Target NeuroML instead of individual simulation engines?

[sanjayankur31]

Hi @orrenravid1 , I ran into your repo from the numpy issue on pyopengl where you had posted. This looks great.

I'm one of the current editors of the NeuroML standard and software ecosystem. The idea of NeuroML is to be a simulator independent language that researchers can use to build their models---the NeuroML stack then supports a number of "simulation backends" that users can then use for simulating their models. We support a bunch of these backends already and we can add more as the need arises in the field. ( I see that you have an example visualisation of the c elegans too: NeuroML is also used for modelling it in the c302 (part of the OpenWorm) project,. @pgleeson is the project owner of both NeuroML and c302 )

The NeuroML software stack is quite mature and we provide lots of tools/utilities in addition to the API for working with models. The documentation has examples on all of this, and so does our recent paper. This also includes a 3D visualiser, also using Vispy.

We do not, at the moment, have a graphical tool to work with NeuroML models, though. Hence my interest in your repo here :)

Open Source Brain v1 is great for working with existing models, but not for creating new ones from scratch. The closest tool to create new NeuroML models currently is perhaps NetPyNE-UI which can read and write NeuroML, but doesn't support it natively. There's also NeuroMLlite, but the main goal there is not to build a GUI tool for NeuroML.

So, we were wondering what you thought of targeting NeuroML instead of individual simulators for your tool here. The advantage of targeting the standard is that you will always work with a well defined specification and API, and you will not need to support the various simulators individually---the NeuroML core tools (pyNeuroML, jNeuroML) will do that for you. You will also be able to re-use a lot of the utilities that we've already implemented in the many libraries.

Please let us know what you think, also very happy to set up a meeting to have a chat if you'd like.

Cheers,

[orrenravid1]

@sanjayankur31 Sounds like an exciting prospect! I actually got my swc files for the c elegans example from c302 and I've played around a bit with the NeuroML things from there. I'll shoot you an email.

[sanjayankur31]

Sounds great. We have some info on morphology files. SWC files aren't always reliable for modelling, unfortunately:

https://docs.neuroml.org/Userdocs/ImportingMorphologyFiles.html
https://docs.neuroml.org/Userdocs/Software/Tools/SWC.html#userdocs-swc

Here's our SWC to NML export which tries to handle different variations in swc files:

https://github.com/NeuroML/pyNeuroML/blob/development/pyneuroml/swc/ExportNML.py#L177

pyneuroml has the pynml-plotmorph -i .. tool (with the -i flag for "interactive") that visualises cells/networks. It creates the mesh and uses it in vispy, but the mesh can be saved and loaded into any other tool/library. pyneuroml has a bunch of other utilities for plotting data too---time series, spikes, and just general plots.

Looking forward to hearing from you, cheers.