Here, we provide documentation and code for the de novo prediction of cross-reactive peptides using the 10LH:PHOX2B/HLA-A*24:02 complex as an example.
For a detailed description of the implementation, please refer to our manuscript: Sun, Florio, & Gupta et al., Structural principles of peptide-centric Chimeric Antigen Receptor recognition guide therapeutic expansion. bioRxiv doi.org/10.1101/2023.05.24.542108
- MacOS/Linux (tested on MacOS 12.6.3)
- Rosetta (tested on 2020.08)
- Anaconda (tested on 4.11.0)
- Python 3.8+ (tested on 3.8.15)
- platformdirs
- PyMOL
- tqdm
- Create a conda environment using the provided YAML file using the following command:
conda env create -f environment.yml
This step takes roughly 5 minutes on a normal desktop computer, depending on internet speed. - Obtain Rosetta via https://www.rosettacommons.org/software/license-and-download
- (Optional) Download the latest version of HLA3DB via https://hla3db.research.chop.edu/ and directly replace the
MHC_pdbs
folder. Note that this repository contains a static version of HLA3DB downloaded on April 28th, 2023.
- Activate the conda environment:
conda activate pc_car
- Run the prediction:
python predict_cr_peptides.py
. The script takes ~40 minutes to run on a normal desktop computer.