Modeling mEPSC single channel noise in neurons

[s.sullivan]

I am interested in using Neuron software to model miniature excitatory post synaptic currents (mEPSCs) built from single channel kinetic schemes (perhaps in channel builder?). These mEPSCs would therefore contain noise fluctuations, allowing for estimates of single channel conductance using peak scaled non-stationary fluctuation analysis. I then want to be able to place these mEPSCs at different locations along the dendritic tree of a multi-compartmental neuron model to look at the effects of dendritic filtering on estimates of single channel conductance. Has this been done? Advice on related downloadable Neuron demos or modelling papers would be helpful.

[ted]

The Channel Builder is the most appropriate tool for implementing a ligand-gated channel mechanism with stochastic transitions between states. You'll probably need to pair that with an NMODL-implemented point process that uses deterministic ODEs to generate an appropriate time course of ligand ("neurotransmitter") when driven by an event. The two mechanisms would be coupled by POINTER variables so that the neurotransmitter concentration is communicated to the stochastic channel mechanism at machine language speeds. Don't worry about these last two sentences; I'm just including them now because the idea occurred to me, and this is the best place to record it for future reference.

These mEPSCs would . . . effects of dendritic filtering on estimates of single channel conductance.

I am not aware of anyone having done this, but I don't follow that particular neighborhood of computational/theoretical neuroscience literature closely.

Advice on related downloadable Neuron demos or modelling papers would be helpful.

Before getting too deeply involved in the intricacies of NEURON, first dig into the comp & theoretical neurosci literature (a likely source may be the Biophysical Journal) to find out what has already been done to investigate the basic scientific issue.

[Bill Connelly]

I answered Steve's question elsewhere, but I thought I'd paste in the same answer I gave there, in this forum, just incase anyone stumbles across it. I will hope that I'm not too terribly wrong.

Obviously this is possible. Like everything in NEURON, you can make it as simple, or as complex as you would like.

For a vague idea, neuron comes with demo.hoc, where you can insert a stochastic single channel HH process.

But if it were me, I'd be tempted to start with a very very simple model. I'd start by coding a point process that was just one single channel. I would open syn.mod and use that as a template. I would have a state parameter of P, that would be the open probability. When t > onset, set P to some value and have it decay via P = pmax * exp(-(t-onset)/tau). It should also have a state parameter O (I think it should be state), which is whether the channel is open or not.

You can generate a random number with scop_random() (uniform random between 0 and 1). If then random number is less than P, then the channel should open (O = 1). I would then say that the channel should be open for some period of time t (presumably selected again from a random process, but could also be fixed for your purposes I believe). Once a period of time greater than that had passed, the channel should close (O = 0) and you're back to square one.

Then your BREAKPOINT block simply is i = O*g*(v-e)

Obviously, this is incredibly simple, and has lots of places where it would fall down (like non fixed time step, and it is very far from efficient) but it will give you some idea, and should in principle work for your purposes. Then you can just insert as many single channels at any given place you want.

As I've stressed, this is very very simple. Have a look in model DB at Marteen Kole's Ih model ( 64195 ) for a more complex approach to single channel dynamics.