pharmpy / modelbuilder Goto Github PK

It shouldn't be possible to set any number smaller than 1 for the number of cutoffs

I think the font should be a fixed width font in the text box displaying the model code. This will make the code look better, but also the mathematical rendering.

For example after starting the app: "Model Type" and "Model format" are labels and "Write model" and "Choose dataset" are buttons, but they look exactly the same (perhaps the font is slightly different). I think it should be possible to know what is clickable just by looking.

These are not important right now

If I try to change the name of "POP_CL" in the starting model I get a crash and the name of the parameter is not changed

When running the app the tab in the browser will be constantly toggling between being named "Pharmpy GUI" and "Updating...". I am using Firefox. In chrome it doesn't switch back and forth but is stable on "Pharmpy GUI"

Perhaps to some pop-up of some kind

Currently it says "PLACEHOLDER Pharmpy GUI Navbar", but since we don't have a navbar yet perhaps it should simply say "Pharmpy model builder"

I see that the file __pycache__/designfile.cpython-311.pyc is in the repo. This is a binary compiled file and shouldn't be commited.

Write a complete user guide, along with examples to help new users

If I increase up to say 3 peripherals it becomes impossible to go back down to 1 or 0. The number changes, but the model code stays the same

We should move the "Model statements" button. If possible the functionality could be replacing the generic code view, but it seems as if the text renderer is buggy. If not possible simply remove the button.

There is not MX-button selected in the route at start, but it seems as if IV is actually the default. Perhaps it should be selected

Zero order elimination is currently missing

Running pip install -r requirements.txt causes error for pywin32on linux due to it being windows specific

Directly switching to MM elimination after starting the app causes a crash.

I think this is actually a bug in Pharmpy:

Traceback (most recent call last):
  File "/home/rikard/devel/modelbuilder/app.py", line 174, in update_abs_elim
    "MM": set_michaelis_menten_elimination(globals()["model"]),
  File "/home/rikard/devel/modelbuilder_venv/lib/python3.10/site-packages/pharmpy/modeling/odes.py", line 573, in set_michaelis_menten_elimination
    model = _do_michaelis_menten_elimination(model)
  File "/home/rikard/devel/modelbuilder_venv/lib/python3.10/site-packages/pharmpy/modeling/odes.py", line 645, in _do_michaelis_menten_elimination
    km_init, clmm_init = _get_mm_inits(model, numer, combined)
  File "/home/rikard/devel/modelbuilder_venv/lib/python3.10/site-packages/pharmpy/modeling/odes.py", line 753, in _get_mm_inits
    dv_max = get_observations(model).max()
  File "/home/rikard/devel/modelbuilder_venv/lib/python3.10/site-packages/pharmpy/modeling/data.py", line 257, in get_observations
    idcol = model.datainfo.id_column.name
  File "/home/rikard/devel/modelbuilder_venv/lib/python3.10/site-packages/pharmpy/model/datainfo.py", line 738, in id_column
    return self.typeix['id'][0]
  File "/home/rikard/devel/modelbuilder_venv/lib/python3.10/site-packages/pharmpy/model/datainfo.py", line 1008, in __getitem__
    raise IndexError(f"No columns of type {i} available")
IndexError: No columns of type id available

No the UI says "Additative" error model. It should be "Additive"

Hi!
Looked around a little and noticed that uploaded datasets are required to be in the repo folder and will give an error message otherwise. Could be nice to allow other paths as well

Currently "PK" is the only "Model type" that can be selected so keep the radio button with only one selection, but "Example type" should be removed.

It feels like the width of the numeric input boxes is too wide in comparison to the rest of the buttons in the same column and also given the small amount of text they will contain. Perhaps they could be made less wide

After creating a parameter it will is added to the parameter list and a THETA value is added to the nonmem code. But when switching to another tab the parameter is vanishing.

Currently, only statements and ETA abstraction are visible

Would be nice to see the structure of the model in a screen, either in a box below the model_code or as a separate option of the model_code.

Alignment of title differ between computers

When setting the model type to oral and then selecting first order absorption (which the model already has), an extra MAT parameter statement will be added.

It seems as if the parameters object gets converted to have different values when going to the parameters tab.

When editing values in the parameter table, it does not update accordingly

Perhaps this should be part of the Pharmpy documentation instead.

Would be nice to have specific links per page with specific informaiton

Perhaps "model view" instead

If stepping up to some transits and down again I cannot change the model code to be 0 transits again.

The app gives an error when directly switching to the "Parameter Variability" tab

Traceback (most recent call last):
  File "/home/rikard/devel/modelbuilder/app.py", line 440, in render_iiv_iov_data
    occ_opts = globals()["model"].datainfo.typeix["idv"].names + globals()["model"].datainfo.typeix["unknown"].names
  File "/home/rikard/devel/modelbuilder_venv/lib/python3.10/site-packages/pharmpy/model/datainfo.py", line 1008, in __getitem__
    raise IndexError(f"No columns of type {i} available")
IndexError: No columns of type idv available

During handling of the above exception, another exception occurred:

IndexError: No columns of type unknown available

When changing from generic model format to nonmem model format after having changed the iiv of some parameters the code is not nonmem code anymore:

