FH model provided in examples Q10 discrepancy

[JimH]

I just started working with Neuron and I had a question regarding the fh.mod file that is provided in the examples folder. I am working with external stimulation and thresholds, and thus an article of interest to me is one by Rattay and Aberham (1993,Modeling Axon Membranes for Functional Electrical Stimulation). This article points out that the FH model does not immediately appear to have a temperature dependence. This of course is not the case, as the k factor and Q10 are both parts of the model, as the fh.mod shows with the alpha and beta functions. What the paper then points out, which is of course confirmed by the source, a 1963 paper by B. Frankenhaeuser and Moore titled The effect of temperature on the sodium and potassium permeability changes in myelinated nerve fibres of Xenopus laevis, is that Q10 was not 3 for all of the alphas and betas functions in the FH model.

What Rattay and Aberham go on to show is that since the m variable is so different (1.8 alpham and 1.7 betam), is that the AP shape, current profiles, and strength-duration profiles are very different. With that said, and being fairly new to this model, is this difference typically ignored, as would be suggested by the provided fh.mod, or is that an error in the file?

[ted]

I just started working with Neuron and I had a question regarding the fh.mod file that is provided in the examples folder.

You mean the fh.mod that is included with the MSWin distribution of NEURON?

I am working with external stimulation and thresholds, and thus an article of interest to me is one by Rattay and Aberham (1993,Modeling Axon Membranes for Functional Electrical Stimulation). This article points out that the FH model does not immediately appear to have a temperature dependence. This of course is not the case, as the k factor and Q10 are both parts of the model, as the fh.mod shows with the alpha and beta functions.

I am not familiar with the paper you cite, but like most statements,
"the FH model does not immediately appear to have a temperature dependence"
is susceptible to a Clintonian diversity of interpretations. Does it simply mean that temp
sensitivity was customarily ignored in most--or even all--computational implementations
of the FH model?

What Rattay and Aberham go on to show is that since the m variable is so different (1.8 alpham and 1.7 betam)

m, the gating variable? 1.8 alpham and 1.7 betam? I don't follow you.

With that said, and being fairly new to this model, is this difference typically ignored, as would be suggested by the provided fh.mod, or is that an error in the file?

I have seen a lot of mod files that totally ignore temperature dependence, and many that
rigidly (and almost certainly incorrectly) assume that q10 is 3. I don't know what the
common practice is regarding the FH model; you will discover that from the literature.
Also, I don't know the provenance of fh.mod--whether it was an oversimplified bit of
"demo" code, or whether it is based on a published model. However, it wouldn't be hard
to revise it so that the rate constants had the appropriate q10s, whatever those may be.

[JimH]

Just to clarify a few points.

You mean the fh.mod that is included with the MSWin distribution of NEURON?

Yes, the fh.mod that is included with the MSWin distribution of NEURON

Does it simply mean that temp sensitivity was customarily ignored in most--or even all--computational implementations
of the FH model?

The classic 1964 FH paper does not include temperature dependence in any of the equations, although, as Rattay suggests, perhaps Frakenhaeuesr and Huxley assumed that the reader should know to put them in, which was done in fh.mod

m, the gating variable? 1.8 alpham and 1.7 betam? I don't follow you.

I apologize for not being clearer, I mean "m" the gating variable, which is a function of alpha and beta functions (HH analog), in the case of "m" I mean alpha and beta are alpham and betam.

However, it wouldn't be hard to revise it so that the rate constants had the appropriate q10s, whatever those may be.

Again, to clarify I mean k=Q10^((T-T0)/10) which in the HH model, is 3 for all state variables, multiplied. However, according to a 1963 paper (mentioned in previous post), Q10 varies for each alpha and beta function.

As you stated, it wouldn't be hard to revise the rate constants for the appropriate q10s, which I did. However, not that it is good practice, but it is easy to see people making the assumption that the example file is correct, and using it in their code without checking it.

I don't know what the
common practice is regarding the FH model; you will discover that from the literature.

As is often the case many papers are scarce on "small" details with something that is so "common." Usually a single reference is given for an approach that today is so well accepted. Especially when the model itself is not under question, but is simply a well established means or technique used to evaluate another question.

With that said, I was curious if people who use this model more often than I do are aware of this variable Q10 parameter in the FH model, and whether they choose to ignore it or not.

[ted]

As you stated, it wouldn't be hard to revise the rate constants for the appropriate q10s, which I did. However, not that it is good practice, but it is easy to see people making the assumption that the example file is correct, and using it in their code without checking it.

Several important but unrelated issues are pertinent. The first is pragmatic: it is always a
good idea to analyze a model before using it. Make sure you understand the
assumptions on which it was based. Verify that it was implemented properly.

The other issues are more philosophical but no less important.

What does it mean to say that a computational model is "correct"?

Any computational model is merely an instantiation, in a computer, of a conceptual model.
A conceptual model of a phenomenon or physical system is just a hypothesis about
some selected aspect of that phenomenon or system. The benefit that derives from
computational modeling is improved insight, i.e. a better understanding of the
consequences of the conceptual model. For a computational model to be correct, it must
be a faithful implementation of the conceptual model on which it is based--it must
correspond, point for point, with the assumptions that are contained in the conceptual
model. Nothing more, nothing less. Otherwise it will not be a reliable tool for gaining insight
into the conceptual model.

So if the conceptual model proposed by F & H ignored temperature effects, it would have
been incorrect for their computational model to include such effects. It would be a
misapplication of that model to try to use it, unaltered, to study temperature effects on
spike generation or propagation.

according to a 1963 paper (mentioned in previous post), Q10 varies for each alpha and beta function

Did that paper provide any confidence estimates for those values? Have there been any
subsequent reports that confirm them? From the standpoint of someone who is
interested in qualitative aspects of temperature effects, it might be sufficient to assume
that q10 is 3 for all temperature-sensitive parameters.