PLIF Validity

This repository contains the code and notebook to reproduce the PLIF analysis experiments in the paper "Assessing interaction recovery of predicted protein-ligand poses" [arXiv:2409.20227].

Structure

Code

The code to calculate the PLIF recovery rates can be found under the plif_utils folder:

• system_prep: for preparing the ligand and protein (binding pocket) objects that will be used for generating the PLIFs,
• file_prep: for loading files and running the above system preparation code,
• analysis: for generating the PLIFs and calculating the PLIF recovery rate from the prepared files,
• settings: for controlling some optional settings (interaction types investigated, number of minimisation steps, suffix of output files...etc.).

Data

The data folder only contains the identifiers from the PoseBusters benchmark study. We also provide the corresponding protein PDB files preprocessed with Spruce and ligand SDF, these can be downloaded on Zenodo.

After running the docking experiments, place the resulting poses in a data/${docking_method}/${posebusters_id}/ folder. More details on the expected file structure can be found in the plif_utils.file_prep.get_files function.

Setting up the environment

Install Python 3.11 in your virtual environment of choice and run the following command:

pip install -r requirements.txt

You will also need Jupyter notebook installed if you wish to run the notebook directly.

Running the analysis code

Open the notebooks/plif_analysis.ipynb file with Jupyter notebook and run all cells.

Citing us

@misc{errington2024assessinginteractionrecoverypredicted,
  title         = {Assessing interaction recovery of predicted protein-ligand poses},
  author        = {David Errington and Constantin Schneider and Cédric Bouysset and Frédéric A. Dreyer},
  year          = {2024},
  url           = {https://arxiv.org/abs/2409.20227},
  eprint        = {2409.20227},
  archiveprefix = {arXiv},
  primaryclass  = {q-bio.BM}
}