Simulation Study: Feature Selection and Collider Bias in Causal Interpretation

This repository contains the code for the simulation study presented in On the construction of molecular signatures of lifestyle exposures, evaluating feature selection strategies under different latent variable scenarios.

📁 Repository Structure

simulation_signatures/
├── run_simulation.R                      # Main simulations (Scenarios 1-3)
├── R/
│   ├── analysis_fxns.R                   # Analysis functions
│   ├── generate_data.R                   # Data generation functions
│   ├── generate_figures.R                # Main figure generation
│   ├── generate_figures_supplementary.R  # Supplementary figure generation
│   ├── run_scenario_1.R                  # Script to run scenario 1
│   ├── run_scenario_2.R                  # Script to run scenario 2
│   ├── run_scenario_3.R                  # Script to run scenarios 3
│   ├── run_scenario_2.sh                 # Bash script to run scenario 2
│   ├── run_scenario_3.sh                 # Bash script to run scenario 3
├── config/
│   ├── scenario_1.R                      # Config file for Scenario 1 
│   ├── scenario_2.R                      # Config file for Scenario 2 
│   ├── scenario_3.R                      # Config file for Scenario 3 
├── renv/                                 # Reproducible R environment
├── renv.lock                             # Package versions used in the study
└── README.md                             # You're here

Clone the repository

If you have Git installed, run:

git clone https://github.com/IARCBiostat/SimulationSignatures/
cd SimulationSignatures

Alternatively, you can download the ZIP file directly from GitHub and unzip it.

Set up the R environment

To ensure the correct package versions, we recommend using the renv environment:

install.packages("renv")        # If not already installed
renv::restore()                 # Restores package versions from renv.lock

This will install the specific versions of all required packages as used in the original analysis.

Running Simulations

Simulations are designed to be run in parallel. The number of cores used is controlled via the NCore parameter defined in the corresponding config file (e.g., config/scenario_1.R).

Run main simulations (Scenarios 1-3)

Rscript run_simulation.R 

Scenario 2 and 3 (SLURM-based parallelization)

Scenario 2 and 3 is parallelized using SLURM job arrays rather than R's internal parallelization.

Each job processes one parameter combination.

Make sure the array size matches the number of parameter combinations:

nrow(AllParams)

Each job uses SLURM_ARRAY_TASK_ID to select its task.

Run supplementary simulations

Rscript run_supplementary.R config/supplementary_scenario_4.R

You can edit the NCore value in the config file to match the number of CPU cores you wish to allocate.

Output

Simulation results will be saved in the results/ directory as .rds files.

Generating Figures

Once the simulations are complete, you can generate the figures used in the paper and supplementary material:

Main paper figures

source("R/generate_figures.R")

Supplementary figures

source("R/generate_figures_supplementary.R")

Figures will be saved to a results/Figures/ or equivalent output directory specified in the figure generation script.

Questions?

For issues, bugs, or questions, feel free to open an issue or contact us.