ProTCR

This package provides an implementation of the inference pipeline of ProTCR.

Requirements

• Python3.9+
• ProFOLD2
• Dependencies

Installation

1. Clone this repository and cd into it.

git clone https://github.com/bigict/ProTCR.git
cd ProTCR
git submodule update --init --remote

2. Create conda an environment and activate it.

conda create -n tcr python=3.11
conda activate tcr

You need to install the ProFOLD2 dependencies

bash profold2/install_env.sh

And then

pip install -r requirements.txt 

Running ProTCR

1. Model Parameters

   The ProTCR model parameters are hosted on Hugging Face.

   huggingface-cli download --local-dir params bigict/ProTCR
   

2. Inference

Once you have installed ProTCR, you can test your setup using e.g. the following input csv file named tcr_pmhc_input.csv

Antigen,TCRB,TCRA,MHC_str
ALSKGVHFV,VSQHPSWVICKSGTSVKIECRSLDFQATTMFWYRQFPKQSLMLMATSNEGSKATYEQGVEKDKFLINHASLTLSTLTVTSAHPEDSSFYICSARGSSGRAEYTQYFGPGTRLTVLE,,SHSMRYFFTSVSRPGRGEPRFIAVGYVDDTQFVRFDSDAASQRMEPRAPWIEQEGPEYWDGETRKVKAHSQTHRVDLGTLRGYYNQSEAGSHTVQRMYGCDVGSDWRFLRGYHQYAYDGKDYIALKEDLRSWTAADMAAQTTKHKWEAAHVAEQLRAYLEGTCVEWLRRYLENGKETLQRTD

./predict.sh --output_dir [OUTPUT_DIR] tcr_pmhc_input.csv

You can run

./predict.sh --help

for further help.

Known Issues

Please create an issue if it is not already listed in the issues tracker