Example Workflow with Certara.RDarwin and R postprocessing
RDarwin_NLME_RPostprocessing.Rmd
library(Certara.RDarwin)
# for tidy style
library(magrittr)
#> Warning: package 'magrittr' was built under R version 4.3.3Introduction
This vignette provides an illustrative workflow for using the
Certara.RDarwin package, a toolkit designed to facilitate
the usage of pyDarwin
with the Certara NLME pharmacometric modeling engine from the R command
line. This particular example demonstrates the step-by-step process of
creating, configuring, and executing a model search using RDarwin in
combination with the pyDarwin framework.
Data
The data set used in this example (pkOralBolus.csv) is
made available in the package examples folder, however, we will copy
them to our current working directory for ease of access. We will also
need to copy an R script to be used for postprocessing.
WorkingDir <- tempdir()
DataFilePath <- file.path(WorkingDir, "pkOralBolus.csv")
file.copy(system.file(package = "Certara.RDarwin", "examples", "pkOralBolus.csv"),
DataFilePath, overwrite = TRUE)
RScriptPath <- file.path(WorkingDir, "penaltyDiffObservedSimulatedCmax.R")
file.copy(system.file(package = "Certara.RDarwin", "examples", "penaltyDiffObservedSimulatedCmax.R"),
RScriptPath, overwrite = TRUE)Setting Up the Model
The first section of the workflow involves setting up the model. In this example, we will configure a first-order absorption model with parameterization set to “Clearance.” We specify several model characteristics, such as the number of compartments, absorption type, and parameterization. Additionally, we introduce parameters to the model and make modifications to their initial estimates. In this workflow, the following actions are taken:
-
Setting Up the Model: We use the
create_ModelPK()function to configure the model’s structures, including the number of compartments, absorption type, and parameterization.
modelPMLCodes <-
create_ModelPK(
CompartmentsNumber = c(1, 2),
Absorption = c("First-Order"),
Parameterization = "Clearance"
)-
Introducing Parameters: Next, we introduce a
parameter named “F” to the model. This parameter is associated with a
LogitNormal distribution, the initial estimates of its corresponding
fixed effect is set to -1, and its associated random effect is set to
“Searched.” This step is accomplished using the
add_StParmfunction.
modelPMLCodes <-
add_StParm(
modelPMLCodes,
StParmName = "F",
Type = "LogitNormal",
State = "Searched",
ThetaStParm = Theta(Name = "tvlogitF", InitialEstimates = -1),
OmegaStParm = Omega(Name = "nlogitF", State = "Searched"),
DosepointArgName = "bioavail"
)-
Modifying Initial Estimates: We modify the initial
estimates for a theta parameter (“tvV”) and an omega parameter (“nCl”)
using the
modify_Thetaandmodify_Omegafunctions, respectively.
# list Thetas available
list_Thetas(modelPMLCodes)
#> [1] "tvCl" "tvV" "tvKa" "tvlogitF" "tvCl2" "tvV2"
# list Omegas available
list_Omegas(modelPMLCodes)
#> [1] "nCl" "nV" "nKa" "nlogitF" "nCl2" "nV2"
# change initial estimates for Theta and Omega
modelPMLCodes <-
modify_Theta(
modelPMLCodes,
Name = "tvV",
InitialEstimates = c(0, 5, 1000)
) %>%
modify_Omega(
Name = "nCl",
InitialOmega = 5
)-
Residual Error Model Configuration: In this
example, we disable the proportional error model (which is enabled by
default) and enable an Additive+ Multiplicative error model using the
modify_Observationfunction.
# list observations available
list_Observations(modelPMLCodes)
#> [1] "CObs"
modelPMLCodes <-
modelPMLCodes |>
modify_Observation(
ObservationName = "CObs",
SigmasChosen =
Sigmas(
Proportional = 0,
AdditiveMultiplicative = list(PropPart = 0.1, AddPart = 10)
)
)- Search Zero-Order Absorption: We expand the search space by adding zero-order absorption on one-compartment models.
