Documentation/justification missing for Cell._update_end_pts

[katduecker]

Disclaimer: If you are new contributor, please do not work on this issue at this time. If you would like to contribute, please find a different issue, and make sure that there is no Pull Request already open for that issue.

I have a question regarding #661 that I can't make sense of.
My understanding is that @raj1701 implemented a function called _update_end_pts in cell.py which allows us to access information on the section end points in python and not in NEURON, which helps with pickeling parallelization etc.

I do not understand why in cells_default.py we define end points that we don't end up using anyway.
E.g. the end points for the L2/3 pyramidal neuron in cells_default.py is

end_pts = {
'soma': [[-50, 0, 765], [-50, 0, 778]],
'apical_trunk': [[-50, 0, 778], [-50, 0, 813]],
'apical_oblique': [[-50, 0, 813], [-250, 0, 813]],
'apical_1': [[-50, 0, 813], [-50, 0, 993]],
'apical_tuft': [[-50, 0, 993], [-50, 0, 1133]],
'basal_1': [[-50, 0, 765], [-50, 0, 715]],
'basal_2': [[-50, 0, 715], [-156, 0, 609]],
'basal_3': [[-50, 0, 715], [56, 0, 609]],
}

but the actual end points (from net.cell_type['L2_pyramidal'].sections[section].end_pts) are:

end_pts = {
'soma': [[0, 0, 0], [0, 0, 22.1]],
'apical_trunk': [[0,0,22.099999999999987], [0,0,81.59999999999998]],
'apical_oblique': [[0,0,81.59999999999998], [-340, 0, 81.59999999999998]],
'apical_1': [[0, 0, 81.59999999999997], [0,0, 387.59999999999997]],
'apical_tuft': [[0, 0, 387.5999999999999], [0, 0, 625.5999999999999]],
'basal_1': [[0,0,0], [0,0,-85.0]],
'basal_2': [[2.842170943040401e-14, 0, -84.99999999999997], [-180.3122292025696, 0, -265.3122292025696]],
'basal_3': [[-2.842170943040401e-14, 0, -84.99999999999997], [180.3122292025696, 0, -265.312229202569]],
}

I think @rythorpe asked the same question in that PR too but I don't see a clear answer to it.

This does not seem to change the behavior of the cell as far as I can tell based on some clamping that I have done,
but it has caused confusion when developing new plots of the network cells.

Why define the "wrong" end points in cells_default.py to then overwrite them in private functions in cell.py?

@ntolley @jasmainak @rythorpe could you explain so that we can document this?

Thanks!

[rythorpe]

My understanding is that the end point coordinates defined in cell.py define a direction vector for each compartment that then gets scaled in magnitude later on in the code. Under the hood, I believe this scaling is done by nrn. I think @raj1701 created the Python function you mentioned to algorithmically mirror the functionality of nrn for plotting purposes, albeit with minor rounding differences (hence why we still let nrn do the scaling for the internal model that simulates neural dynamics instead of scaling and fixing the network in Python).

[katduecker]

Thanks @rythorpe! I still don't understand why the end points for soma are defined as [0,0,0],[0,0,23] instead of [0,0,0],[0,0,39], or if it's just the direction, why not [0,0,0],[0,0,1]? Why 23?

[rythorpe]

Thanks @rythorpe! I still don't understand why the end points for soma are defined as [0,0,0],[0,0,23] instead of [0,0,0],[0,0,39], or if it's just the direction, why not [0,0,0],[0,0,1]? Why 23?

Fair point. I'm guessing the model as it is now was repurposed from other code where the desired length of that segment was truly 23. You could probably open a PR to test and fix this :)

[katduecker]

I believe they are the values from Bush and Sejnowski 1993 ...

Well comparison suggests there are similar. There are some inconsistencies..