search

Open-Systems-Pharmacology / MoBi

MoBi® is a software tool for multiscale physiological modeling and simulation
Other
29 stars 9 forks source link

Start values are set to NaN after Calculate Scale Divisor #194

Open UlrichSi opened 6 years ago

UlrichSi commented 6 years ago

When "Calculate Scale Divisor" is used after simulation, the Start values of the molecules appear to be set to NaN.

Use the (internal) example project provided in Mantis, then:

  1. Open the simulation.
  2. Press "F5" to simulate.
  3. Click on "Run&Analyze"/"Calculate Scale Divisor", then "Calculate", then "Ok".
  4. Expand the simulation's configuration in the simulations browser. The MSV building block is flagged as different. Right-click it, then select "Show differences".

The differences list not only the changed scale divisors, but also the start values. If these are committed, they will show up in the MSV BB. No current workaround, either divisors or start values have to be entered manually in MSV, since a "Refresh from source" will both override concentrations and divisors.

My observation is that this effect shows only in

StephanSchaller commented 6 years ago

What is the consequence for the simulation?

msevestre commented 6 years ago

Hum was not aware of this issue. Going to look at that and see if we can squeeze it in upcoming 7.3

msevestre commented 6 years ago

I can reproduce exactly and only with the observations defined above A concentration based project AND start value constant

@UlrichSi Well done.

msevestre commented 6 years ago

ok I'd like to discuss with @Yuri05 the actual implementation of Scale Divisir in case of concentration-based project. I feel that this might not be quite right but I might be mistaken

msevestre commented 6 years ago

All: In case of concentration project, shouldn't the scale factor be calculated for concentration outputs instead of amount?

UlrichSi commented 6 years ago

I guess the scale divisors should be used on the number that is calculated for a molecule container. If internally amount is used (even in a concentration-based project), then it should be calculated for amount. The intention is to avoid numerical underflow or overflow.

StephanSchaller commented 6 years ago

Yes, it depends how the models are handled internally / by the solver: do amount and conc. dependent projects look the same for the solver (and SFs are integrated in the same way for both) but are just displayed differently on the user interface?

msevestre commented 6 years ago

@all: Yeah ok. Then it's just a Mobi bug then :)

StephanSchaller commented 6 years ago

Yeah ok. Then it's just a Mobi bug then :)

Yes...HIGH PRIORITY!

msevestre commented 6 years ago

nope

StephanSchaller commented 6 years ago

MoBi is all the headstart we have on other PBPK software... don't start neglecting it!

msevestre commented 6 years ago

@StephanSchaller This is not high priority because this has been in the software for over 3 years and reported for the first time in November. Moreover this does not affect simulation results whatsoever.

Issues are analyzed and looked at. Their priority is defined according to their impact. This issue has very low impact and as such is not high priority.

We are doing our best with the resources we have to make MoBi better. Insinuating that we are neglecting MoBi when we thrive to always make it better and more user friendly is simply not acceptable.

I will now ask you one last time, to moderate your comments in the future

UlrichSi commented 6 years ago

I am sure this bug will be fixed. However, only very few people seem to use the Scale Divisor functionality - like the two of us, Stephan.

So, let's not get too emotional about this issue.

StephanSchaller commented 6 years ago

I will now ask you one last time, to moderate your comments in the future

Please excuse, no harm intended.