Add Lang_PLOSComputBiol2024

[paulflang]

Is the readme up to date that you still welcome additional benchmark models? If so, I'd like to contribute this.

A few points to mention on this PR:

• The following works locally:

>>> import benchmark_models_petab as models
>>> problem = models.get_problem('Lang_PLOSComputBiol2024')
>>> assert problem.measurement_df is not None

• petablint complains about using observables in the noiseFormula. saCeSS (and I believe parPE) did not have a problem with that. This can be changed but will be less concise.

Missing parameter(s) in the model or the parameters table: {'obs_cycA__cyt_median', 'obs_p21__cyt_median', 'obs_cycE__nuc_median', 'obs_pRB__cyt_median', 'obs_p21__nuc_median', 'obs_cycA__nuc_median', 'obs_E2F1__cyt_median', 'obs_Skp2__nuc_median', 'obs_Skp2__cyt_median', 'obs_cycB1__cyt_median', 'obs_p27__cyt_median', 'obs_cycB1__nuc_median', 'obs_E2F1__nuc_median', 'obs_cycE__cyt_median', 'obs_p27__nuc_median', 'obs_pRB__nuc_median'}   
Not OK

• The other models contain observable and noise parameter overrides in the measurement table, which is not needed in my case, but can be changed, too, of course.

• The SBML file contains the optimized parameters. I've never tried an optimization with this file, but I don't see a reason why it shouldn't work 🤞

• More generally, please lmk what else I need to do.

[dweindl]

Thanks Paul. Sorry for the late response. Contributions are still very welcome!

petablint complains about using observables in the noiseFormula. saCeSS (and I believe parPE) did not have a problem with that. This can be changed but will be less concise.

Right, it's supported by amici/parpe, but unfortunately not in PEtab v1 (PEtab-dev/PEtab#543). Will be supported in the next PEtab version. I guess for now it would be better to substitute the observable IDs.

[paulflang]

Thanks @dweindl , I pushed the substitution.

[m-philipps]

Hi Paul, thank you for the contribution, it is very welcome!

I'm checking the model based on this checklist and already added the simulated data and updated the visualisation specification to match Figure 7 in your publication. Is that fine with you?

It would be great if you could provide some information on the points below in the corresponding issue #210

• A brief model description (one or two sentences)
• A brief data description (one or two sentences)
• Differences between the implementation and the original publication are described
• Experience of fitting / uncertainty analysis (e.g. optimizer used, hyperparameters, reproducibility of best fit)
• Source of nominal parameters (e.g.: taken from the original publication, or from your own fitting)

and if you could check this point:

• The model ID and model name attributes in the SBML model file match the problem name (example)

Thank you already!

[m-philipps]

Thank you, this is ready to go!

[dweindl]

Hi @paulflang, the following parameters are set log-scale, but have 0 as nominalValue, is this correct?

kDpApc_1
kDpE2f1
kPhC25A
kPhNup
kDpNup_1
kIm_2
kEx_2

Three out of those are marked as estimated. Shouldn't they be different from 0?

[paulflang]

Hmm. So the last four are set to zero and not estimated (I wanted to switch nuclear pore phosphorylation and unspecific nuclear import and export off).

For the first three, I always used zero as my guess (Note to myself: kDpApc_1 and kDpE2f1 are B55 independent background dephosphorylation rates. kPhC25A is CDC25 phosphorylation by CCNA. It is known that CCNB:CDK1 can phosphorylate CDC25, so I thought I let the optimizer figure out if CCNA:CDK1 can do it too). I just checked how David ran the optimization, and he just set the lower bound to -1e24 (in log space) and the optimal result for all three parameters was around -9.98e23. Bit strange that the optimizer did not move much from the lower bound in any of the cases.

Anyway, since the PEtab problem passes petablint, I think it is up to the optimizer how to handle zero lowerBound for log parameter scale. But I could also set lowerBound and nominalValue to a very low value, maybe 1e-100). @dweindl , what do you think?

[dweindl]

Makes sense. Thanks for the explanation, @paulflang.

since the PEtab problem passes petablint

There is the open question whether it should :).