# add zero order absorption to one compartment model
modelPMLCodes <-
modelPMLCodes |>
add_StParm(
StParmName = "Rate",
Type = "LogNormal2",
State = "Searched",
ThetaStParm = Theta(Name = "tvlogRate", InitialEstimates = 0),
OmegaStParm = Omega(Name = "nlogRate", State = "Present"),
DosepointArgName = "rate",
PMLStructures = c("PK1FOC")
)
# see the resulted template
print(modelPMLCodes)
#> PK1FOC
#> test() {
#> cfMicro(A1, Cl / V, first = (Aa = Ka))
#> C = A1 / V
#> dosepoint(Aa{_F[1]}{_Rate[1]}, idosevar = AaDose, infdosevar = AaInfDose, infratevar = AaInfRate)
#> {_F[2]}
#> {_Rate[2]}
#> error(CEps = 10)
#> observe(CObs = C + CEps*sqrt(1 + C*C*(CMultStdev/sigma())^2))
#> stparm(CMultStdev = tvCMultStdev )
#> fixef(tvCMultStdev= c(, 0.1, ))
#>
#>
#> stparm(Cl = tvCl * exp( nCl ))
#> fixef(tvCl= c(, 1, ))
#> ranef(diag(nCl) = c(5))
#> stparm(V = tvV * exp( nV ))
#> fixef(tvV= c(0, 5, 1000))
#> ranef(diag(nV) = c(1))
#> stparm(Ka = tvKa * exp( nKa ))
#> fixef(tvKa= c(, 1, ))
#> ranef(diag(nKa) = c(1))
#>
#> }
#> PK2FOC
#> test() {
#> cfMicro(A1, Cl / V, Cl2 / V, Cl2 / V2, first = (Aa = Ka))
#> C = A1 / V
#> dosepoint(Aa{_F[1]}, idosevar = AaDose, infdosevar = AaInfDose, infratevar = AaInfRate)
#> {_F[2]}
#> error(CEps = 10)
#> observe(CObs = C + CEps*sqrt(1 + C*C*(CMultStdev/sigma())^2))
#> stparm(CMultStdev = tvCMultStdev )
#> fixef(tvCMultStdev= c(, 0.1, ))
#>
#>
#> stparm(Cl = tvCl * exp( nCl ))
#> fixef(tvCl= c(, 1, ))
#> ranef(diag(nCl) = c(5))
#> stparm(V = tvV * exp( nV ))
#> fixef(tvV= c(0, 5, 1000))
#> ranef(diag(nV) = c(1))
#> stparm(Cl2 = tvCl2 * exp( nCl2 ))
#> fixef(tvCl2= c(, 1, ))
#> ranef(diag(nCl2) = c(1))
#> stparm(V2 = tvV2 * exp( nV2 ))
#> fixef(tvV2= c(, 1, ))
#> ranef(diag(nV2) = c(1))
#> stparm(Ka = tvKa * exp( nKa ))
#> fixef(tvKa= c(, 1, ))
#> ranef(diag(nKa) = c(1))
#>
#> }Engine and Simulation Setups
The next step in our workflow involves setting up the engine and simulation parameters. In this example, we use the FOCE-ELS (First-Order Conditional Estimation with Interaction) method for optimization and specify simulation settings. These are the actions performed:
-
Engine Setup: We use the
specify_EngineParamsfunction to set up the engine parameters. In this case, we choose the FOCE-ELS method, configure standard error estimation, and set the number of iterations for Maximum A Posterior (MAP) initial Naive Pooling optimization.
engine_Setup <- specify_EngineParams(
method = "FOCE-ELS",
stdErr = "None",
numIterMAPNP = 3
)-
Simulation Setup: We specify the simulation
parameters using the
specify_SimParamsfunction. In this example, we configure the number of replicates and set a seed for reproducibility.
Sim_Setup <- specify_SimParams(
numReplicates = 10,
seed = 12345678
)-
Specifying Output Tables: The tables are specified
using the
Table()function. In this example, we introduce two output tables. The first table, “ObsData.csv” is generated after estimation step to get observations (CObs) from the source file. The second table, “SimData.csv”, is generated during simulation (by setting the argumentForSimulation = TRUE) and includes simulated data for CObs.
Generating Template and Token Files
Once the model and parameter setups are complete, we use the
write_ModelTemplateTokens function to generate the template
(template.txt) and token (tokens.json) files,
which define the model structure, its characteristics, and data mappings
- all of which are essential to the model search. This function also
allows the specification of omega
structures to be searched through the argument
OmegaSearchBlock.