The nonmem code looks as follows:
{"parameters": {"parameters": [{"name": "POP_CL", "init": 0.01, "lower": 0, "upper": null, "fix": false}, {"name": "POP_VC", "init": 1, "lower": 0, "upper": null, "fix": false}, {"name": "IIV_CL", "init": 0.1, "lower": null, "upper": null, "fix": false}, {"name": "IIV_VC", "init": 0.1, "lower": null, "upper": null, "fix": false}, {"name": "IIV_CL_IIV_CL", "init": 0.01, "lower": -Infinity, "upper": Infinity, "fix": false}, {"name": "sigma", "init": 0.09, "lower": -Infinity, "upper": Infinity, "fix": false}]}, "random_variables": {"dists": [{"class": "JointNormalDistribution", "names": ["ETA_CL", "ETA_VC"], "level": "IIV", "mean": "MutableDenseMatrix([[Integer(0)], [Integer(0)]])", "variance": "MutableDenseMatrix([[Symbol('IIV_CL'), Symbol('IIV_CL_IIV_CL')], [Symbol('IIV_CL_IIV_CL'), Symbol('IIV_VC')]])"}, {"class": "NormalDistribution", "name": "epsilon_p", "level": "RUV", "mean": "Integer(0)", "variance": "Symbol('sigma')"}], "eta_levels": {"levels": [{"name": "IIV", "reference": true, "group": "ID"}, {"name": "IOV", "reference": false, "group": "OCC"}]}, "epsilon_levels": {"levels": [{"name": "RUV", "reference": true, "group": null}]}}, "statements": {"statements": [{"class": "Assignment", "symbol": "Symbol('CL')", "expression": "Mul(Symbol('POP_CL'), exp(Symbol('ETA_CL')))"}, {"class": "Assignment", "symbol": "Symbol('VC')", "expression": "Mul(Symbol('POP_VC'), exp(Symbol('ETA_VC')))"}, {"class": "CompartmentalSystem", "compartments": [{"class": "Output"}, {"class": "Compartment", "name": "CENTRAL", "amount": "Symbol('A_CENTRAL')", "dose": {"class": "Bolus", "amount": "Symbol('AMT')", "admid": 1}, "input": "Integer(0)", "lag_time": "Integer(0)", "bioavailability": "Integer(1)"}], "rates": [[1, 0, "Mul(Symbol('CL'), Pow(Symbol('VC'), Integer(-1)))"]], "t": "Symbol('t')"}, {"class": "Assignment", "symbol": "Symbol('IPRED')", "expression": "Mul(Symbol('A_CENTRAL'), Pow(Symbol('VC'), Integer(-1)))"}, {"class": "Assignment", "symbol": "Symbol('IPREDADJ')", "expression": "Piecewise(ExprCondPair(Float('2.2250000000000002e-16', precision=53), Equality(Symbol('IPRED'), Integer(0))), ExprCondPair(Symbol('IPRED'), true))"}, {"class": "Assignment", "symbol": "Symbol('IPREDADJ1')", "expression": "Piecewise(ExprCondPair(Float('2.2250000000000002e-16', precision=53), Equality(Symbol('IPRED'), Integer(0))), ExprCondPair(Symbol('IPRED'), true))"}, {"class": "Assignment", "symbol": "Symbol('Y')", "expression": "Add(Symbol('IPRED'), Mul(Symbol('IPREDADJ1'), Symbol('epsilon_p')))"}]}, "estimation_steps": {"steps": [{"method": "FOCE", "interaction": true, "cov": null, "evaluation": false, "maximum_evaluations": 99999, "laplace": false, "isample": null, "niter": null, "auto": null, "keep_every_nth_iter": null, "solver": null, "solver_rtol": null, "solver_atol": null, "tool_options": {}}]}, "datainfo": {"columns": [{"name": "ID", "type": "id", "scale": "nominal", "continuous": false, "categories": null, "unit": "1", "datatype": "int32", "drop": false, "descriptor": null}, {"name": "TIME", "type": "idv", "scale": "ratio", "continuous": true, "categories": null, "unit": "1", "datatype": "float64", "drop": false, "descriptor": null}, {"name": "DV", "type": "dv", "scale": "ratio", "continuous": true, "categories": null, "unit": "1", "datatype": "float64", "drop": false, "descriptor": null}, {"name": "AMT", "type": "dose", "scale": "ratio", "continuous": true, "categories": null, "unit": "1", "datatype": "float64", "drop": false, "descriptor": null}, {"name": "DVID", "type": "unknown", "scale": "ratio", "continuous": true, "categories": null, "unit": "1", "datatype": "float64", "drop": false, "descriptor": null}], "path": null, "separator": "\\s+"}, "value_type": "PREDICTION", "dependent_variables": {"Y": 1}, "observation_transformation": {"Symbol('Y')": "Symbol('Y')"}, "initial_individual_estimates": null, "__magic__": "Pharmpy Model", "__version__": "0.99.0"}

In the same way we can generate e.g. NONMEM code, there should be a tab for R/Python code generation.

• Include a folder called /models that is the default path unless otherwise specified.

• Fix issue with write not naming models correctly

Move callbacks to separate files corresponding to each tab, then import from there into app.py
Will improve maintaining , debugging and readability.

Perhpas simple black

Perhaps the numeric input field could always be visible and have 0 as default. Then we don't need the toggle and hide functionality.

Perhaps to only "Model Builder"

Currently this selector is confusing. The "editable" option seems too low level. The "dataset path" has some strange default value.