To enable fitting a model to experimental data resulting from applying the sinusoidal waveform.
Use cases:
i) to reproduce Kylie's paper/Michael's Method 4 (DONE in the hacky prototype)
ii) to fit slightly different data (Dom's mutant) with same model and same fitting spec (SHOULD WORK ALREADY in the hacky prototype, not tested)
iii) to fit with Chon's staircase protocol instead (should already work, not tested)
iv) to fit any HH/Markov ion channel model to the same sinusoidal data with reaction rates of the correct form with the same fitting spec (NEEDS WORK). We'll want to tag
i) 'A' parameters,
ii) 'B' parameters, and
iii) forward/backward/voltage-independent rates 'k', of the form k_forward=Aexp(BV), k_backward=Aexp(-BV), k_voltage_independent=A, where A and B are positive constants. And then system should automatically figure out what rates (and therefore what parameters) are needed for a given current's kinetics.
We've confirmed that application of the Kylie fitting spec (protocol, optimiser, with log-linear transforms, uniform prior in that space etc.) all works for this in https://github.com/sanmitraghosh/hergModels although bigger models do need more runs to get global optimum reliably, as you'd expect.
To enable fitting a model to experimental data resulting from applying the sinusoidal waveform.
Use cases: i) to reproduce Kylie's paper/Michael's Method 4 (DONE in the hacky prototype) ii) to fit slightly different data (Dom's mutant) with same model and same fitting spec (SHOULD WORK ALREADY in the hacky prototype, not tested) iii) to fit with Chon's staircase protocol instead (should already work, not tested) iv) to fit any HH/Markov ion channel model to the same sinusoidal data with reaction rates of the correct form with the same fitting spec (NEEDS WORK). We'll want to tag
We've confirmed that application of the Kylie fitting spec (protocol, optimiser, with log-linear transforms, uniform prior in that space etc.) all works for this in https://github.com/sanmitraghosh/hergModels although bigger models do need more runs to get global optimum reliably, as you'd expect.