Individual and Population Simulations

knitr::opts_chunk$set(echo = TRUE, message=FALSE, eval=FALSE)


The workshop ( provides several examples of how to perform simulations in ubiquity. To make a copy of these scripts and other supporting files in the current working directory run the following:

fr = workshop_fetch(section="Simulation", overwrite=TRUE)

This should create the following scripts:

These rely on a PK model of mAbs in humans (Davda etal. mAbs, 6(4), 1094-1102). The contents of the system file for this model can be seen at the bottom (use ?system_new to see a list of the available system file examples). The first step in any analysis (simulation, estimation, etc) is building the system file creating the ubiquity model object (cfg).

system_new(file_name="system.txt", system_file="mab_pk", overwrite = TRUE)
cfg = build_system(system_file = "system.txt")
cfg_orig = cfg
save(cfg_orig, file="Simulation_cfg_orig.RData")

After building any system you can then create templates. To create a template script for running simulations use system_fetch_template and specify "Simulation" as the template argument:

system_fetch_template(cfg, template="Simulation")

This will create the file analysis_simulate.R in the working directory. This template should have common options and dosing information commented out. You simply need to uncomment them and run the simulation.

Simulating an Individual Response (analysis_single.r)

We'll begin by demonstrating how to simulate an indiviudal response to dosing. The system parameters for the default parameter set can be pulled out of this object:

cfg = build_system(system_file = "system.txt")
parameters = system_fetch_parameters(cfg)

To alter system parameters at the scripting level, you can just reassign the elements in parameters. For example to change the celarance to a value of .015 you could simply use: parameters$CL = 0.15.

Next different simulation options can be set. For example the following will set the duration of the simulation to three months $\left(3\ \mbox{months} \times 4\frac{\mbox{weeks}}{\mbox{month}}\times7\frac{\mbox{days}}{\mbox{week}}\right)$ in days:

cfg = system_set_option(cfg, group  = "simulation", 
                             option = "output_times", 

The system file is written to allow both IV (Cp) and SC (At) dosing. So we zero out any default dosing specified in the system file. The next line specifies the SC dosing we want to simulate.

cfg = system_zero_inputs(cfg)
cfg = system_set_bolus(cfg, state  = "At", 
                            times  = c(  0,  14,  28,  42 ), #  day
                            values = c(200, 200, 200, 200 )) #  mg

Next we run the simulation:

som = run_simulation_ubiquity(parameters, cfg)

The variable som is a list containing the mapped simulation output, and the time course is stored in the simout element. The first column (time) contains the simulation time in the units of the simulation, days in this case. Next there is a column for each state (At, Cc, Cp) and a column for each output (C_ng_ml, C_DOSE). Each system parameter is passed through the simulation into the output (F1, ... MW). This model has two covariates specified DOSE and WT. The initial value of these covariates is passed through as well as the values at each time point. For the covariate DOSE this is SIMINT_CVIC_DOSE and DOSE, respectively. Next secondary parameters are also provided (kel, ... kpc). Lastly each timescale specified in the system file is also passed through with a "ts." prefix:

save(som, file="Simulation_som_single.RData")
rhandsontable(som$simout, width=600, height=300)
p = ggplot() + 
    geom_line(data=som$simout, aes(x=ts.days, y=C_ng_ml), color="blue")  +
    xlab("Time (days)")+
    ylab("C (ng/ml) (units)")
p = gg_log10_yaxis(p, ylim_min=1e3, ylim_max=3e5)
p = prepare_figure("print", p)

Simulating Population Response From IIV (analysis_multiple.r)

Next we want to simulate the response of multiple subjects. This system has IIV specified in the following manner:

message(paste(system_view(cfg_orig, 'iiv'), collapse="\n"))

