Page 1 of 1

How much detail do we need in our models?

Posted: Thu Nov 05, 2009 11:46 am
by Keivan
I decided to start this topic and include every piece of information I find in this regard. It may help other people after me. To start I refer you the one nice article I find. (Excuse me If I have my own referencing method - this method seems to be enough for at least 100 years which is enough for me)

Code: Select all

Schutter05 [Biophysically detailed modelling of microcircuits and beyond]
= {author: Schutter, year: 2005, Title: Biophysically detailed modelling of microcircuits and beyond]

I don't know Erik De Schutter but It seems he concerns for details in modeling.
also his book seems to be interesting for people who concern about this question.

Code: Select all

Computational Modeling Methods for Neuroscientists
+ Lately we had a little discussion about detailed modeling of synapses
please share your experience with us.

Active Spines are stochastic in nature

Posted: Thu Nov 05, 2009 1:54 pm
by Keivan
on the head of synaptic spine, there may be only a few dozen channels and so it becomes questionable whether to treat the fraction of open channels as a continuous variable. Rather, we may have to treat each species or channel as a random individual. The randomness arises because the opening and closing of individual channels is typically a stochastic process.
In some situations the choice is clear: a typical synaptic spine has a volume of about 0.1 femtoliters. If the resting concentration of calcium is 100 nM, the spine contains about six free Ca2+ ions, which is bound to be stochastic. The point at which it becomes acceptable to use ODEs depends both on the number of molecules and on the reaction kinetics. In practice, any biological reaction involving fewer than 100 molecules is likely to require stochastic methods to describe its behavior.
Adopted from "Computational Modeling Methods for Neuroscientists"

Extracellular resistance is Not negligible.

Posted: Sun Nov 08, 2009 11:41 am
by Keivan
Extracellular resistance is usually assumed to be negligible in models. While this is clearly not true (because extracellular recordings could not be made otherwise and ephaptic coupling has been observed among tightly packed neurons; Jefferys, 1995), it is a reasonable assumption in most situations when the extracellular space is large (e.g., see Rall, 1959), the neuron is isolated, or the concern of the model is a single neuron.
Adopted from "Computational Modeling Methods for Neuroscientists"

When we can use Equivalent-Cylinder models?

Posted: Sun Nov 08, 2009 1:26 pm
by Keivan
The equivalent-cylinder model is a useful construct for studying the effect of current input or a voltage clamp applied at the soma on voltage changes throughout the dendritic tree. Its simplicity allows mathematical analysis and ready insight into dendritic function. However, it is less useful for studying dendritic input. Any input to a location in the equivalent cylinder would have to be divided among all dendrites in the fully branched structure at the same electrotonic distance for the responses to be the same. The equivalent-cylinder model is also useful for estimating parameter values. Clearly no dendritic tree is likely to satisfy all of the constraints of the equivalent-cylinder model, particularly the constraint of all dendrites terminating at the same electrotonic distance from the soma.
Adopted from "Computational Modeling Methods for Neuroscientists"
Also, we should have evidence that the target neuron can be simplified this way structurally.

Benchmarcking Neuron models

Posted: Thu Dec 10, 2009 3:29 pm
by Keivan
Gerstner09 [How Good Are Neuron Models]
See this new science journal article. To be honest, I didn't know that integrate and fire neuron model have the potential to be a good approximation of a real neuron even better than compartmental ones.