# Generate template and token files
ModelTemplateTokens <-
write_ModelTemplateTokens(
TemplateFilePath = file.path(WorkingDir, "template.txt"),
TokensFilePath = file.path(WorkingDir, "tokens.json"),
Description = "search--123F-12Tlag-1Rate",
Author = "Certara",
DataFilePath = DataFilePath,
DataMapping = c(
id = "ID",
time = "time",
AMT = "Dose",
CObs = "CObs"
),
PMLParametersSets = modelPMLCodes,
EstArgs = engine_Setup,
SimArgs = Sim_Setup,
Tables = list(simple_table01, Sim_table01)
)Generating Options File
The next step in the workflow is the generation of the options file,
options.json. This file is crucial for configuring the
pyDarwin project according to specific requirements. The following
actions are taken:
We configure the options for the pyDarwin project using the
create_pyDarwinOptions function. This includes specifying
the algorithm, timeouts, directories, and post-processing settings. In
this example, we use post run R code, which is specified through the
postprocess argument.
It is assumed to have NLME Engine installed/unzipped and
INSTALLDIR environment variable pointed to the installation
directory (C:/Program Files/Certara/NLME_Engine on
Windows). Please modify it accordingly.
# relative paths are used for convenience
data_dir <- "../../../.."
working_dir <- "{project_dir}/Results"
output_dir <- "{project_dir}/Results/output"
temp_dir <- "{project_dir}/Results/temp"
optionSetup <- create_pyDarwinOptions(
algorithm = "GA",
model_run_timeout = 2400,
data_dir = data_dir,
working_dir = working_dir,
output_dir = output_dir,
temp_dir = temp_dir,
nlme_dir = Sys.getenv("INSTALLDIR"),
remove_temp_dir = FALSE,
postprocess =
pyDarwinOptionsPostprocess(
use_r = TRUE,
post_run_r_code = RScriptPath
)
)We use the write_pyDarwinOptions function to generate
the options.json file based on the setup created in the
previous step.
write_pyDarwinOptions(pyDarwinOptions = optionSetup,
file = file.path(WorkingDir, "options.json"))Postprocessing R Code Requirements
When using R code for postprocessing with pyDarwin, the R script must return a character vector of length 2, structured as follows:
Penalty Value: The first element of the vector is a string that can be interpreted as a numeric value, representing the penalty. This value is used directly in the pyDarwin model selection process, contributing to the fitness/reward value of a particular model configuration. The penalty can be any real number, including negative values, but is typically positive. This penalty influences the selection of subsequent populations during the evolutionary search.
Comment for Documentation: The second element of the returned vector is a string that is appended as a comment to the output file generated by pyDarwin. This comment is meant to document the output of the R code but has no effect on the pyDarwin process itself. The content of this element is entirely up to the user and can be used to provide additional context or information about the results of the postprocessing step.
The content of the second element (the comment) has no impact on pyDarwin’s model search, while the penalty value plays a significant role in guiding the evolutionary search process.
Running pyDarwin
With the template, token, and options files in place, we are ready to
execute the pyDarwin model search. In this final step, we run the search
using the run_pyDarwin function. The following actions are
performed:
- Specify Python Path: We specify the path to the Python interpreter executable. Please follow pyDarwin installation guide for instructions. Current row should be modified accordingly.
pythonPath <- "./venv/Scripts/python.exe"-
Run pyDarwin: We launch the pyDarwin search process
using the
run_pyDarwinfunction. This function monitors the search progress and, depending on the configuration, either waits for the search to complete or exits immediately.
job <- run_pyDarwin(InterpreterPath = pythonPath,
DirectoryPath = WorkingDir,
TemplatePath = "template.txt",
TokensPath = "tokens.json",
OptionsPath = "options.json",
Wait = TRUE) Running pyDarwin v3.0.0
Options file found at /tmp/RtmplOWzal/options.json
Preparing project working folder...
Preparing project output folder...
Preparing project temp folder...
Using darwin.MemoryModelCache
Writing intermediate output to /tmp/RtmplOWzal/Results/output/results.csv
Models will be saved in /tmp/RtmplOWzal/Results/models.json
Template file found at /tmp/RtmplOWzal/template.txt
Tokens file found at /tmp/RtmplOWzal/tokens.json
Algorithm: GA
Engine: NLME
random_seed: 11
Project dir: /tmp/RtmplOWzal
Data dir: ../../../..
