Feature request (maybe?): allow nlmixrSim to work on models that have not been solved

[billdenney]

I'm trying to generate the dataset for a reprex for a different issue, and for that, I'm trying to simulate a dataset. When trying to simulate the dataset, I thought that I would try using nlmixrSim(), but that doesn't seem to work. I think that the right answer is: "Bill, you're supposed to use RxODE for simulations like that." But, it would be handy if I could just use nlmixrSim() since I think it has most of the machinery built-in.

Here is the reprex of what I want to do. Ideally, either of the paths to get to simulation at the bottom would work:

library(nlmixr)

linear_twocmt_iiv_growth <- function() {
  ini({
    tcl <- fixed(log(0.01)) ; label("Clearance (L/kg/hr)")
    tq <- fixed(log(0.05)) ; label("Intercompartmental clearance (L/kg/hr)")
    tv <- fixed(log(1.5)) ; label("Volume central (L/kg)")
    tv1 <- fixed(log(2)) ; label("Volume peripheral (L/kg)")
    
    tktr <- log(c(0.00001, 0.03, 3)) ; label("Transit compartment rate (1/hr)")
    slope <- c(-2, -0.02, 2) ; label("Slope on growth rate ((ng/mL)/hr)")
    ltumorgrowth <- log(c(0.0001, 0.01, 1)) ; label("Baseline tumor exponential growth rate (1/hr)")

    iiv_tumorgrowth ~ 0.1 ; label("Inter-individual variability in baseline tumor exponential growth rate")

    prop_sd <- 0.2 ; label("Proportional error for tumor volume (fraction)")
    add_sd <- 100 ; label("Additive error for tumor volume (mm^3)")
  })
  drake::no_deps(model({
    cl <- exp(tcl)
    q <- exp(tq)
    v <- exp(tv)
    v1 <- exp(tv1)
    kcp <- q/v
    kpc <- q/v1
    ktr <- exp(tktr)
    TUMOR(0) <- TUMORBL
    
    d/dt(CENTRAL) = -cl/v * CENTRAL - kcp*CENTRAL + kpc*PERIPH1
    d/dt(PERIPH1) = kcp*CENTRAL - kpc*PERIPH1
    # unit conversion
    cp <- CENTRAL/v*1000
    d/dt(TRANSIT) = ktr*(cp - TRANSIT)
    tumorgrowth <- exp(ltumorgrowth + iiv_tumorgrowth)
    ktumor <- tumorgrowth + slope*TRANSIT
    d/dt(TUMOR) = ktumor*TUMOR
    TUMOR ~ add(add_sd) + prop(prop_sd)
  }))
}

sim_dose <-
  data.frame(
    ID=1:50,
    time=0,
    amt=rep(seq(0, 20, by=5), each=10),
    evid=1,
    cmt="CENTRAL",
    mdv=1,
    TUMORBL=100
  )

sim_obs <-
  expand.grid(
    ID=1:50,
    time=(0:60)*24,
    amt=0,
    evid=0,
    cmt="TUMOR",
    mdv=0,
    TUMORBL=100
  )

sim_data <- rbind(sim_dose, sim_obs)
sim_data <- sim_data[order(sim_data$ID, sim_data$time), ]

simulation <- nlmixrSim(linear_twocmt_iiv_growth, events=sim_data)
#> Error: object of type 'closure' is not subsettable
model <- nlmixr(linear_twocmt_iiv_growth)
simulation <- nlmixrSim(model, events=sim_data)
#> Error: object of type 'closure' is not subsettable

^{Created on 2021-07-26 by the reprex package (v2.0.0)}

[mattfidler]

I am moving the UI to RxODE to allow the UI functions to be solved with RxODE alone.

So I agree and am already working on this in RxODE.