Convert an nlmixr2 model to a NONMEM control stream or Monolix project

These write the files that nlmixr2(model, data, est="nonmem") or nlmixr2(model, data, est="monolix") would write, without running NONMEM or Monolix. The file names are the same as the ones from nonmemControl(modelName=) and monolixControl(modelName=):

Usage

as.nonmem(
  x,
  data = NULL,
  modelName = NULL,
  timeVarying = character(0),
  dose = NULL,
  doseType = "bolus",
  ss = FALSE,
  cens = FALSE,
  limit = FALSE,
  control = nonmemControl(),
  ...
)

as.monolix(
  x,
  data = NULL,
  modelName = NULL,
  timeVarying = character(0),
  dose = NULL,
  doseType = "bolus",
  ss = FALSE,
  cens = FALSE,
  limit = FALSE,
  control = monolixControl(),
  ...
)

Arguments

x: nlmixr2 model function, rxode2 ui or fit
data: optional data to translate with. When NULL and x is a fit (and none of the template arguments are given), the fit's data are used; otherwise a template dataset is made from the arguments below.
modelName: the base name of the files. When NULL, the control's modelName is used, then the name of the model variable.
timeVarying: character vector of the covariates that change with time (by default none do). Time-varying covariates are not mu-referenced (and are Monolix regressors).
dose: the compartments receiving doses (names or numbers); by default the first compartment of the model (a model without compartments has no doses). Numbers are in the translated order, where a linCmt() model's compartments come before its ODEs
doseType: the type of dose for each dose compartment (recycled): "bolus", "infusion", "modelRate" (a modeled rate()) or "modelDur" (a modeled dur())
ss: logical; does the data have steady state doses (adds SS and II)
cens: logical; does the data have censored observations (adds CENS)
limit: logical; does the data have censoring limits (adds LIMIT, and implies cens=TRUE)
control: the nonmemControl() or monolixControl() options for the translation; runCommand is ignored since the program is not run
...: other arguments (currently ignored)

Value

invisibly, a named character vector of the written files

Details

NONMEM: <modelName>-nonmem/<modelName>.nmctl and <modelName>-nonmem/<modelName>.csv
Monolix: <modelName>-monolix.mlxtran, <modelName>-monolix.txt and <modelName>-monolix.csv

Part of the translation depends on the data: which covariates are time-varying (these are not mu-referenced), which compartments are dosed and how (the Monolix administrations, RATE), and if the data has steady state (SS/II) or censoring (CENS/LIMIT) columns. When data is given, all of these come from the data. Otherwise a one-subject template dataset is made from the model and the timeVarying, dose, doseType, ss, cens and limit arguments. This template dataset is written as the .csv file so you can see the columns the model expects; replace it with your data before running NONMEM or Monolix. With the template dataset, covariate expressions (like log(WT/70)) are written in the model instead of as derived data columns (muRefCovAlg=FALSE), so the files only need the model's covariates.

When files with the same name but a different model already exist, the new files are numbered (like <modelName>-001-nonmem), as in nlmixr2(); files of the same model are left alone.

Author

Matthew L. Fidler

Examples

# \donttest{
one.cmt <- function() {
  ini({
    tka <- 0.45
    tcl <- log(c(0, 2.7, 100))
    tv <- 3.45
    cl.wt <- 0.1
    eta.ka ~ 0.6
    eta.cl ~ 0.3
    eta.v ~ 0.1
    add.sd <- 0.7
  })
  model({
    ka <- exp(tka + eta.ka)
    cl <- exp(tcl + eta.cl + cl.wt * WT)
    v <- exp(tv + eta.v)
    linCmt() ~ add(add.sd)
  })
}

withr::with_tempdir({
  print(as.nonmem(one.cmt))
  print(as.monolix(one.cmt, timeVarying = "WT"))
})
#>  
#>  
#> ℹ parameter labels from comments are typically ignored in non-interactive mode
#> ℹ Need to run with the source intact to parse comments
#>  
#>  
#> ℹ translating linCmt() to NONMEM's closed-form solution
#> → loading into symengine environment...
#> → pruning branches (`if`/`else`) of full model...
#> ✔ done
#> ℹ writing nonmem files
#> ℹ done
#> ℹ only exported NONMEM control stream/data
#>                          nmctl                            csv 
#> "one.cmt-nonmem/one.cmt.nmctl"   "one.cmt-nonmem/one.cmt.csv" 
#>  
#>  
#> ℹ parameter labels from comments are typically ignored in non-interactive mode
#> ℹ Need to run with the source intact to parse comments
#>  
#>  
#> ℹ translating linCmt() to Monolix's closed-form solution
#> ℹ writing monolix files
#> ℹ done
#> ℹ run Monolix manually and rerun nlmixr()
#>                   mlxtran                     model                       csv 
#> "one.cmt-monolix.mlxtran"     "one.cmt-monolix.txt"     "one.cmt-monolix.csv" 
# }