Project working dir: /tmp/RtmplOWzal/Results
Project temp dir: /tmp/RtmplOWzal/Results/temp
Project output dir: /tmp/RtmplOWzal/Results/output
Key models dir: /tmp/RtmplOWzal/Results/key_models
Search space size: 16
Estimated number of models to run: 264
NLME found: /home/user/Desktop/installdir/
GCC found: /usr
Search start time: Wed Nov 1 15:59:04 2023
Starting generation 1
RScript found at /usr/lib/R/bin/Rscript
Post Run R code found at /tmp/RtmplOWzal/penaltyDiffObservedSimulatedCmax.R
Not using Post Run Python code
Checking files in /tmp/RtmplOWzal/Results/temp/1/1
Generation = 1, Model 2, Done, fitness = 7570.350, message = No important warnings
Generation = 1, Model 1, Done, fitness = 7615.266, message = No important warnings
Generation = 1, Model 3, Done, fitness = 7615.266, message = No important warnings
Generation = 1, Model 4, Done, fitness = 7406.655, message = No important warnings
Current generation best genome = [0, 1, 1, 0], best fitness = 7406.6551
Best overall fitness = 7406.655099, iteration 1, model 4
No change in fitness for 1 generations, best fitness = 7406.6551
Starting generation 2
Time elapsed: 1.4 min.
Estimated time remaining: 89 min.
Generation = 2, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 2, Model 2, Cache(1-2), fitness = , message = From NLME_1_2: No important warnings
Generation = 2, Model 3, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 2, Model 4, Cache(1-1), fitness = , message = From NLME_1_1: No important warnings
Starting downhill generation = 2 at Wed Nov 1 16:00:27 2023
current best model(s) =
generation 2, ind 1, fitness = 7406.655099211558
generation 2, ind 3, fitness = 7406.655099211558
code for niche (minimal binary) 1 = [0, 1, 1, 0], fitness = 7406.655099211558
Starting downhill step 1, total of 4 in 1 niches to be run.
Time elapsed: 1.4 min.
Estimated time remaining: 86 min.
Generation = 2D01, Model 1, Cache(1-1), fitness = , message = From NLME_1_1: No important warnings
Generation = 2D01, Model 3, Cache(1-2), fitness = , message = From NLME_1_2: No important warnings
Generation = 2D01, Model 2, Done, fitness = 8705.072, message = No important warnings
Generation = 2D01, Model 4, Done, fitness = 8843.169, message = No important warnings
Begin local exhaustive 2-bit search, generation = 2, step = 2
Model for local exhaustive search = 2, phenotype = ([('PML', 0), ('_F', 1), ('_Rate', 1), ('_nlogitF', 0)]) model Num = 1, fitness = 7406.655099211558
4 models in local exhaustive search, 1 bits
10 models in local exhaustive search, 2 bits
Time elapsed: 6.2 min.
Estimated time remaining: 3 h. 43 min.
Generation = 2S01, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 2S01, Model 3, Cache(1-3), fitness = , message = From NLME_1_3: No important warnings
Generation = 2S01, Model 5, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 2S01, Model 6, Done, fitness = 7549.176, message = No important warnings
Generation = 2S01, Model 8, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 2S01, Model 2, Done, fitness = 7562.779, message = No important warnings
Generation = 2S01, Model 10, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 2S01, Model 9, Done, fitness = 7564.889, message = No important warnings
Generation = 2S01, Model 4, Done, fitness = 7538.376, message = No important warnings
Generation = 2S01, Model 7, Done, fitness = 8705.072, message = No important warnings
Done with downhill step, 2. best fitness = 7406.655099211558
Current generation best genome = [0, 1, 1, 0], best fitness = 7406.6551
Best overall fitness = 7406.655099, iteration 1, model 4
No change in fitness for 2 generations, best fitness = 7406.6551
Starting generation 3
Time elapsed: 9.4 min.
Estimated time remaining: 2 h. 48 min.
Generation = 3, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 3, Model 2, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 3, Model 3, Cache(1-1), fitness = , message = From NLME_1_1: No important warnings
Generation = 3, Model 4, Clone(3), fitness = , message = From NLME_1_1: No important warnings
Current generation best genome = [0, 1, 1, 0], best fitness = 7406.6551
Best overall fitness = 7406.655099, iteration 1, model 4
No change in fitness for 3 generations, best fitness = 7406.6551
Starting generation 4
Time elapsed: 9.4 min.
Estimated time remaining: 2 h. 44 min.
Generation = 4, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 4, Model 2, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 4, Model 3, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 4, Model 4, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Starting downhill generation = 4 at Wed Nov 1 16:08:30 2023
current best model(s) =
generation 4, ind 1, fitness = 7406.655099211558
generation 4, ind 2, fitness = 7406.655099211558
code for niche (minimal binary) 1 = [0, 1, 1, 0], fitness = 7406.655099211558
Starting downhill step 1, total of 4 in 1 niches to be run.
