Error: the complex F is not supported by babelmixr2

[john-harrold]

I'm testing the models in the model library (nlmixr2lib) trying to convert them to different formats. I'm having trouble converting this one to monolix.

library(babelmixr2)
#> Loading required package: nlmixr2
#> Loading required package: nlmixr2data
library(rxode2)
#> rxode2 2.1.3.9000 using 5 threads (see ?getRxThreads)
#>   no cache: create with `rxCreateCache()`
rx_fun = 
function() {
    covariateData <- list(WT = "Body weight in kg")
    description <- "Dupilumab PK model (Kovalenko 2020)"
    reference <- "Kovalenko P, Davis JD, Li M, et al. Base and Covariate Population Pharmacokinetic Analyses of Dupilumab Using Phase 3 Data. Clinical Pharmacology in Drug Development. 2020;9(6):756-767. doi:10.1002/cpdd.780"
    units <- list(time = "day", dosing = "mg")
    ini({
        lvc <- 0.908258560176891
        label("central volume (L)")
        lke <- -2.92994453301599
        label("elimination rate (1/d)")
        lkcp <- -1.54646311327271
        label("central-to-peripheral rate (1/d)")
        Mpc <- 0.686
        label("ratio of kcp and kpc (kpc is peripheral to central rate with units of 1/d)")
        lka <- -1.36257783450257
        label("absorption rate (1/d)")
        lMTT <- -2.25379492882461
        label("mean transit time (d)")
        lVm <- 0.0676586484738149
        label("maximum target-mediated rate of elimination (mg/L/d)")
        Km <- fix(0.01)
        label("Michaelis-Menten constant (mg/L)")
        lfdepot <- -0.441610554744518
        label("Bioavailability (fraction)")
        e_wt_vc <- 0.711
        label("Exponent of weight on central volume (unitless)")
        CcpropSd <- c(0, 0.15)
        label("Proportional residual error (fraction)")
        CcaddSd <- fix(0, 0.03)
        label("Additive residual error (mg/L)")
        etalvc ~ 0.192
        etalke ~ 0.285
        etalka ~ 0.474
        etalvm ~ 0.236
        etamtt ~ 0.525
    })
    model({
        vc <- exp(lvc + etalvc) * (WT/75)^e_wt_vc
        ke <- exp(lke + etalke)
        kcp <- exp(lkcp)
        ka <- exp(lka + etalka)
        MTT <- exp(lMTT + etamtt)
        Vm <- exp(lVm + etalvm)
        kpc <- kcp/Mpc
        ktr <- (3 + 1)/MTT
        d/dt(depot) <- -ktr * depot
        d/dt(transit1) <- ktr * (depot - transit1)
        d/dt(transit2) <- ktr * (transit1 - transit2)
        d/dt(transit3) <- ktr * transit2 - ka * transit3
        d/dt(central) <- ka * transit3 - ke * central - kcp * 
            central + kpc * periph - central * (Vm/(Km + central/vc))
        d/dt(periph) <- kcp * central - kpc * periph
        f(depot) <- exp(lfdepot)
        Cc <- central/vc
        Cc ~ add(CcaddSd) + prop(CcpropSd)
    })
}

rx_obj = rxode2::rxode2(rx_fun)
export_name = "test"

library(tibble)
dataset = 
tribble(
~id,  ~low,  ~time,  ~high,  ~cmt,  ~amt,  ~rate,  ~ii,  ~addl,  ~evid,  ~ss,  ~dur,  ~WT,  ~dv,
  1,    NA,      0,     NA,  'Cc',    NA,     NA,   NA,     NA,      0,   NA,    NA,    1,    0,
  1,    NA,      1,     NA,  'Cc',    NA,     NA,   NA,     NA,      0,   NA,    NA,    1,    0,
  2,    NA,      0,     NA,  'Cc',    NA,     NA,   NA,     NA,      0,   NA,    NA,    1,    0,
  2,    NA,      1,     NA,  'Cc',    NA,     NA,   NA,     NA,      0,   NA,    NA,    1,    0)

res = nlmixr2::nlmixr2(rx_obj, dataset, "monolix", babelmixr2::monolixControl(modelName=export_name, runCommand=NA))
#> Error : the complex F is not supported by babelmixr2
#> Error: the complex F is not supported by babelmixr2

^{Created on 2024-08-20 with reprex v2.1.1}

[mattfidler]

That is because it isn't supported because f(depot) is an equation. More parsing could probably support this particular expression, though I am not likely to get to this soon.

[mattfidler]

The error could be a bit more clear, I suppose.

[john-harrold]

Is there a way I can detect it's presence in rx_obj above?

[mattfidler]

Nope.

[mattfidler]

At least not as the nlmixr2/rxode2 interface is currently written.

[john-harrold]

Can I add that to the props? I know this isn't the place to request it :)

[mattfidler]

That is where it would likely go, though I believe I wouldn't support this directly since it is too much of a edge case for properties, rather I would support what properties (aka initial condition, f, rate, dur, alag) are present in the model and for what state).

Then you could use modelExtract() to parse this out yourself. I would say that if this changes, it wouldn't work so this idea is fragile. I think you should base it on if the model translates instead. But that is just my thoughts right now.

There are likely cases where a model won't import from nonmem or export to nonmem, and the same is true for monolix.

[john-harrold]

I'm sorry, you would rather support f, but isn't this f?

[john-harrold]

Oh I see. Can you add what properties are supported? I can easily use that.

[mattfidler]

I will add what properties are defined by the model.

For example in this model you would have some way of figuring out that f(depot) is in the model.

[john-harrold]

Awesome. Thank you.

[john-harrold]

If I wanted to conditionally skip tests for Monolix conversion where f is causing failure would I look at the Property column from the props$cmtProp data frame for a value of f?

$cmtProp
  Compartment Property
1       depot        f

[mattfidler]

No value of f is in the data frame. You would have to query it from the UI.

You can probably get that information from modelExtract(). The data.frame simply states there is a property there (not the value of it).

See

https://nlmixr2.github.io/rxode2/articles/Modifying-Models.html#creating-more-complex-model-modification-functions

[mattfidler]

I do not plan to add this to the data.frame either since it is can be queried elsewhere...

[john-harrold]

If I wanted to conditionally skip tests for Monolix conversion where f is causing failure would I look at the Property column from the props$cmtProp data frame for a value of f?

$cmtProp
  Compartment Property
1       depot        f

This is what is in the props$cmtProp field.

[mattfidler]

Yes but it doesn't have the value of f(depot), just that f(depot) is in the model in some form. If you want to get the value you would have to query it.

[mattfidler]

If you want to skip all monolix models with f, then yes this would work, though some monolix models with F work just fine.

[john-harrold]

I just need a way to detect models that will fail during testing. Can you tell me what babelmixr2 is looking at to throw the error?

[mattfidler]

It looks for a f that depends on derived variables as opposed to an f that depends on only initial estimates.

[mattfidler]

As said by the bug report, this example is likely a bug.

[john-harrold]

I'm sorry. I missed that.

…

On Sun, Sep 29, 2024 at 3:45 PM Matthew Fidler ***@***.***> wrote: As said by the bug report, this example is likely a bug. — Reply to this email directly, view it on GitHub <#115 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAYF5VCPAZGZVUTRSLWPR6DZZB7HJAVCNFSM6AAAAABM26UOLWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGOBRGYZTQNZYHE> . You are receiving this because you modified the open/close state.Message ID: ***@***.***>

[mattfidler]

It's not obvious. It's linked to this issue in babelmixr2 issues...

[mattfidler]

#116