Non-uniform channel densities

[bgmeulem]

Hi

I'm working on writing an exporter for our own codebase to SONATA. We have many biophysically detailed neuron models with channel densities that vary along the dendrites. We embed these in networks for exact synapse placement, and synapse activations are saved as dataframes.

I thought it would make sense as a first step to export the neuron models. However, I can't figure out how to specify channel distributions of varying density.
the SONATA spec sheet is rather brief on this. One example shows a neuron model in both XML and JSON format.
I'd like to commit to the Allen cell types JSON format (https://github.com/AllenInstitute/sonata/blob/master/docs/SONATA_DEVELOPER_GUIDE.md#allen-cell-types-database-json), but i can't deduce from this documentation how to do so. hte docs on this page simply state "It is worth noting that the scope of the current work was with neuron models which had uniform ion channel distributions. For future versions of the common format, an example with non-uniform channel distributions should also be considered."

My questions:

1. Are non-uniform channel densities supported by SONATA or the file formats included in SONATA?
2. Where can I find a more complete specification of the Allen JSON format, if it supports non-uniform channel densities?

[jdcourcol]

Hello @bgmeulem,
indeed there are many way to specify biophysical detailed models in the spec: neuroML, Allen json or hoc file.

You can find here an example with non uniform channel models for hoc. ( you will need to login with a github or google account).

We are making available many biophysical detailed sonata models on the https://www.openbraininstitute.org platform (based on the work of Blue Brain Project at EPFL).

We would be happy to integrate your circuit on our platform and make it available and simulatable there. Please contact me if you are interested.

EDIT: fixing link.

[kaeldai]

Hi @bgmeulem,

SONATA is somewhat intentionally vague on things like ion channel placement because it is more focused on circuit description and tries to leave cell-specific implementation up to already established formats. As Jean-Denis said above you can use NeuroML2 specification to specify individual cells or cell-types with variable channel placement. You can also try another format if you want or define your own properties.

As for the Allen Cell-Types JSON format, unfortunately there is no complete specification. Nor does it allow non-uniform distribution of ion channels (the closest you can do is assign different values have apical and basal branches). (As a side note: the developers at DendroTweaks have extend the Allen JSON format to allow for advanced cell descriptions. It's a new tool but does look promising).

Please let us know if you have any more questions? If you want to talk more in-depth us feel free to reach out over eamil (kaeld at alleninstitute.org and antona at alleninstitute.org).