Pirana/PsN
Working with Pirana and PsN
Create a VPC in Pirana
Select a model, open the menu (right mouse click) and go to
'PsN' -> 'Model Diagnotics' -> 'VPC'.
vpc -samples=1000 -bin_by_count=0 -bin_array=2,4,6,8 -idv=TAD -predcorr run001.mod -directory=vpc001
- vpc: refers to the Perl VPC directory (sometimes the whole path is shown 'C:\strawberry\perl\bin\vpc').
- -samples = 1000: number of simulated datasets
- -bin_by_count = 0 of 1: binning based on the number of observations (1) or width of interval (0).
- -bin_array = 2,6,8,14: the number of observation (-bin_by_count=1) or boundries of the bins (-bin_by_count=0).
- -no_bins = 8: number of bins of approx. equal size.
- -idv = TAD: independend variable. In this case the TAD (time after dose). The default is TIME (total time from start of id).
- -predcorr: prediction corrected for simulation (e.g. due to differences in dosages).
- 001.mod: the model file that is used to create the VPC
- -directory = vpc001: the directory to run the VPC and store the results
- -stratify_on = BLOCK: stratify on specific categorical covariate (note that this has to be available in the table files)
- For other options and settings see 'vpc -h' or 'vpc -help'
Bootstrap
Select a model, open the menu (right click), go to 'PsN' -> 'Model diagnostics' -> 'Bootstrap'.
bootstrap -samples=1000 -threads=7 -dir=bs_001 001.mod
- bootstrap: refers to Perl boostrap code
- -samples: number of simulated datasets (usually about 500, 1000 or 5000)
- -threads = 4: processor (CPU) threads used for the bootstrap (meestal max -1)
- -dir = bs_001: directory for the bootstrap
- 001.mod: the model file
Stepwise Covariate Model-building (SCM)
To run a SCM the best way is to create a config file. This can be used for multiple models (with some minor tweaking)
For functionality and options refer to the scm guide or -help function. Note that the confif file has to be stored as .scm file.
See: PsN documentation
; In the model parameters should be defined as typical values, e.g.:
; TVCL = THETA(2)
; CL = TVCL*EXP(ETA(1))
model=008.mod
directory=scm_dir1
search_direction=both
p_forward=0.1
p_backward=0.05
continuous_covariates=WT,AGE,HT,BMI,BSA
categorical_covariates=GNDR,CYP1,CYP2
missing_data_token=-99
linearize=0
nm_version=default
retries=1
threads=6
tweak_inits=1
picky=0
[test_relations]
CL=WT,AGE,HT,BMI,BSA,GNDR,CYP1,CYP2
V=WT,AGE,HT,BMI,BSA,GNDR,CYP1,CYP2
[valid_states]
continuous=1,2,3
categorical=1,2
Select a model, open the menu (right click) and go to 'PsN' -> 'Covariates' -> 'SCM'.
scm -config_file=scm6.scm -model=011.mod -dir=scm6
R scripts for pirana
R-scripts can be added to pirana by putting the R file in the pirana 'scripts' directory. This directory can be found in the pirana installation location. This folder also contains a template. Pirana exports a file with a lot of information concerning the selected model. This can be loaded in R and used in many ways. The scripts can be run by selecting and double clicking the script (or right click for more options) Scripts can be opened within R by selecting model, right click script and select 'Open script in R':
Pirana scripts
Basic DV versus TIME plot
# Below an example of an R script which can be run from pirana to plot the DV and TIME from the SDTAB table produced by nonmem.
# It saves the plot in a pdf and opens the file from its location. Hence it can all be done without manually having to use R.
# Note that it makes use of the info sent by pirana ('#PIRANA_IN').
# Table naming in nonmem can be arbitrary, preferably use conventional naming.
library(tidyverse)
library(wrapr)
models <- #PIRANA_IN
model_names <- names(models)
model_1 <- models[[model_names[1]]]
for(i in 1:length(model_1$tables)){
if(startsWith(model_1$tables[i],'SD')){
SDTAB <- model_1$tables[i]
SDTAB <- read.table (SDTAB, skip=1, header=T)
}
else if (startsWith(model_1$tables[i],'PA')){
PATAB <- model_1$tables[i]
PATAB <- read.table (PATAB, skip=1, header=T)
}
else if (startsWith(model_1$tables[i],'CO')){
COTAB <- model_1$tables[i]
COTAB <- read.table (COTAB, skip=1, header=T)
}
else{
print('Not a table starting with SD, PA or CO')
}
}
##### Start rest of script with loaded data
SDTAB %>% filter(DV > 0) %.>% plot(.$DV~.$TIME)
## Create output file
if (!file.exists("pirana_temp")) {dir.create ("pirana_temp")}
filename <- "pirana_temp/test.pdf"
pdf (file = filename)
plot (x=tab$TIME, tab$DV, main=model_1$descr);
dev.off()
## open created file
cat (paste("OUTPUT: ", fname, sep=""))
if (file.exists(filename) && open_res) {
if (Sys.info()['sysname'] == 'Windows') { shell.exec(paste(getwd(),"/",filename,sep="")) } # windows
else if (Sys.info()['sysname'] == 'Darwin') { system(paste ("open ",filename, sep="")) } # mac
else { system(paste("xdg-open ", fname, sep=""), ignore.stdout=TRUE, ignore.stderr=TRUE, wait=FALSE) } # linux
}
quit()
VPC plot
library(xpose4)
models <- #PIRANA_IN
model_names <- names(models)
model_1 <- models[[model_names[1]]]
### VPC ###
vpcnam <- paste('/vpc_',model_names, sep='')
vpcdir <- getwd()
vpcdir <- paste(vpcdir, vpcnam, sep='')
setwd(vpcdir)
vpc <- xpose.VPC(vpc.info = "vpc_results.csv", vpctab = "vpctab",
main = list("Visual Predictive Check"), PI.ci = "area", PI.real= T, type ="p")
### Create output file ###
fname <- paste(names(models)[1],"_VPC.pdf", sep="")
pdf (file = fname)
print(vpc)
dev.off()
### Open created file ###
cat (paste("OUTPUT: ", fname, sep=""))
if (file.exists(fname) && open_res) {
if (Sys.info()['sysname'] == 'Windows') {
shell.exec(paste(getwd(),"/",fname,sep=""))
} # windows
else if (Sys.info()['sysname'] == 'Darwin'){
system(paste ("open ",fname, sep=""))
} # mac
else {
system(paste("xdg-open ", fname, sep=""), ignore.stdout=TRUE, ignore.stderr=TRUE, wait=FALSE)
} # linux
}
quit()