cardiacmodelling / fitting-notebooks Goto Github PK

Fast and slow inactivation --> Need an extra constraint to ensure the slow gate is slower than the fast gate

in the "Four ways of fitting"30666-6) pap

Probably in multiple places

Check they're all in \text{mV}

Voltage is tab:green, current is tab:blue

Could re-use wires example

I think it's a bit unfair to say we don't know the reason for log-transformation. We do know quite a bit why it helps...

• positively constrained to unconstrained
• skewed distributions to less skewed
• similar variance for all parameters

I think these are all good reasons to mention already!

For the AP example, might worth pointing it to https://arxiv.org/abs/2001.04230/ (figures 1-3). Make sure people are aware of the pitfalls when the kinetics of the ion channels is slightly incorrect.

For papers always use doi, always cite authors or title in text

Or maybe just shorten?

Or maybe even something like

## Title goes here [!image|view with notebook viewer] [!image|view on github](link) [!image|play with binder]

E.g. a section "Running multiple fits simultaneously on a big server"

Perhaps split into two: one looking at the results closely, a second looking at validation data sets

• Repeated runs (move out of starting points notebook) validates fitting.

  • Order runs by RMSE, should have cluster of good points near the top. Percentage = robustness
  • With same order plot |p_i - p_best_i| / |p_best_i|, should be very similar inside good cluster! See page 17 four ways supp
  • Plot results in the prior
  • Run MCMC around best point, looking for unidentifiability

• Training, validation, and test sets validates modelling: REF wires

In chemistry

a * k_1 is a rate
k_1 is a "rate constant"

even if it isn't constant

Have to check wording everwhere, come up with something good?

• Ideally use Myokit from pypi, with some minimum version if required
• Ideally use PINTS from pypi, but may need the repo version instead?

• Rename old notebook told transformations-old
• Create copy, add notice near top of new notebook linking to old, add notice near top of old linking to new
• Update the new notebook to use transformation objects
• Update the notebooks following up from it to use them too
• Update the shared / example code

And split into AP and CaT chapter

Base on current-models pdf

• energy
• alternative formulation
• figure 5
• Give, don't derive, equations for Vhalf and time constants

Will make intro chapter a bit shorter

"Fitting tutorial"

"Ion current fitting tutorial"

"Ion current and AP fitting tutorial" ?

"Ion current and AP fitting tutorial with Myokit and PINTS" ?

Just show how, say it's open area, speed-up would depend on machine used

Just link to e.g. Tsien

With two inactivation gates, show how to add constraint that "fast" time constant < "slow" time constant

(or other strategies, e.g. re-order points after fitting)

Should be https://doi.org/10.1113/JP275733

But instead points to editorial note

Re-use wires example?

Simplified or real?

Might well include some temperature dependence? Bottom line, no more (or be careful when using) Q10.

Update wording. See optimisation; first example