Bartolucci-Passini-Severi model in tissue

Question

Hello everyone,

I was trying to use the BPS ventricular cell model in a 2D simulation of the “Basic tissue EP” example. 

The problem is that the model is particularly stiff, and with bench the simulation converges only for very small time steps (<2 us) or using cvode (max step of 0.3 ms). In either case a simple 2x2 cm tissue takes tens of minutes to solve 20 ms.

I also tried solving the 2 markovian models (for ICaL and IKr) separately with the markov_be method, but I didn’t get a benefit in the maximal step I could use for the integration with bench and so the simulations are still too slow... how can I fix this?

Here is the cellML of the model:

https://models.cellml.org/workspace/5fd/file/36c7253c0426d8fcd7326ec104e162b44b656174/BPS2020.cellml 

(with a mistake in Cao: the control value is 1.8 not 0.9), I'm also attaching how I imported it in EasyML:

https://opencarp.org/q2a/?qa=blob&qa_blobid=11054773068091813872

Thank you for the support,

Eugenio

Answer 1

Hi Eugenio,

Which ODEs are most stiff in your model? Some sodium gates?

One option could be to solve those with a higher order method, which could relax the requirements on the integration time step (a bit).

Do you have other options in mind? Otherwise, I think that one has to deal with the higher computational costs for more involved models.

Cheers,

Axel

Comments

Hi Axel,

thank you for the reply. Actually my question was not clear: I was not expecting the BPS model, which is a development of the Tomek one with only few state variables more (due to the markovian formulation for ICaL), to require much smaller time steps and much longer simulation times.

Now I constrained the gating variables between 0 and 1 and added lookup tables for Vm and Cai and the BPS model is integrated without cvode in only slightly longer times than the Tomek one, as expected.

Thank you,

Eugenio