Different aspects of the Monte Carlo simulations can be specified. For example the following states that we want to simulate 20 subjects and the simulation is run using simulate_subjects:

cfg=system_set_option(cfg, group  = "stochastic",
                           option = "nsub",
                           value  = 20)

som  = simulate_subjects(parameters, cfg)
# Creating a reduced form of the simulation output to save it and use less disk space
tmpsom = som
som = list()
som$tcsummary = data.frame(o.C_ng_ml.mean  =  tmpsom$tcsummary$o.C_ng_ml.mean,
                           o.C_ng_ml.ub_ci =  tmpsom$tcsummary$o.C_ng_ml.ub_ci,
                           o.C_ng_ml.lb_ci =  tmpsom$tcsummary$o.C_ng_ml.lb_ci,
                           ts.days         =  tmpsom$tcsummary$ts.days)
save(som, file="Simulation_som_multiple.RData")

The output here, som, has a different structure than the output from an individual simulation. It is a list with the following elements

p = ggplot(som$tcsummary, aes(x=ts.days, y=o.C_ng_ml.mean)) +
                           alpha=0.6) +
           geom_line(linetype="solid", size=0.7, color="blue")  +
           geom_line(aes(x=ts.days, y=o.C_ng_ml.ub_ci), linetype="dashed", size=0.2, color="blue")  +
           geom_line(aes(x=ts.days, y=o.C_ng_ml.lb_ci), linetype="dashed", size=0.2, color="blue")  +
           xlab("Time (days)")+
           ylab("C (ng/ml) (units)")+

p     = gg_log10_yaxis(p    , ylim_min=1e3, ylim_max=3e5)
p     = prepare_figure("print", p    )

Simulating Population Response From File (analysis_multiple_file.r)

Subject information can be pulled from a data file. First we need to load the dataset using system_load_data, here the dataset is named SUBS.

cfg = system_load_data(cfg, 
                       dsname    = "SUBS", 
                       data_file = system.file("ubinc", "csv", "mab_pk_subjects.csv", 
                                               package = "ubiquity"))

The format of the dataset is shown below. There needs to be an column for the subject ID (SIMINT_ID) and the simulation time (SIMINT_TIME). Next the subject level parameters can be specified where the column headers correspond to the parameter names. If a parameter is not specified, the default value will be taken from the parameters input. In this case MW was not specified in the dataset, so it will be taken from parameters.

Optionally, covariates can also be specified by name as well. For time varying covariates, multiple records can be specified for the same subject (the parameter values for that subject should remain constant between records). In this example WT and SEX are covariates but only WT has been defined in the system file, so SEX will be ignored. The covariate DOSE is not defined so the default from the system file will be used.

SUBSCSV= read.csv(system.file("ubinc", "csv", "mab_pk_subjects.csv",  package = "ubiquity"))
rhandsontable( SUBSCSV, width=800, height=300)

With the dataset loaded, we need to link that subject file to the stochastic simulations using the dataset name (SUBS) and then we can tell the scripts how to handle sampling from the dataset:

cfg=system_set_option(cfg, group  = "stochastic",
                           option = "sub_file",
                           value  = "SUBS")

cfg=system_set_option(cfg, group  = "stochastic",
                           option = "sub_file_sample",
                           value  = "with replacement")

Then we just use simulate_subjects to run the simulation as before:

som  = simulate_subjects(parameters, cfg)


If you are using simulate_subjects to perform population simulations and have multiple cores on your computer you can utilize those cores by setting the following options in the "simulation" group:

cfg=system_set_option(cfg, group  = "simulation",
                           option = "parallel",    
                           value  = "multicore")

cfg=system_set_option(cfg, group  = "simulation",
                           option = "compute_cores", 
                           value  = detectCores() - 1)

Contents of system.txt

cat(readLines(system.file("ubinc", "systems","system-mab_pk.txt", package="ubiquity")), sep="\n")

Try the ubiquity package in your browser

Any scripts or data that you put into this service are public.

ubiquity documentation built on Nov. 2, 2023, 6:26 p.m.