Reducing morphology (soma) complexity, how?

[selfdestructo]

Hi!

I'd just want to know if there exist a simple way of reducing morphology complexity using NEURON, either while converting from .swc- to .hoc-format, or after loading the .hoc-file I just created (redefining things). The reason is that for some morphologies I convert from neuromorpho.org, I end up with soma's with multiple compartments, i.e. soma[1],...,soma[n], all with nseg=1, while I really want just one compartment (soma[1], nseg=1).

The reason is that I work with extracellular potential modeling from reconstructed neurons, and the thin cylindrical slabs that we get for multicompartment soma's, do not go well with the line-source approximation of Holt&Koch for the soma compartments. I need some thing that I can approach with a sphere or line. (or simple cylinder for that matter)

[ted]

I work with extracellular potential modeling from reconstructed neurons, and the thin cylindrical slabs that we get for multicompartment soma's, do not go well with the line-source approximation of Holt&Koch for the soma compartments.

Really? Why?

for some morphologies I convert from neuromorpho.org, I end up with soma's with multiple compartments, i.e. soma[1],...,soma[n], all with nseg=1

Yes, because that's how the morphometric measurements were made. It is easy to manually revise the pt3d data that describe a series of soma slices so that the result is a single section with complex geometry, but that doesn't convert a complex shape into a cylinder or sphere.

I need some thing that I can approach with a sphere or line. (or simple cylinder for that matter)

Real somas often have complex structures that are not well captured by a single sphere or cylinder. This is particularly the case for cells that are located in noncortical structures, and especially cells at the periphery of brainstem nuclei.

I'd just want to know if there exist a simple way of reducing morphology complexity using NEURON

Sorry, nothing like that is built into NEURON. If you find an existing solution, or devise one yourself, it might be possible to implement a tool for NEURON in hoc or Python that will do the job.

[selfdestructo]

I work with extracellular potential modeling from reconstructed neurons, and the thin cylindrical slabs that we get for multicompartment soma's, do not go well with the line-source approximation of Holt&Koch for the soma compartments.

Really? Why?

It's just that the line-source method assumes that the length of the line-source should be long relative to the diameter of the piece of "dendrite" it is supposed to represent. But it's a while since I read those papers, so I might be mistaken of course.

In any case, it's most convenient for us with our model framework that we approximate the (single) soma compartments of neurons as sphere's, since that is what is already implemented, and that the potential in extracellular space from a sphere source is the same as for a point source, i.e. Phi(r) = I_mem/(4*pi*sigma_ex*r), r > r_soma. I don't believe that the outcome should be wildly different anyway.

[ted]

the line-source method assumes that the length of the line-source should be long relative to the diameter of the piece of "dendrite" it is supposed to represent.

Sounds like a rational assumption for something called "line-source method".

it's most convenient for us with our model framework that we approximate the (single) soma compartments of neurons as spheres, since that is what is already implemented, and that the potential in extracellular space from a sphere source is the same as for a point source, i.e. Phi(r) = I_mem/(4*pi*sigma_ex*r), r > r_soma.

Whether or not this introduces significant error remains an open question that requires verification on a case-by-case basis. The shape of the soma is only one factor, and is an issue only when it falls between the extremes of "potato" (longest diameter no more than 2-3 x shortest diameter, which can probably be approximated by a sphere) and "string bean" (very long compared to diameter, which is suitable for the line-source approximation). The other factor is how much contribution somatic membrane current makes to the field at different points in space. Most cell membrane, and most membrane current, is nonsomatic, so for subthreshold fluctuations at least the soma would probably have little effect on the field (except in the "near vicinity" of the soma). But for a cell with sparse dendrites, and a medium-thick soma whose centroid has an irregular curvature that deviates widely from a straight line over the course of several tens of microns in 3-space, it seems likely that approximation by a sphere would introduce significant errors into field calculations at distances on the order of the largest somatic dimension.

I don't believe that the outcome should be wildly different anyway.

Differences of opinion are what horse races (and experiments) are about.