This guide provides step-by-step instructions for setting up Python environments, running simulations, and analyzing results for the paper Back to the Future of Lead Optimization: Benchmarking Compound Prioritization Strategies.

The code has been tested using Rocky Linux release 8.8 (Green Obsidian).

Follow these instructions to set up the required Python environments using Conda:

conda create -n RetroDMTA python=3.11 -y
conda activate RetroDMTA
pip install --no-deps -r RetroDMTA.txt

Note: TMAP requires Python ≤ 3.9.

conda create -n RetroDMTA_TMAP python=3.9 -y
conda activate RetroDMTA_TMAP
pip install --no-deps -r TMAP.txt

Run the initialization script to set up necessary folders and files. Replace $DATASET with your actual dataset name:

./scripts/0_initialize.sh $DATASET

After running, the folder structure will be created. Place your dataset and blueprint files accordingly:

RetroDMTA/
├── app/
├── code/
│   └── $DATASET/
├── config/
├── data/
│   └── $DATASET/
│       ├── blueprint.csv
│       ├── data_aggregated.csv
│       └── data.csv
├── experiments/
├── figures/
├── scripts/
├── src/
└── README.md

Manually run the parameters notebook (./code/$DATASET/0_2_simulation_parameters.ipynb) to set dataset parameters (initial date, final date, timestep).

Parameters are auto-calculated but can be manually adjusted and saved in ./data/common/dataset_config.json.

Run preprocessing to generate necessary files and figures:

./scripts/1_preprocessing.sh $DATASET

This script will automatically:

• Define best molecules (saved in ./data/$DATASET/top_molecules.pkl).
• Compute similarity and distance matrices (saved in ./data/similarity_matrix.parquet and ./data/$DATASET/distance_matrix.parquet).
• Compute the Matched Molecular Pairs (saved in ./data/$DATASET/mmp.parquet).
• Generate all exploratory data analysis figures (saved in ./figures/$DATASET/).

Manually run the ./code/$DATASET/0_3_tmap_manual.ipynb notebook to:

• Determine the TMAP coordinates (saved in ./data/$DATASET/tmap_coordinates.pkl).
• Generate the interactive TMAP (saved in ./data/$DATASET/tmap.html).

Create a simulation config file in ./config/. Look at ./config/template.json for an example file. This file will be noted $PATH_TO_SIMULATION_CONFIG.

Execute the run script to launch the simulation:

./scripts/2_simulation.sh $PATH_TO_SIMULATION_CONFIG

This will launch the simulations and store the results in ./experiments/$DATASET/

Execute the postprocessing script:

./scripts/3_postprocessing.sh $DATASET

This automatically:

• Compute exploitation metrics (saved in ./data/$DATASET/exploitation.parquet).
• Compute exploration metrics (saved in ./data/$DATASET/exploration_*.parquet).
• Generate all TMAPs figures (saved in ./figures/$DATASET/).

To visualize results, you can launch a local app with the following command be sure to have the correct conda env enabled:

./scripts/4_app.sh