Matthew T. Sit (UC Berkeley)
Roy Ben-Shalom (UCSF, LBNL)
Kevin Bender (UCSF)
We can use a compartmental model to represent neurons. [1] Further, changes to ion channel distribution or density affect neuronal output. We ask, what are the distributions of ion channels at the different compartments of the neuron? [2]
We can use a genetic optimization algorithm to fit models to electrophysiological data. A stimulus and values for free parameters can be inputed to our model to obtain a voltage response representing the simulated waveform. We can compare this to the target experimental data using a score function to assess the similarity. This information can then be used to refine our free parameter values iteratively.
In order to constrain the free parameters, we employ the use of large volumes of synthetic data, including that of stimuli, parameter sets, and voltage responses. The current methodology is to store all this data in separate csv files, one for each datum instance. The goal is to abandon csv files and to leverage NWB instead.
All three goals have been accomplished as of the weekend following the Hackathon! The next steps are to integrate usage of NWB into our existing scripts. We have already integrated into our stimuli generating script. The next task is to integrate it into our scripts which use Neuron [3] so that it can accept stimuli and parameter sets via NWB, and write its voltage response outputs to NWB as well.
Not available at this time.
[1] Canavier & Landry. 2006.
[2] Mainen & Sejnowski. 1996.
[3] https://neuron.yale.edu/neuron/