Importing NONMEM into rxode2

In nonmem2rx: Converts 'NONMEM' Models to 'rxode2'

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
library(rxode2)
setRxThreads(1L)
library(data.table)
setDTthreads(1L)

The goal of nonmem2rx is to convert a NONMEM control stream to rxode2 for easy clinical trial simulation in R.

Here is a quick example of a conversion:

library(nonmem2rx)

# First we need the location of the nonmem control stream Since we are running an example, we will use one of the built-in examples in `nonmem2rx`
ctlFile <- system.file("mods/cpt/runODE032.ctl", package="nonmem2rx")
# You can use a control stream or other file. With the development
# version of `babelmixr2`, you can simply point to the listing file

mod <- nonmem2rx(ctlFile, lst=".res", save=FALSE, determineError=FALSE)
mod

Setting up nonmem2rx for your model

Some common options that you may want to change when importing NONMEM control stream are:

• The default NONMEM output extension; By default it is .lst. You can set it to something else, like .res, using the following option: options(nonmem2rx.lst=".res").

• Turn on extended control stream support. You can turn it on by options(nonmem2rx.extended=TRUE)

You probably also want to change the name of parameters and compartments. The easiest way to name the parameters whatever you want is to pre-specify the names. For example:

mod <- nonmem2rx(system.file("mods/cpt/runODE032.ctl", package="nonmem2rx"), lst=".res", save=FALSE,
                 thetaNames=c("lcl", "lvc", "lq", "lvp", "prop.sd"),
                 etaNames=c("eta.cl", "eta.vc", "eta.q","eta.vp"),
                 cmtNames = c("central", "perip"))

mod

This checks the parameter names to make sure they are the same length as the input names, if they are not, the model will skip parameter renaming and keep the default translation names theta# and eta#.

As a note, sigma parameters are not currently renamed; So for the following model (which grabs the parameter automatically labels to generate variables), sigma is simply eps#.

mod <- nonmem2rx(system.file("Theopd.ctl", package="nonmem2rx"), save=FALSE)
mod

You can still rename however you wish, though, using model piping (rxRename() or dplyr::rename() would both work):

mod <- mod %>% rxRename(add.var=eps1)
mod

This model does not specify the residuals in a way that makes sense to nlmixr2. If you want, you can still convert the rxode2 model to a nlmixr2 fit.

Technical details about reading NONMEM to rxode2

The key files to import are the NONMEM control stream (or related file) and the NONMEM output (often with a .lst or .res extension).

The import process steps are below:

• Read in the nonmem control stream and convert the model to a rxode2 ui function.

• Try to determine an endpoint/residual specification in the model (if possible), and convert to a fully qualified ui model that can be used in nlmixr2 and rxode2. If it cannot be determined automatically, you can manually fix this and still convert to a nlmixr2 object (if the data/estimates are available of course).

• If available, nonmem2rx will read the final parameter estimates and update the model.

• The converter will read in the nonmem input dataset, and search for the output files with IPRED, PRED and the ETA values. The translated rxode2 model is run for the population parameters and the individual parameters. This will then compare the results between NONMEM and rxode2 to make sure the translation makes sense. This only works when nonmem2rx has access to the input data and the output with the IWRES, IPRED, PRED and the ETA values.

• Converts the upper case NONMEM variables to lower case (can be turned off with nonmem2rx(..., toLowerLhs=FALSE)))

• Replaces the NONMEM theta / eta names with the label-based names like an extended control stream (can be turned off with nonmem2rx(thetaNames=FALSE, etaNames=FALSE))

• Replaces the compartment names with the defined compartment names in the control stream (ie COMP=(compartmenName))

nonmem2rx documentation built on April 3, 2025, 11:05 p.m.