In IntiQuan/populationIsoboles: Calculate isoboles efficiently

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(populationIsoboles)
library(ggplot2)
library(IQRtools)
.outputFolder <- "Output"

Introduction

An R-script version of this vignette and the resources are available in the folder system.file("examples/Example-2", package = "populationIsoboles")

This vignette demonstrates how to use the fast isobole algorithm fastIsoboles in the context of population simulation of NLME models. The underlying PKPD model describes malaria parasite clearance by two drugs acting jointly. Implementing the model is less explicit than in the other vignette, where each step is outlined how to get from the individual simulation to the population isobole. With IQRtools out-of-the-box population simulation functionalities, one just needs to define the objective function which evaluates success rates for different doses, given a parameter set. More information on IQRtools is available at

The population isoboles simulation uses populationIsoboles::runPopulationIsoboles() to calculate population isoboles. Finally, three different visualizations are shown.

Load Model and Parameter information

# system.file("examples/Example-2", package = "populationIsoboles")

# .. Model -----
model <- IQRmodel(system.file(file.path("examples","Example-2","Resources","Model.txt"), package = "populationIsoboles"))

# .. Parameter information -----
parameterGPF <- load_GPF(system.file(file.path("examples","Example-2","Resources","Parameters.xlsx"), package = "populationIsoboles"))

# .. Show the basic PKPD model for one parameter -----
dosing <- IQRdosing(c(0,0), c(1,2), c(500,500))
pars <- sampleIndParamValues(parameterGPF)$typicalIndParamValues
pars <- pars[setdiff(names(pars), c("ID", "ID.POP"))]
pars <- cbind(pars, PLbase = 9.)
simtime    <- seq(0,24*28,24)
sim <- sim_IQRmodel(model, simtime = simtime,
                    parameters = pars,
                    dosingTable = dosing, FLAGoutputsOnly = TRUE)
plot(sim) +
  geom_hline(yintercept = log(10), linetype = 2, color = "blue") +
  annotate("text", 600, 3, label = "LLOQ", color = "blue") +
  labs(x = "Time [h]", y  = "log Parasites / mL") +
  guides(color = FALSE)

Define functions to simulate populations

#' Simulate model and determin cure rate
#'
#' @param AMTx1,AMTx2 Vector of doses
#' @param parsIndiv output from [IQRtools::sampleIndParamValues()]
#'
#' @return numeric(length(AMTx1)) with cure rates
objectiveFunction <- function(AMTx1,AMTx2, parsIndiv) {
  pars <- parsIndiv$indParamValues
  pars <- pars[setdiff(names(pars), "ID.POP")]
  # Optimal opportunity to parallelize simulations
  cureRates <- lapply(
    X = seq_along(AMTx1),FUN = function(i) {
      dosing <- IQRdosing(rep(0, length(AMTx1[i])*2),
                          ADM = rep(c(1,2), each = length(AMTx1[i])),
                          AMT = c(AMTx1[i], AMTx2[i]))
      eventTable <- create_IQReventTable(dosing, pars)
      sim <- sim_IQRmodel(model,simtime = seq(0,24*28,24),
                          eventTable = eventTable,
                          FLAGoutputsOnly = TRUE)
      PLfinal <- sim[sim$TIME==24*28, "OUTPUT1"]
      mean(PLfinal < log(10)) # Parasitemia less than 10parasites/mL
    })
  unlist(cureRates)
}

# Illustrate ObjectiveFunction
parsIndiv <- sampleIndParamValues(parameterGPF, NULL, 10, Npop = 1)
AMTx1 <- AMTx2 <- c(100,500)
objectiveFunction(AMTx1, AMTx2, parsIndiv)

Run population isoboles

set.seed(1234)
# Nsubj and Npop Values should be higher to appropriately
# sample the parameter space. Needs powerful computer though
Npop <- 3
Nsubj <- 20
objectiveValue <- 0.95 # Target cure rate in each population
runPopulationIsoboles(
  Npop                  = Npop,
  Nsubj                 = Nsubj,
  gridInfo              = list(AMTx1Max = 5000, AMTx2Max = 5000),
  objectiveValue        = objectiveValue,
  objectiveFunction     = objectiveFunction,
  sampleIndividuals     = sampleIndParamValues,
  argsSampleIndividuals = list(spec = parameterGPF, Nsamples = Nsubj),
  FLAGverbose           = TRUE,
  .outputFolder         = .outputFolder
)

Plot results

Spaghetti plot of isoboles

plotSpaghetti(file.path(.outputFolder, "Simulations"))

Spaghetti with confidence levels overlaid

plotQuantiles(file.path(.outputFolder, "Simulations"))

Confidence level response surface

plotCLRS(file.path(.outputFolder, "Simulations"))

Clean up output folder

unlink(.outputFolder, recursive = TRUE)

IntiQuan/populationIsoboles documentation built on Jan. 13, 2022, 8:29 p.m.