Increasing diffusion coefficient speeds up CPM

[Carlijn Lems]

Hi everyone,

I am working on a multiscale model that includes an ODE, CPM and PDE layer. When I increase the diffusion coefficient of the PDE layer without adjusting any other settings or model parameters, the CPM dynamics appear to speed up, i.e. the cells grow/move faster per simulation time unit. I have tried to correct for this by increasing the MCS duration, but this has no effect on the simulation outcomes. Adjusting the relative time scaling under Global > System > time-scaling also does not help. I run into the same problem with the example multiscale models, such as Autocrine Chemotaxis. Am I doing something wrong, or could this be a bug?

Thank you,

Carlijn

[Jörn Starruss]

Hi Carlijn,

sorry for the late answer.

We suspect that you just run into a feedback-coupling of your model. As you mentioned the Autocrine Chemotaxis example: The movement and aggregation of the cells in that model is bound to the chemotactic behaviour of the  cells. The faster the attractant moves, the faster the gradient can establish and cells can respond via aggregation. So that's key of the model.
In addition, faster diffusion will take longer to simulate, as the forward-euler diffusion solver we currently employ has a time step limit depending on the diffusion rate.

If you still feel there is something odd, please don't hesitate to contact us again, maybe even including an example.

Best, Jörn

Am 17.05.21 um 11:29 schrieb Carlijn Lems:

--
You received this message because you are subscribed to the Google Groups "Morpheus users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to morpheus-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/morpheus-users/126c27ca-a8aa-4a36-8890-77c7eba2b9f8n%40googlegroups.com.

-- 
--------------
Jörn Starruß
Tel +49 351 463 38554
APB 1026

Technische Universität Dresden
Zentrum für Informationsdienste und Hochleistungsrechnen (ZIH)
Abteilung Innovative Methoden des Computing

[Carlijn Lems]

Hi Jörn,

I understand that for the Autocrine Chemotaxis example, the movement and aggregation of the cells is dependent on the attractant, however in my own model this is not the case. The local field concentration is only used in CellType > System > DiffEqn to calculate the value of a property that is unrelated to movement/aggregation. Still, I experience the same problem. I also understand that increasing the diffusion coefficient decreases the diffusion time step (indicated with 'TimeStepListener "Diffusion" propagates its time step ...'). However, I noticed that if the diffusion time step is smaller than the CPM time step that is based on the MCS duration, then the diffusion time step is also used for the CPM, and the MCS duration that I include in the model is ignored. As a result, the movement of the cells increases. I assume that it should be possible to have a CPM time step that is higher than the diffusion time step, but please tell me if I'm wrong. To illustrate the problem, I included a simple example of a cell that increases in size independent of a field g, and with cell property A that is dependent on g. When the diffusion coefficient of g is set to 0, the cell is still increasing in size after 100 time steps. However, if the diffusion coefficient is set to 1, the cell reaches its maximum volume after ~28 time steps. 

I look forward to hearing from you,

Carlijn

Op woensdag 26 mei 2021 om 18:43:04 UTC+2 schreef Jörn Starruss:

[Jörn Starruss]

Hi Calijn,

Thanks to your example I could confirm the issue. We solved it along with a couple of other improvements.

Please find an updated release 2.2.3 at https://morpheus.gitlab.io

Best,
Joern

Am 02.06.21 um 10:09 schrieb Carlijn Lems:

To view this discussion on the web visit https://groups.google.com/d/msgid/morpheus-users/1c973ab3-5b78-4af4-a4cc-694d0e145539n%40googlegroups.com.