Optimising covariate allocation at design stage using Fisher Information Matrix for Non-Linear Mixed Effects Models in pharmacometrics

  
  This repository contains the code and data used in the analysis for the publication "Optimising covariate allocation at design stage using Fisher Information Matrix for Non-Linear Mixed Effects Models in pharmacometrics"

  
Main Script

  
 **`run_all.R`**: Master script to run the entire analysis pipeline.

  
Folder Structure
  

  - **`/my_pmxcopula`**  
  Source code for the R package `pmxcopula` [@pmxcopula], used for copula fit diagnostic plots. This version includes minor edits. Original source: vanhasseltlab/pmxcopula.

- **`/PFIM_6_1_beta_cov`**  
  Contains `PFIM6.1_beta_cov` [@fayette_2024_13692989], an adapted version of the `PFIM6.1` R package [@pfim6], available at Zenodo.

- **`/Concentrations_covariates`**  
  Includes figures showing the evolution of concentration \( f(t) \) with respect to fixed and covariate effects.

- **`/HepaticFunction`**  
  Results and scripts for the Hepatic Function (HF) example.

- **`/RenalFunction`**  
  Results and scripts for the Renal Function (RF) example.

- **`/NHANES`**  
  Contains NHANES datasets (2009–2020) [@NHANES20092020] and the transformed covariate dataset used in the analysis.

- **`/tikzDictionary`**  
  TikZ dictionary for LaTeX-based plotting.

  
R Scripts

  
Data Preparation & Plotting
 

- `00_NHANES_create_database_RF_HF.R`: Creates the transformed covariate database from NHANES data.
- `01_Concentrations_covariates.R`: Plots concentration evolution figures.

Utility Functions

- `funct.R`: Loads required libraries and general functions.
- `funct_diag_copula.R`: Diagnostic functions for copula fits.
- `funct_Plot_Latex.R`: Plotting functions using `tikzDevice` for LaTeX output.
- `funct_ProjectedGradient.R`: Functions for the Projected Gradient Descent (PGD) algorithm.
- `funct_resOptim.R`: Functions to extract optimisation results.

 Hepatic Function (HF) Example

- `HF_02_fit_copula_pooled_vine.R`: Fits copulas and generates diagnostics.
- `HF_03_PFIM_GQ.R`: Computes FIM using Gauss-Legendre Quadrature.
- `HF_03_PFIM_MC.R`: Computes FIM using Monte Carlo methods.
- `HF_04_Cov_Opti.R`: Optimises covariate distribution.

Renal Function (RF) Example

- `RF_02_fit_copula_pooled_vine.R`: Fits copulas and generates diagnostics.
- `RF_03_PFIM_GQ_varying_beta.R`: Computes FIM using Gauss-Legendre Quadrature with varying beta.
- `RF_03_PFIM_MC.R`: Computes FIM using Monte Carlo methods.
- `RF_04_Cov_Opti_varying_beta.R`: Optimises covariate distribution.

- `Supp_RF_03_PFIM_baseModel.R`: Computes FIM for the model without covariates.
- `Supp_RF_BMI4_02_fit_copula_pooled_vine.R`: Sensitivity analysis with 4 BMI classes, fits copulas and generates diagnostics. 
- `Supp_RF_BMI4_03_PFIM_GQ_varying_beta.R`: Sensitivity analysis with 4 BMI classes, computes FIM using Gauss-Legendre Quadrature with varying beta. 
- `Supp_RF_BMI4_04_Cov_Opti_varying_beta.R`: Sensitivity analysis with 4 BMI classes, optimises covariate distribution.