Time elapsed: 9.4 min.
Estimated time remaining: 2 h. 37 min.
Generation = 4D01, Model 1, Cache(1-1), fitness = , message = From NLME_1_1: No important warnings
Generation = 4D01, Model 2, Cache(2D01-2), fitness = , message = From NLME_2D01_2: No important warnings
Generation = 4D01, Model 3, Cache(1-2), fitness = , message = From NLME_1_2: No important warnings
Generation = 4D01, Model 4, Cache(2D01-4), fitness = , message = From NLME_2D01_4: No important warnings
Begin local exhaustive 2-bit search, generation = 4, step = 2
Model for local exhaustive search = 4, phenotype = ([('PML', 0), ('_F', 1), ('_Rate', 1), ('_nlogitF', 0)]) model Num = 1, fitness = 7406.655099211558
4 models in local exhaustive search, 1 bits
10 models in local exhaustive search, 2 bits
Time elapsed: 9.4 min.
Estimated time remaining: 2 h. 06 min.
Generation = 4S01, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 4S01, Model 2, Cache(2S01-2), fitness = , message = From NLME_2S01_02: No important warnings
Generation = 4S01, Model 3, Cache(1-3), fitness = , message = From NLME_1_3: No important warnings
Generation = 4S01, Model 4, Cache(2S01-4), fitness = , message = From NLME_2S01_04: No important warnings
Generation = 4S01, Model 5, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 4S01, Model 6, Cache(2S01-6), fitness = , message = From NLME_2S01_06: No important warnings
Generation = 4S01, Model 7, Cache(2S01-7), fitness = , message = From NLME_2S01_07: No important warnings
Generation = 4S01, Model 8, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 4S01, Model 9, Cache(2S01-9), fitness = , message = From NLME_2S01_09: No important warnings
Generation = 4S01, Model 10, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Done with downhill step, 4. best fitness = 7406.655099211558
Current generation best genome = [0, 1, 1, 0], best fitness = 7406.6551
Best overall fitness = 7406.655099, iteration 1, model 4
No change in fitness for 4 generations, best fitness = 7406.6551
Starting generation 5
Time elapsed: 9.4 min.
Estimated time remaining: 109 min.
Generation = 5, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 5, Model 2, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 5, Model 3, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 5, Model 4, Cache(2S01-4), fitness = , message = From NLME_2S01_04: No important warnings
Current generation best genome = [0, 1, 1, 0], best fitness = 7406.6551
Best overall fitness = 7406.655099, iteration 1, model 4
No change in fitness for 5 generations, best fitness = 7406.6551
Starting generation 6
Time elapsed: 9.4 min.
Estimated time remaining: 106 min.
Generation = 6, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 6, Model 2, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 6, Model 3, Cache(2S01-9), fitness = , message = From NLME_2S01_09: No important warnings
Generation = 6, Model 4, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Starting downhill generation = 6 at Wed Nov 1 16:08:30 2023
current best model(s) =
generation 6, ind 1, fitness = 7406.655099211558
generation 6, ind 2, fitness = 7406.655099211558
code for niche (minimal binary) 1 = [0, 1, 1, 0], fitness = 7406.655099211558
Starting downhill step 1, total of 4 in 1 niches to be run.
Time elapsed: 9.4 min.
Estimated time remaining: 99 min.
Generation = 6D01, Model 1, Cache(1-1), fitness = , message = From NLME_1_1: No important warnings
Generation = 6D01, Model 2, Cache(2D01-2), fitness = , message = From NLME_2D01_2: No important warnings
Generation = 6D01, Model 3, Cache(1-2), fitness = , message = From NLME_1_2: No important warnings
Generation = 6D01, Model 4, Cache(2D01-4), fitness = , message = From NLME_2D01_4: No important warnings
Begin local exhaustive 2-bit search, generation = 6, step = 2
Model for local exhaustive search = 6, phenotype = ([('PML', 0), ('_F', 1), ('_Rate', 1), ('_nlogitF', 0)]) model Num = 1, fitness = 7406.655099211558
4 models in local exhaustive search, 1 bits
10 models in local exhaustive search, 2 bits
Time elapsed: 9.4 min.
Estimated time remaining: 68 min.
