Causality-based Therapy Recommendation (CTR)

A Python and R implementation of Causality-based Therapy Recommendation (CTR) method in paper "Causal Recommendation Method for Personalised Chemotherapy Optimisation in Breast Cancer".

Infrastructure used to run experiments:

• OS: Ubuntu 24.04.3 LTS (Windows Subsystem for Linux WSL2)
• CPU: Intel(R) Core(TM) i7-1255U @ 1.70GHz).
• RAM: 16 GB.

Dataset

We used observational data from two sources: the DUKE dataset (Saha et al., 2021) and the TransNEO dataset (Sammut et al., 2022, Earl et al., 2015).

Installation

Installation requirements for CTR:

• Python >= 3.12
• numpy 2.2.2
• pandas 2.2.3
• scikit-learn 1.6.1
• scipy 1.15.1
• matplotlib 3.10.0
• seaborn 0.13.2
• R-base = 4.3.3
  • causalTree
  • rpart

Detailed Guidelines for Environment Setup using Conda on Linux

1. Create a Conda Environment

Firstly, follow the link to install Conda

Create a new Conda environment named ctr_env with specific versions of Python and R: (bash)

conda create -n ctr_env python=3.12 r-base=4.3 -c conda-forge -y
conda activate ctr_env

2. Install Python Packages: Install essential Python packages using pip: (bash)

pip install numpy pandas scikit-learn scipy matplotlib seaborn

3. Install R packages:

Install the r-devtools package via Conda: (bash)

conda install r-devtools

Launch R within the environment: (bash)

R

In the R session, install the required R packages: (R script)

# Install the 'causalTree' package from GitHub
devtools::install_github("susanathey/causalTree")

# Install additional R packages from CRAN
install.packages(c("dplyr", "graph", "rpart"), repos = "https://cloud.r-project.org", dependencies = TRUE)

# Verify installed R packages
rownames(installed.packages())

Exit the R session: (R script)

q()

Reproducing the Paper Results

1. Run the CTR model with 2 datasets

Rscript run_CTR.R

2. Run 6 baselines with 2 datasets

python run_baselines.py

3. Generate Evaluation Results in the paper

python survival_analysis.py
python recovery_comparison.py