Using an analytical model

[vskantze]

Hi,

I'm trying to use nlmxir2 to fit response data and it works great with the ode model I provided. However, I get an error when I use an analytical solution to the model for faster convergence. I haven't found any help on this subject when providing your own analytical solution instead of the premade ones.

It seems as if RxODE tries to solve the ode system when there isn't one to be solved.

My code:

analytical_model_2 <- function() {
  ini({
    alpha <- log(0.01)
    A = log(100)
    w0 <- log(0.005)
    g0 <- log(100)
    add.err <- 0.1
    
    eta.alpha ~ 0.001
    eta.A ~ 0.001
    eta.w0 ~ 0.001
    eta.g0 ~ 0.001
    
  })
  model({
    alpha <- exp(alpha + eta.alpha)
    omega <- exp(w0 + eta.w0)
    A = exp(A + eta.A)
    g0 <- exp(g0 + eta.g0)
    

    
    t_since_dose = max(time - tDose, 0)
    
    gh = exp(-alpha * t_since_dose) *  A * sin(omega * t_since_dose) + g0
    gh ~  add(add.err)
  })
}

# Fit the model
 > fit <- nlmixr2(analytical_model_2, data = df_filter, est = "saem")#,
→ loading into symengine environment...
→ pruning branches (`if`/`else`) of saem model...
✔ done
→ finding duplicate expressions in saem model...
[====|====|====|====|====|====|====|====|====|====] 100%; 0:00:00 
✔ done
using C compiler: 'gcc.exe (GCC) 13.2.0'

ℹ calculate uninformed etas
Model:

param(alpha,A,w0,g0,eta.g0,tDose);
rx_pred_=.prd;
cmt(gh);
dvid(1);

Error in rxSolveSEXP(object, .ctl, .nms, .xtra, params, events, inits,  : 
  The following parameter(s) are required for solving: .prd
Error: The following parameter(s) are required for solving: .prd

[mattfidler]

Hi @vskantze if this is about nlmixr2, then I believe the version released yesterday will fix your issue.

Unfortunately without data I cannot really reproduce your issue, but the model seems to work.

the max function translation to symengine was refactored in rxode2 because symengine didn't like it. Now it uses if/else expressions instead.

library(nlmixr2)
#> Loading required package: nlmixr2data

analytical_model_2 <- function() {
  ini({
    alpha <- log(0.01)
    A = log(100)
    w0 <- log(0.005)
    g0 <- log(100)
    add.err <- 0.1

    eta.alpha ~ 0.001
    eta.A ~ 0.001
    eta.w0 ~ 0.001
    eta.g0 ~ 0.001

  })
  model({
    alpha <- exp(alpha + eta.alpha)
    omega <- exp(w0 + eta.w0)
    A = exp(A + eta.A)
    g0 <- exp(g0 + eta.g0)

    t_since_dose = max(time - tDose, 0)

    gh = exp(-alpha * t_since_dose) *  A * sin(omega * t_since_dose) + g0
    gh ~  add(add.err)
  })
}

analytical_model_2 <- analytical_model_2()

summary(rxode2::rxode2(analytical_model_2$saemModel))
#> → loading into symengine environment...
#> → pruning branches (`if`/`else`) of saem model...
#> ✔ done
#> → finding duplicate expressions in saem model...
#> [====|====|====|====|====|====|====|====|====|====] 0:00:00
#> → optimizing duplicate expressions in saem model...
#> [====|====|====|====|====|====|====|====|====|====] 0:00:00
#> ✔ done
#> rxode2 NA model named rx_890be4b4237babfecac11f024d27cdf3 model (✔ ready). 
#> DLL: /tmp/RtmpWHXVwU/rxode2/rx_890be4b4237babfecac11f024d27cdf3__.rxd/rx_890be4b4237babfecac11f024d27cdf3_.so
#> NULL
#> 
#> Calculated Variables:
#> [1] "rx_pred_"
#> ── rxode2 Model Syntax ──
#> rxode2({
#>     param(alpha, A, w0, g0, eta.g0, tDose)
#>     rx_expr_0 ~ t - tDose
#>     rx_expr_1 ~ rx_expr_0 > 0
#>     rx_pred_ = exp(A - exp(alpha) * (rx_expr_0) * (rx_expr_1)) * 
#>         sin(exp(w0) * (rx_expr_0) * (rx_expr_1)) + exp(eta.g0 + 
#>         g0)
#>     cmt(gh)
#>     dvid(1)
#> })

^{Created on 2024-09-20 with reprex v2.1.1}****

[mattfidler]

I am closing this, for now. I think I will close all issues here and archive this repository, since I think it is causing too much confusion and it hasn't been updated for years.