Skip to contents

Getting started

Source: vignettes/PKPDsim.Rmd
PKPDsim.Rmd

Installation

PKPDsim can be installed from CRAN:

install.packages("PKPDsim")

Or you can install the development version from GitHub:

devtools::install_github("InsightRX/PKPDsim")

First simulation

The main simulation function in PKPDsim is sim(). To be able to simulate a dosing regimen for a specific drug, at least the following three arguments are required:

  • ode: the compiled ODE model (created using the new_ode_model() function)
  • parameters: a list of parameter values for the model
  • regimen: the dosing regimen (created using the new_regimen() function)

The model library in PKPDsim contains a small library of built-in PK and PD models, but of course more interesting is its ability to handle user-specified ODE systems. However, as a first example, let’s implement the most simple example from the library:

## Error in get(paste0(generic, ".", class), envir = get_method_env()) : 
##   object 'type_sum.accel' not found
p <- list(CL = 1, V  = 10, KA = 0.5)
pk1 <- new_ode_model("pk_1cmt_oral")
r1 <- new_regimen(
  amt = 100,
  n = 5,
  interval = 12
)
dat <- sim(
  ode = pk1,
  parameters = p,
  regimen = r1
)

You probably noticed that the new_ode_model()-step took a few seconds to finish, while the simulation itself was in the order of milliseconds. In new_ode_model(), the model is compiled from C++ source code to binary code, which takes a few seconds. However, this has to be done only once. After compilation, the model is then available to be used in sim() for as long as the R session is open.

Let’s look at the output. PKPDsim will output data in the “long” format, i.e. one row per observed timepoint, and split by compartment:

head(dat)
##     id t comp         y obs_type
## 1    1 0    1 100.00000        1
## 4    1 1    1  60.65307        1
## 40   1 2    1  36.78794        1
## 76   1 3    1  22.31302        1
## 112  1 4    1  13.53353        1
## 148  1 5    1   8.20850        1

To check what output was produced by sim(), let’s plot it (installation of ggplot2 required).

ggplot(dat, aes(x = t, y = y)) +
  geom_line() +
  facet_wrap(vars(comp))

Custom ODE model

We can write the same exact model using custom ODE code. The following example will also only output the concentration (only_obs = TRUE), and not the amounts in each compartment. We’re also going to use an infusion instead of bolus injection:

pk2 <- new_ode_model(
  code = "dAdt[1] = -(CL/V) * A[1]",
  obs = list(cmt = 1, scale = "V"),
  dose = list(cmt = 1)
)
r2 <- new_regimen(
  amt = 100,
  n = 5,
  interval = 12,
  type = "infusion",
  t_inf = 2
)
dat2 <- sim(
  ode = pk2,
  parameters = p,
  regimen = r2,
  only_obs = TRUE
)
ggplot(dat2, aes(x = t, y = y)) +
  geom_line()

Included models

PKPDsim includes definitions of several PK/PD models used in the InsightRX platform.

These models can be installed as R packages using the model_from_api() function.

model_from_api(
  system.file("models", "pk_vanco_anderson.json5", package = "PKPDsim"),
  to_package = TRUE,
  install_all = TRUE
)

Please note that these models may differ from those used in production by InsightRX. They are provided for demonstration purposes only, and should not be used for dosing patients.

Example usage

To simulate PK, provide the model and parameters to sim() along with a dosing regimen and covariates.

## Create dosing regimen and covariates
reg <- new_regimen(
  amt = 100,
  n = 3,
  interval = 12,
  type = "infusion",
  t_inf = 1
)

covs <- list(
  WT = new_covariate(4),
  PMA = new_covariate(42),
  CR = new_covariate(0.5),
  CL_HEMO = new_covariate(0)
)

## Perform simulation
sim(
  ode = pkvancoanderson::model(),
  parameters = pkvancoanderson::parameters(),
  regimen = reg,
  covariates = covs
)