest version only cell_model.probe("voltage", '"stim"', 10000) works not cell_model.probe("voltage", "stim", 10000).

[noraabiakar]

We've been working on the user facing interface in the past few months. One of the things we've added in v0.4 is the ability to have arbitrary strings as labels for regions/locsets. For example, you can rename your everywhere region in the label dictionary to full cell or soma & dend & axon and that would work, but because there's now spaces and special characters involved, we need to add the quotes to signify that this is a region/locset label as opposed to a region/locset expression.
As an example of why the extra quotes are needed we can look at 2 different methods to paint a pas mechanism every where on the cell:

cell.paint('(join (region "soma") (region "dend") (region "axon") (region "apic"))', 'pas')

or

# label dictionary defines the "full cell" region
cell.paint('"full cell"', 'pas')

If we don't add extra quotes around "full cell" region in the second method, it could be interpreted as a region s-expression/definition. We add the quotes to signify this is instead a region label. It's the same explanation for why we need quotes around stim.

The API changed slightly between releases as we try to settle on an easy interface for the users. We're also working on keeping our documentation up-to-date, the python examples are a good place to start to see what has changed since the last release.

[lungsi]

I am afraid this was not the solution. I still get a similar RuntimeError

Starting with the (modified) suggested solution

# Set mechanisms
leaky_chnl = arbor.mechanism('passive/e=-54.3', {'g': 0.0003})
hh_na_k_chnl = arbor.mechanism("hh", {"gnabar": 0.120,
                                      "gkbar": 0.036,
                                      "gl": 0.}) # to avoid duplication

# Attach mechanism
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "na", rev_pot=50)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "k", rev_pot=-77)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', hh_na_k_chnl)
cell.paint('"everywhere"', leaky_chnl)

and then cell_model.run(100) returns

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-41323ba3c1b4> in <module>
----> 1 cell_model.run(100) # 100ms

RuntimeError: no mechanism passive in catalogue

Also, if I used leaky_chnl = arbor.mechanism("pass", {'e': -54.3, 'g': 0.0003}) the result is similar

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-41323ba3c1b4> in <module>
----> 1 cell_model.run(100) # 100ms

RuntimeError: no mechanism pass in catalogue

And exactly the same (above) if I use leaky_chnl = arbor.mechanism('pass/e=-54.3', {'g': 0.0003})

[noraabiakar]

Oh that's my bad, I forgot to change the passive/pas mechanism in the suggested code, here's what that should look like:

# Set mechanisms
leaky_chnl = arbor.mechanism('pas/e=-54.3', {'g': 0.0003})
hh_na_k_chnl = arbor.mechanism("hh", {"gnabar": 0.120,
                                      "gkbar": 0.036,
                                      "gl": 0.}) # to avoid duplication

# Attach mechanism
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "na", rev_pot=50)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "k", rev_pot=-77)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', hh_na_k_chnl)
cell.paint('"everywhere"', leaky_chnl)

I will edit the above answer to use pas to help anyone else who may read this discussion.

[lungsi]

Ok Thanks @noraabiakar But I get another RuntimeError, this time the error message says mechanism pas/e has no parameter called e

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-41323ba3c1b4> in <module>
----> 1 cell_model.run(100) # 100ms

RuntimeError: mechanism pas/e=-54.3 has no parameter e

PS: Although you already mentioned above that the pas mechanism does not have the el parameter I just wanted to give it a try with that and I got a similar RuntimeError, RuntimeError: mechanism pas/el=-54.3 has no parameter el

[lungsi]

Ok I think I got it to work with

# Set mechanisms
leaky_chnl = arbor.mechanism("pas", {'e': -54.3, 'g': 0.0003})
hh_na_k_chnl = arbor.mechanism("hh", {"gnabar": 0.120,
                                      "gkbar": 0.036,
                                      "gl": 0.}) # to avoid duplication

# Attach mechanism
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "na", rev_pot=50)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', "k", rev_pot=-77)
cell.paint('(join (region "soma") (region "dend") (region "axon"))', hh_na_k_chnl)
cell.paint('"everywhere"', leaky_chnl)

and then cell_model.run(100) is successful (i.e, no RuntimeError) but

I don't know if the simulation runs with the intended changes.

Note:

• leaky_chnl = arbor.mechanism('pas/e=-54.3', {'g': 0.0003}) results in RuntimeError

[noraabiakar]

Arbor's pas.mod mechanism has the following header:

NEURON {
    SUFFIX pas
    NONSPECIFIC_CURRENT i
    RANGE g, e
}

The g and e parameters are RANGE parameters in this definition, meaning that every instance of this mechanism has its own g and e values which are controlled/provided by the user, making them inputs to the mechanism. If g and e had been GLOBAL parameters instead, they wouldn't be something to be provided by the user, but more of variables that control the behavior of the mechanism, which makes them a property of the mechanism itself.

In practice, the distinction between RANGE and GLOBAL is often murky in NMODL. But the classification is important in determining whether we need to use arbor.mechanism("pas/e=-54.3") or arbor.mechanism("pas", {'e': -54.3}).

The first method arbor.mechanism("pas/e=-54.3") is used when e is a GLOBAL parameter: something that controls the behavior of pas. Changing e changes the intended definition of the mechanism, so we need to derive a new mechanism similar to pas but with different GLOBAL parameters, the resulting mechanism is a new one called pas/e=-54.3.

The second method arbor.mechanism("pas", {'e': -54.3}) is used when e is a RANGE parameter: something that is meant to be set by the user. Changing e doesn't change the mechanism itself, it's like passing arguments to a function. We don't derive a new mechanism in this case, we use the same pas mechanism with different inputs.

So your final solution is indeed correct, and the simulation should run with the intended changes. I hope this explains the runtime errors you have seen. We also have a page in our docs about mechanisms, which may provide more insight.