search

openworm / muscle_model

Model of C elegans body wall muscle based on Boyle & Cohen 2008
http://www.opensourcebrain.org/projects/muscle_model
Other
47 stars 24 forks source link

SciUnit test to reproduce additional muscle patch clamp data in the muscle model #64

Open slarson opened 7 years ago

slarson commented 7 years ago

Currently, a notebook that executes one of the muscle models being run out of c302 is here

We want to fit this to existing data.

Figures are here

Ideally we'd be running both the model and the comparison with data inside SciUnit.

slarson commented 7 years ago

@rgerkin Can we set up a time in a couple weeks to discuss this?

rgerkin commented 7 years ago

@slarson Yes, most mid-day times will work for me.

rgerkin commented 7 years ago

@slarson In progress.

rgerkin commented 6 years ago

Currently neuronunit is giving each processor its own copy of the root NML/LEMS file so that it may rewrite it as needed for certain tasks. This is important when simulations are done via jNeuroML since jNeuroML takes a path to a file which then represents the structure of the model which cannot be changed in memory.

c302 NML files contain local include directives to other c302 NML files in same directory. In order to copy the root c302 NML file elsewhere, I need jNeuroML to know to look in a list of include directories that includes the original root c302 NML file directory, so that it can find these include files. Padraig implemented this in NeuroML/jNeuroML@a714ffa23c018a65f97a0c1f8c812bd6b667cee8.

However, in order to use this development branch, I have to get it to compile, which is currently failing as described in NeuroML/jNeuroML#52.

When that issue is resolved I can continue making progress on this one.

rgerkin commented 6 years ago

The above issues are now solved and I have a prototype shown here. This just demonstrates wrapping the model with a SciUnit Model class and getting the c302 simulation to run. The next step is to actually do the testing, but that should be the easy part.

rgerkin commented 6 years ago

@slarson Is the data in the figure in the "Figures are here" link in your first comment on this issue available in digitized form anywhere? Panel C is easy enough to do by hand, but A, D, and E probably require the raw values or a digitization of the figure in order to test. Does muscle_model have its own version of channelworm-django?

rgerkin commented 6 years ago

@slarson Here I have the muscle model being tested (against a suite of two tests) in SciUnit. I can add more tests from the table in panel C of the figure later. Showing an overlaid comparison probably requires getting a digitized figure.