Generation = 6S01, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = 6S01, Model 2, Cache(2S01-2), fitness = , message = From NLME_2S01_02: No important warnings
Generation = 6S01, Model 3, Cache(1-3), fitness = , message = From NLME_1_3: No important warnings
Generation = 6S01, Model 4, Cache(2S01-4), fitness = , message = From NLME_2S01_04: No important warnings
Generation = 6S01, Model 5, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 6S01, Model 6, Cache(2S01-6), fitness = , message = From NLME_2S01_06: No important warnings
Generation = 6S01, Model 7, Cache(2S01-7), fitness = , message = From NLME_2S01_07: No important warnings
Generation = 6S01, Model 8, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = 6S01, Model 9, Cache(2S01-9), fitness = , message = From NLME_2S01_09: No important warnings
Generation = 6S01, Model 10, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Done with downhill step, 6. best fitness = 7406.655099211558
Current generation best genome = [0, 1, 1, 0], best fitness = 7406.6551
Best overall fitness = 7406.655099, iteration 1, model 4
No change in fitness for 6 generations, best fitness = 7406.6551
-- End of GA component at Wed Nov 1 16:08:30 2023 --
0] Best individual GA is [0, 1, 1, 0] with fitness of 7406.655099
code for niche (minimal binary) 1 = [0, 1, 1, 0], fitness = 7406.655099211558
Starting downhill step 1, total of 4 in 1 niches to be run.
Time elapsed: 9.4 min.
Estimated time remaining: 48 min.
Generation = FND01, Model 1, Cache(1-1), fitness = , message = From NLME_1_1: No important warnings
Generation = FND01, Model 2, Cache(2D01-2), fitness = , message = From NLME_2D01_2: No important warnings
Generation = FND01, Model 3, Cache(1-2), fitness = , message = From NLME_1_2: No important warnings
Generation = FND01, Model 4, Cache(2D01-4), fitness = , message = From NLME_2D01_4: No important warnings
Begin local exhaustive 2-bit search, generation = FN, step = 2
Model for local exhaustive search = 6, phenotype = ([('PML', 0), ('_F', 1), ('_Rate', 1), ('_nlogitF', 0)]) model Num = 1, fitness = 7406.655099211558
4 models in local exhaustive search, 1 bits
10 models in local exhaustive search, 2 bits
Time elapsed: 9.4 min.
Estimated time remaining: 17 min.
Generation = FNS01, Model 1, Cache(1-4), fitness = , message = From NLME_1_4: No important warnings
Generation = FNS01, Model 2, Cache(2S01-2), fitness = , message = From NLME_2S01_02: No important warnings
Generation = FNS01, Model 3, Cache(1-3), fitness = , message = From NLME_1_3: No important warnings
Generation = FNS01, Model 4, Cache(2S01-4), fitness = , message = From NLME_2S01_04: No important warnings
Generation = FNS01, Model 5, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = FNS01, Model 6, Cache(2S01-6), fitness = , message = From NLME_2S01_06: No important warnings
Generation = FNS01, Model 7, Cache(2S01-7), fitness = , message = From NLME_2S01_07: No important warnings
Generation = FNS01, Model 8, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Generation = FNS01, Model 9, Cache(2S01-9), fitness = , message = From NLME_2S01_09: No important warnings
Generation = FNS01, Model 10, Clone(1), fitness = , message = From NLME_1_4: No important warnings
Done with final downhill step. best fitness = 7406.655099211558
Final output from best model is in /tmp/RtmplOWzal/Results/output/FinalResultFile.txt
Number of considered models: 80
Number of models that were run during the search: 11
Number of unique models to best model: 4
Time to best model: 1.4 minutes
Best overall fitness: 7406.655099, iteration 1, model 4
Elapsed time: 9.4 minutes
Search end time: Wed Nov 1 16:08:30 2023
Key models:
Generation = 1, Model 4, Done, fitness = 7406.655, message = No important warningsThe results will be available as a list with elements including
results (a data frame with results of all runs),
FinalResultFile (textual file with the properties of final
model chosen) and FinalControlFile with the final metamodel
as a result of the search.
Conclusion
This vignette demonstrates a complete workflow for using
Certara.RDarwin to configure and execute a model search
with the pyDarwin framework. It highlights the importance of setting up
the model, configuring parameters, and creating the necessary files for
a successful model selection process. By following these steps, users
can effectively perform automated model selection using RDarwin and
pyDarwin.