Baseline Calcium Level -- Isn't this cheating?

[mdemirtas]

Hello,

I am currently modifying a model, trying to get the calcium dynamics to work properly. (Unfortunately, calcium levels in many models are significantly higher than the experimental data. I have tried the Grunditz and Sterratt models so far.) I have noticed that in every model that I have seen so far, the author includes the baseline calcium concentration level (around 50-70nM) in the code. Isn't this simply cheating?

Without any stimulation (synaptic stimulation, current clamp etc.) the calcium concentration in the cell must stabilize around 50-70nM. BUT, this must happen simply because the calcium influx by NMDA receptors and calcium channels should be canceled out by the calcium efflux by the calcium pumps. NOT because the rate/strength of calcium pumps/channels/NMDAs depend directly on the desired equillibrium point. (that we have to read off a book and include in the code.)

Some examples:

Example 1:

Code: Select all

INITIAL {

	P  = (1-exp(-65*-0.0755))/(10*Area*14564*(50e-09-(2e-03*exp(-65*-0.0755))))*k	:converting conductance to permaebility 
}

Here, the baseline calcium level is included in the mod file of the NMDA receptor.

Example 2:

Code: Select all

DERIVATIVE state { 

	drive_channel =  - (10000) * ica / (2 * FARADAY * depth)
	if (drive_channel <= 0.) { drive_channel = 0.  }   : cannot pump inward 
         
	:ca' = drive_channel + (cainf-ca)/taur
        ca' = drive_channel/18 + (cainf -ca)/taur*7
	cai = ca
}

Here, the baseline calcium level is included in the calcium pump dynamics. (cainf).

Example 3:

Code: Select all

DERIVATIVE integrate {
	cai' = -ica/depth/F/2 * (1e7) + (cai0 - cai)/tau
}

Here, cai0 is the baseline calcium level. This is included in the mod file that directly controls the calcium concentration. This is even worse that the others, the calcium levels are adjusted by the hand of god so that after a while, calcium concentration will always go back to the baseline level.


I see that this method is used everywhere, is it impossible (or very difficult, or unnecessary) to create a model that will reach equillibrium without including the desired equillibrium level in the code? I believe that this would be a perfect way to test a model's calcium dynamics, if it reaches an equillibrium and if the calcium concentration level agrees with the experimental data; this would be a good reason to pop a champagne and celebrate.

Sorry for the long post. I can provide the full files of these examples, if needed.

Mehmet

[ted]

I have noticed that most models that involve sodium and potassium channels assume fixed values for E_Na and E_K. Isn't this cheating?

[mdemirtas]

Thank you for the fast response,

Firstly, I do not mean to be disrespectful to any of these authors. I have little experience on the field of computational neuroscience and I am definitely not in a position to judge anyone.

I understand that we cannot model everything perfectly. We have to take some short-cuts, make approximations etc. However, I am working specifically on the calcium dynamics in spines. Therefore, I wish to model the calcium concentration as precisely as possible. In short, I can "cheat" about anything except of the calcium components.

Since I have limited experience and knowledge, I have attempted modifying existing published models instead of creating my own. I have tried several models, all specifically about calcium dynamics, all published in recognized journals. Not one had calcium concentration levels that were even slightly close to the experimental values. As a matter of fact, one of them had calcium concentration on the order of tens of millimolars. I honestly cannot understand how such a model can be used or published.

Are there any models out there that you trust whole-heartedly about the calcium dynamics?

[ted]

I do not mean to be disrespectful to any of these authors. I have little experience on the field of computational neuroscience and I am definitely not in a position to judge anyone.

I wasn't defending or criticizing; just playing devil's advocate.

I am working specifically on the calcium dynamics in spines. Therefore, I wish to model the calcium concentration as precisely as possible. In short, I can "cheat" about anything except of the calcium components.

First it would be good to drop pejorative terms. Next it would be useful to recognize the difference between "functional representation" and "mechanistic representation" in models, and the key role of the modeler's intent in determining what combination of these two should be used in any particular model. All mechanistic models tend to become phenomenological or functional at the edges. Where those edges lie, and how much mechanistic detail is included, are matters of judgment. A level of detail and balance between mechanistic and functional representation that may be just right for some purposes could be overkill for another purpose, or inadequate for yet another.

Given your particular interests, you might want to look for models that focus on calcium dynamics; it's unlikely that you'll find much mechanistic representation in models for which calcium is a bit player. The topic is of wide interest, so the literature is full of pertinent conceptual models. Computational models will be fewer, but not necessarily harder to find. You might want to start by searching ModelDB for "calcium pump" or "calcium accumulation". You might also want to check out Biomodels DB. When you find something that looks promising, be sure to read the associated paper(s) to discover the authors' true intent and assumptions.